Tech Trends 2026 PDF Free Download

Name: Tech Trends 2026 PDF
Author: Rachel N. Garcia

1 / 67

0 views•67 pages

Tech Trends 2026 PDF Free Download

Tech Trends 2026 PDF free Download. Think more deeply and widely.

MAKE THE

LEAP FROM

NOW TO

NEXT!

TECH TRENDS

DISRUPTION DEEPENS. OPPORTUNITY WIDENS. EIGHT TECH TRENDS IN 2026.

2026

2 3

What will the world look like in 2030?

Answering such a question seems

impossible as most wonder what

tomorrow looks like. But sometimes it's

useful to consider what comes later before

we guess what comes next.

TECH TRENDS 2026

Info-Tech Research Group

4 5

What will the

world look like

in 2030? An-

swering such a

question seems impossible as

most wonder what tomorrow

looks like. But sometimes it’s

useful to consider what comes

later before we guess what

comes next.

The world has never been

more uncertain about the fu-

ture. It’s measured: the World

Uncertainty Index (WUI) has

tracked the percent of the

word “uncertain” in Economist

Intelligence Unit country re-

ports since 2008. Since the start

of 2025, it’s climbed 481% to

surpass 84,000. It’s higher than

it was during its previous peak

during the COVID-19 pandemic

(57,000) and a far cry from the

low of 9,000 in July 2008.

There’s little doubt about

the singular driver of that un-

certainty: tariffs. A new US

economic policy regime has

quickly transformed long-

held free trade corridors into

fee-laden channels. Beyond

tariffs, there is increasing

conflict around the world

and a mounting rift between

the east and the west. After

decades of crafting supply

chains in a relatively stable

global order, organizations

are now contemplating de-

globalization and financial

market volatility.

Our “deglobalization un-

certainty” theme will exam-

ine how these recent geopoliti-

cal shifts have sent shockwaves

through organizations, dis-

rupting their supply chains

and calling into question the

approach organizations take

to manage risk.

At the same time, emerging

technology is disrupting orga-

nizations faster than ever. AI

is leading the way. Last year,

Info-Tech predicted that AI

would make the jump from

an emerging technology to a

transformative technology on

our index. This indicates that

most organizations are already

invested in the technology and

that investment continues to

grow. Our Future of IT 2026

VXUYH\ GDWD FRQÀUPV WKH SUH-

diction was accurate, with our

current investment index for

AI or machine learning climb-

ing from -3 to 64, with a growth

rate of 80. Generative AI is the

most popular variety of AI for

current investments, with the

newer agentic AI showing fast

growth as an emerging tech-

nology.

Our “guided intelligent au-

tonomy” theme will examine

trends tracking the evolution

of AI as it becomes entrenched

in the enterprise tech stack

and is given more agency to

autonomize processes. We’ll

examine the opportunities to

restructure business models

and the risks that could foil the

promise of autonomous and

abundant intelligence.

To bolster their organiza-

tions against exponential un-

certainty, IT must harness the

new capabilities of emerging

technology. But an IT depart-

ment’s ability to deliver on

creating value from emerging

technologies is dependent on

its maturity level. Only about

one-quarter of IT departments

identify themselves as inno-

vators, but out of that group,

 VD\ WKH\ DUH FRQÀGHQW ,7

can create exponential value

from emerging technology. A

bit more than half of IT depart-

ments describe themselves

as average – either as trusted

operators or as business part-

ners – and of that group, 35%

DUHFRQÀGHQWWKH\FDQGHOLYHU

exponential value.

Our “Exponential IT” theme

will examine the reshaping of

IT’s role in the face of uncer-

tainty and rapid technological

transformation. It paints a pic-

ture of an IT function that’s an

integrated enabler of innovative

capabilities, orchestrating plat-

forms that are purpose-crafted

to organizational strategy.

That vision might seem

like a long journey from the

uncertainty of now. But to get

from now to next, you must

consider what comes later.

The world in 2030 looks very different

from the world in 2026.

That vision might seem

like a long journey from

the uncertainty of now.

But to get from now to

next, you must consider

what comes later.

TECH TRENDS 2026

Info-Tech Research Group

6 7

TechTrends

2026

From global market to multipolar

uncertainty

RESILIENT SUPPLY CHAIN SOURCING /16

6KLIWLQJIURPORZIULFWLRQJOREDOVRXUFLQJWRDGDSWDEOHGLYHUVLÀHGDQGUHOLDEOHVXSSO\IRUPDWHULDOV

and technology.

INTEGRATED ORGANIZATIONAL RESILIENCE /30

Moving beyond IT risk management to an integrated enterprise approach that proactively

responds to risk.

From digital tools to guided

intelligent autonomy

MULTI-AGENT ORCHESTRATION /44

Evolving from individual task-based agents to coordinated ecosystems of agents in pursuit of a

shared goal.

SMART SENSING NETWORKS /56

IoT is becoming more sophisticated with the convergence of advanced sensors and edge AI enabling

real-time autonomy.

AI AS ADVERSARY AND ALLY /68

AI is already escalating the cyber arms race between criminals and organizations, augmenting both

attacks and defenses. But organizations must consider whether AI itself is a threat.

From back-office operator to

Exponential IT

FEDERATED DATA GOVERNANCE /82

Overcoming poor data quality with a domain-based decentralized ownership architecture that can

be automated.

PURPOSE-BUILT PLATFORMS /96

$SSURDFKLQJPRGHUQLQIUDVWUXFWXUHLQDZD\WKDW·VH[SOLFLWO\WDLORUHGIRUVSHFLÀFEXVLQHVVJRDOV

rather than relying on commodity solutions.

SERVICE AS SOFTWARE /110

Instead of paying for access to cloud-delivered software, enterprises are paying for software that

delivers outcomes through AI automation and integration.

MAKE THE

LEAP FROM

NOW TO

NEXT!

TECH TRENDS 2026

Info-Tech Research Group

8 9

Last year we forecast the breakthrough of

AI technology from our emerging quad-

rant into the transformative quadrant,

and we were right. Overall, AI ranks close

to cloud computing and cybersecurity

solutions as a technology that most organizations

have now invested in and continue to invest in.

We broke down the AI category this year to examine

more details about investment trends. Generative

AI is the fastest growing area of AI investment and

is about equal with traditional AI for current levels

of investment. Agentic AI, a newer variety, lags be-

hind in terms of adoption compared to others. But

if you compare where agentic AI is debuting on this

index to generative AI just three years ago, it’s start-

ing off with a much higher rate of adoption and is

positioned for rapid growth. Physical AI – or AI that

allows self-driving cars and autonomous robots to

interact in the world – is more nascent, starting in

the niche quadrant.

• Quantum computing saw the largest year-over-

year change in growth. It remains in the niche

quadrant, but it shows a lot of momentum with

both current investment and investment intent.

• %XVLQHVV SURFHVV DXWRPDWLRQ DOVR VKRZHG VLJQLÀ-

cant growth, perhaps in part driven by AI’s contri-

EXWLRQWRWKLVÀHOG

• 0DQ\PRUHÀUPVDUHLQYHVWLQJLQKDUGZDUHWRDF-

celerate AI training or inference, with four in ten

organizations saying they’re invested already and

another four in ten saying they intend to invest in

the future.

OTHER NOTABLE SHIFTS ON THE INDEX:

Technology Investment Index

AI GRADUATES TO TRANSFORMATIONAL TECHNOLOGY

…if you compare where agentic AI is

debuting on this index to generative AI

just three years ago, it’s starting off with

a much higher rate of adoption and is

positioned for rapid growth.

10 11

TECH TRENDS 2026

Info-Tech Research Group

01 05

1818

-20

100

-40

-60

-80

-100

-100 100-80 80-60 60-40 40-20 200

1414

CURRENT INVESTMENT

GROWTH RATE

0606

TECHNOLOGIES CURRENT INVESTMENT GROWTH RATE

64%

49%

85%

-16%

80%

71%

77%

45%

34%

37%

72%

40%

-21%

-7%

-2%

67%

69%

12%

-40%

80%

70%

77%

-4%

83%

76%

75%

57%

50%

11%

41%

22%

-3%

12%

35%

63%

78%

65%

-7%

AI or Machine Learning

Business Process Automation (RPA/IPA)

Cybersecurity Solutions

Blockchain

Cloud Computing

Data Management Solutions

Integration Technologies (APIs)

No-Code / Low-Code Platforms

CI/CD Tools

Private Cellular (LTE or 5G)

On-Premises Servers / Compute

IoT

Robotics / Drones

Quantum Computing

Mixed Reality (AR/VR)

Hardware to Accelerate AI Training/Inference

Traditional AI

Generative AI

Agentic AI

Physical AI (self-driving cars, robots)

TECHNOLOGY INVESTMENT INDEX 2026 (N=525)

12 13

TECH TRENDS 2026

Info-Tech Research Group

Please estimate the total head count of your entire organization. (n=690)

What best describes your current level of IT maturity? (n=525)

In which country or region is your organization’s primary headquarters? (n=690) Count %

256

124

112

166

150

139

364

206

37.10%

18.00%

16.20%

13.20%

12.90%

31.60%

28.60%

26.50%

12.20%

1.10%

52.80%

29.90%

7.80%

1.90%

1.60%

251-1,000

1,001-2,500

50-250

2,501-5000

More than 5,000

IT is a business/organization partner

IT is a trusted operator

IT is an innovator

IT acts as firefighter

IT is unstable

United States

Canada

Australia

Africa

Asia

Info-Tech’s Tech Trends 2026 report is based on the results of its Future of IT 2026 survey,

conducted in May and June 2025. The online survey received 738 responses from IT decision-

makers. Each data point included in the report will specify the sample size received for the

VSHFLÀFTXHVWLRQRUUHVSRQGHQWJURXS([SHUWLQWHUYLHZVZHUHDOVRFRQGXFWHGEHWZHHQ0D\DQG

-XO\ DQGSURYLGHDGGLWLRQDOFRQWH[WWRWKHWUHQGVDVZHOO DVVSHFLÀFFDVHVWXG\H[DPSOHVRI

how organizations are responding to them. View the expert contributors section to see a complete

list of external contributors. In addition, the Future of IT survey and Tech Trends 2026 report were

developed through discussions with many Info-Tech research advisors, practice leads, executives,

ZRUNVKRSIDFLOLWDWRUVDQGH[HFXWLYHFRXQVHORUV)XUWKHUÀUPRJUDSKLFFRQWH[WRQWKH)XWXUHRI,7

2025 survey results provided here.

METHODOLOGY Count %

Count %

14 15

TECH TRENDS 2026

Info-Tech Research Group

Choice Count %

12.0

9.7

9.5

8.5

8.2

7.8

6.5

6.1

5.5

4.8

3.8

3.5

3.0

2.5

2.3

1.6

1.5

1.2

0.6

0.4

0.3

Government (State)

Financial Services

Professional & Technology Services

Government (Local/Municipal/County)

Manufacturing (Durable Goods)

Education (Higher Ed.)

Other

Education (K-12)

Insurance

Manufacturing (Non-Durable Goods)

Government (Federal)

Utilities

Healthcare Delivery

Government (Provincial)

Professional Associations & Non-Profits

Transportation & Logistics

Oil & Gas Operations

Retail

Casinos, Gambling & Lottery

Construction

Hotels, Resorts & Hospitality

Sports Entertainment

Healthcare Insurance

Throughout the report, we’ll compare

how innovators responded on the survey

to an “Average” group.

In this case,

we’ll consider a combination

of the respondents in the “trusted operator”

and “business/organization partner”

categories to be “Average.”

WHAT IS YOUR ORGANIZATION’S PRIMARY INDUSTRY? (N=690)

TECH TRENDS 2026

Info-Tech Research Group

16 17

Resilient

Supply Chain

Sourcing

New risks caused by geopolitical shifts and

increasingly powerful large corporations are creating

drivers for organizations to shift away from a globally

sourced supply chain. Where sourcing was primarily

motivated by price sensitivity before, there’s now a

shift toward resilience.

TECH TRENDS 2026

Info-Tech Research Group

18 19

Where availabili-

ty and low costs

were the dom-

inant consid-

erations for both the physical

and digital supply chains pre-

viously, organizations are now

considering a broader geopo-

OLWLFDOSLFWXUHDQGWKHUDPLÀFD-

tions of vendor entrenchment.

Pressures across multiple

fronts are threatening to crack

open the protective structure

of existing supply chains:

TARI FF S

US-introduced tariff policies

could increase hardware costs

between 9% and 45% if they’re

in place for the long term (IDC,

2025). Manufacturers pass on

the cost of tariffs to custom-

ers and impact IT budgets. The

uncertainty of what tariffs

are in place and whether they

will remain in place long term

also creates unpredictability

around quotes for procure-

ment. With the global supply

of semiconductors dependent

on Taiwan’s production, avoid-

ing cost increases from tariffs

is unlikely.

GEOPOLITICAL TENSIONS

Aside from tariffs, rising ten-

sions between different re-

gions is driving a bigger wedge

in trade connections. Supply

chains are becoming more re-

oriented around shared values

and security interests. With ICT

VSHFLÀFDOO\ZHVWHUQFRXQWULHV

are blocking Chinese-made

manufacturers from partici-

pating in infrastructure proj-

ects. For example, the US

Federal Communications Com-

mission designated both Hua-

wei and ZTE as national securi-

ty threats due to close ties with

the Chinese Communist Par-

ty (FCC, 2020). More recently,

TikTok has been the subject of

regulatory scrutiny in the US

and Canada because of its Chi-

nese ownership and the mobile

app’s collection of sensitive

data (uOttawa, 2025).

VENDOR RISKS

Technology vendor contracts

have tended toward a sub-

scription model that allows

enterprises to leverage their

operating budgets instead of

making capital investments,

especially since the cloud com-

puting era began. Negotiating

with vendors on a price per

user per month is the standard,

but enterprises risk becom-

ing locked into platforms and

being prone to steep vendor

price hikes. Further, enterpris-

es have seen the risk of “supply

chain attacks” on very large

vendors that cater to millions

of customers. Even internal

errors can result in outages

for clients with many negative

downstream effects.

Enterprise resource and supply chains

were put under strain during the

pandemic, but 2025 added new and

different strains to contend with.

To what degree organizations can bring their supply

chains closer to home or alter vendor relationships is

uncertain. But the need for supply chain management

that’s more holistic, proactive, and adaptive is clear.

DISRUPTION FROM

EMERGING TECHNOLOGY

Dependence on large vendors for technology

platforms is creating risks for organizations

SURQH WR WKH F\EHUDWWDFNV DQG VHOILQÁLFWHG

outages of their third-party partners. A recent

example was a major cloud service outage with

Google Cloud on June 12, 2025, that spiraled

out to affect internet services including Cloud-

ÁDUH 6SRWLI\ 7ZLWFK 6QDSFKDW DQG 'LVFRUG

(The Verge, 2025). AI-driven attacks are also

EHFRPLQJ PRUH FRPPRQSODFH DQG PRUH GLIÀ-

cult to detect, from constructing fake vendor

emails to deepfake voicemails that sound like

a company executive (Supply & Demand Chain

Executive, 2025).

INCREASED REGULATORY COMPLEXITY

There’s more regulatory variation in different

jurisdictions around the world, in particular

around environmental, social, and governance

(ESG) considerations. Manufacturers are be-

ing held accountable for their product’s entire

lifecycle in some regions. Other regions require

detailed carbon emissions reporting, including

“scope 3” emissions from their supply chain. A

new wave of human rights due diligence laws in

the EU and the US put pressure on companies to

prove no forced labor is involved in their supply

chains (Inspectorio, 2025).

TECH TRENDS 2026

Info-Tech Research Group

20 21

Despite a year of upheaval in trade due

to new US-introduced tariffs and other

geopolitical turmoil, overall IT organi-

zations are not more concerned about

disruption from government-enacted regulato-

ry changes. This suggests a disconnect with busi-

ness executives, as 95% cite tariffs as a primary

disruptor, leading to shifts in sourcing (Inspec-

torio, 2025).

In fact, on average, organizations rated that

disruption factor as slightly lower (3.23) com-

pared to last year’s survey (3.35). Government

regulations were considered less disruptive than

AI or other emerging technologies (3.42) and cy-

bersecurity incidents (3.27). But they are more

disruptive than the talent shortage (3.17) or mis-

information and disinformation (2.9) (n=525).

Here’s how IT maturity shaped how organi-

zations view disruption:

I N N OVAT O R S (n=125)

• Overall, innovators have a heightened expec-

tation of disruption across all factors com-

pared to the average IT department.

• 1RWDEO\ LQQRYDWRUV DUH PRUH VLJQLÀFDQWO\

concerned about misinformation and climate/

environment/health than the average.

• Innovators are about twice as likely as the av-

erage to exclusively use sovereign AI models

(59% vs. 31%). But they are also more likely to

say they use foreign AI as well (28% use for-

eign and sovereign AI compared to 17% of the

average).

• When it comes to the biggest areas of risk

with critical vendors, innovators are most

concerned about unanticipated increases in

licensing and renewal costs (68%) and vendor

roadmap alignment (63%) than supply chain

attacks (17%) or logistical disruption (20%)

(n=87).

AVERAGE (n=249)

• Rank cybersecurity incidents as relatively

higher compared to other disruptive factors,

placing it second after AI and emerging tech.

,QQRYDWRUVUDQNLWÀIWKRYHUDOO

• Rank government-enacted regulatory change

above the talent shortage, while innovators re-

verse those rankings.

• One in six say they have no immediate plans

for the adoption of sovereign AI models. No

innovators describe their position in this way.

• When considering risk with critical vendors,

average departments are also most concerned

with unanticipated price increases (58%) but

less concerned with strategic roadmaps (33%).

They fret over compliance, security, and priva-

cy concerns instead (50%) and less about ven-

dor quality (n=112).

SIGNALS

Innovators have a

heightened expectation

of disruption across all

factors compared to the

average IT department.

TECH TRENDS 2026

Info-Tech Research Group

22 23

IT departments can increase

organizational resilience by

rearchitecting their solutions

to be component-based and re-

usable. Getting away from frag-

mented solutions that solve

different functional problems

will simplify IT and reduce

costs. IT will less often build its

own solutions and instead seek

to assemble reuseable compo-

nents provided by fewer ven-

dors that can offer end-to-end

ecosystems. This will help IT

organizations avoid accruing

more technical debt. “Building

is the easy part, they’ve found,

but then it’s the supportability,

the maintenance, the updating

from a security and resiliency

standpoint that makes it real-

ly, really hard for these busi-

nesses to keep up as they’ve

done some of that homegrown

development,” says Stu Brad-

ley, sr. vice president, risk,

fraud and compliance solu-

tions, SAS. With less time

spent on managing technical

debt, IT departments can fo-

cus on solutions that help dif-

ferentiate their organizations,

such as customizing AI with

organizational data.

Resilient supply chains can

mitigate the price increases

and uncertain nature of tariffs

through a series of tactics:

• Strategically time purchas-

es and contract renewals to

avoid peak prices.

• Extend lifecycle of existing

devices and avoid refreshing

at high costs. Bulk purchase

to get a better price per unit,

hopefully at a time when mar-

ket conditions favor the buyer.

• Shift sources to domestic or

nearshore locations where

possible to reduce tariff expo-

sure and get better long-term

price predictability (AIMMS,

2025).

Organizations that can in-

crease supply chain agility and

integrate ESG considerations

can position themselves as

OHDGHUV IXWXUHSURRÀQJ VXS-

ply chains for future upheav-

als while avoiding the negative

effects of the current scenario.

Consumers are increasingly

demanding more transparency

from organizations in terms of

their ethical and environmen-

tally friendly operations. IT

can provide traceability tools

that anchor brand credibility

to provable data and help the

organization stand out from

the competition. By accepting

less optimal near-term costs

in their supply chains, orga-

nizations are setting up for

longer-term sustainability (In-

spectorio, 2025).

OPPORTUNITIES

A RATIONALIZED

IT ENVIRONMENT AVOID TARIFF UNCERTAINTY THRIVE IN VOLATILITY

With less time spent on managing

technical debt, IT departments can focus

on solutions that help differentiate

their organizations.

More than half of IT departments expect to in-

crease spending between 1% and 10% in 2026, and

almost one-quarter expect to increase spending

by more than 10%, according to the Future of IT

Survey 2025. It’s a good thing, because higher

FRVWVDUHOLNHO\(YHQÀUPVWKDWDYRLGWDULIIVE\

shifting sourcing locations in-market will likely

face increased costs in the near term, as domes-

tic suppliers are likely to have higher prices. Piv-

oting supply chain sources can also incur more

logistics expenses and transitional pain.

Organizations seeking to build supply chain re-

silience by shifting operations or sourcing to dif-

IHUHQWUHJLRQVPD\ÀQGWKHFRPSOH[LW\RIFRP-

pliance more burdensome than they can keep

up with. Commitments to ethical sourcing and

environmental standards will require stringent

traceability deployed through the supply chain

that is above and beyond what most organiza-

tions are capable of today. It’s no wonder one in

ÀYHH[HFXWLYHVZRUU\WKDW´UHJXODWRU\RYHUKHDGµ

will divert resources from trying to improve

their businesses (Inspectorio, 2025).

With more barriers forming around different

PDUNHWV RUJDQL]DWLRQV PD\ ÀQG WKH\ DUH DW D

competitive disadvantage compared to compa-

nies on the other side of that barrier. Certain

key goods aren’t widely available from different

markets, including semiconductors made in Tai-

wan or rare earth elements mined in China. At

the same time, organizations may be diverting

investment away from innovation and research

and development to restructure their supply

chains.

RISKS

HIGHER SPENDING

AUDIT FATIGUE

COMPETITIVE DISADVANTAGE

TECH TRENDS 2026

Info-Tech Research Group

24 25

CASE STUDY

Taiwan Semiconductor Manufacturing Compa-

ny (TSMC) invested $165 billion into the US tech

sector to strengthen domestic chip fabrication

capacity, including three new fabrication plants,

two advanced packaging facilities, and a major

R&D team center (Atlantic Council, 2025).

0LFURQ D 86EDVHG VHPLFRQGXFWRU ÀUP LV LQ-

YHVWLQJ VHYHUDO ELOOLRQ GROODUV LQWR D FHUWLÀHG

fabrication facility in India. Although the region

isn’t known for semiconductor manufacturing,

labor costs are comparable to other options and

there are less threats of tariffs (McKinsey, 2025).

Intel is doing both onshoring and nearshoring. It

promised $100 billion plus into new US fabrica-

tion facilities and plans to invest $92 billion over

the next ten years into the EU to help develop

a full semiconductor value chain, including ad-

vanced fabs, assembly and testing facilities, and

research and development labs (Intel, 2025).

Batteries Plus restructured its suppliers to shift

from sourcing one-third of its inventory from

China to just 4% between 2018 and 2025. Moti-

vated by tariffs placed on China, CEO Scott Wil-

liams sought products from Vietnam, Malaysia,

the US, and several other countries. By distrib-

uting suppliers around the world, Williams cre-

DWHGPRUHÁH[LELOLW\WRVKLIWEHWZHHQFRXQWULHV

as new tariffs come into effect in the future. At

the same time, the company increased automa-

tion in its warehouses and renegotiated with

suppliers to mitigate the cost of tariffs (Business

Insider, 2025).

THE SEMICONDUCTOR INDUSTRY DIVERSI-

FIES ITS MANUFACTURING BASE

RECHARGING SUPPLY LINE DIVERSITY

Governments around the world are making ef-

forts to foster homegrown AI models to protect

economic and security sovereign interests. Do-

mestic telcos are getting involved in providing

the compute infrastructure and services toward

the effort:

• In Canada, Bell Canada’s Bell AI Fabric project

commits to construct 500 MW of hydroelec-

tric-powered (a renewable energy) AI com-

pute capacity across six data centers. Telus is

building two “Sovereign AI Factories” in BC and

Quebec, featuring NVIDIA-powered supercom-

puters. It plans to provide GPU-as-a-service to

enterprises through its Fuel iX Gen AI platform.

• In Europe, Orange plans to offer “sovereign-

ty-as-a-service” and provide customers control

over their data residency, ethical aspects of

LLMs, and ecosystem partners. Deutsche Tele-

kom offers GPU-as-a-service featuring NVID-

IA H100 Tensor Core processors from its data

centers in Germany, the Netherlands, and Swit-

zerland. FastWeb+Vodafone offers its own lo-

cal language model, FastwebMIIA, running on

an NVIDIA SuperPOD Computer from its infra-

structure in Italy.

• Telekom Indonesia is partnering with IBM to

provide sovereign AI services based on watsonx

(Taafe).

CALLING FOR SOVEREIGN AI

TECH TRENDS 2026

Info-Tech Research Group

26 27

Organizations will try to avoid creating supply

chains that are over reliant on any one suppli-

er or region to reduce exposure to tariffs, geo-

political risk, or predatory vendor practices. A

PRUHÁH[LEOHDSSURDFKZLOOLQFOXGHGLYHUVLI\LQJ

inputs and may add complexity as well as higher

logistics expenses. At the same time, they may

not be able to replace the specialized capabilities

of certain hubs. Organizations will be seeking to

optimize their own distribution, balancing pro-

duction times, and conduct location risk anal-

ysis when selecting new supplier hubs (AIMMS,

2025).

With the global trade system in a constant state

RIÁX[OHDGHUVDUHFRQFHUQHGDERXWPDNLQJGH-

cisions to restructure supply chains. Choosing a

supplier that may offer lower costs today could

turn out to be more costly tomorrow after trade

policies are updated. Instead, organizations

need more permanent competitive advantages

in their supply chain strategy. To build that, or-

ganizations are more often turning to scenario

modeling and forecasting simulations to provide

better visibility into future possibilities.

The cloud computing era introduced the concept

of data sovereignty, with countries concerned

about the location where sensitive data was host-

ed or even where it transited. Highly regulated

LQGXVWULHVOLNHWKHSXEOLFVHFWRURUÀQDQFLDOVHU-

vices were more restricted in requirements to

locate data domestically when using cloud-com-

puting services from foreign providers. The EU

extended the concept even further with GDPR,

bringing the private information of individuals

into consideration. Now, AI is extending sover-

eign considerations to the knowledge available

in LLMs, the bias they show, and even the cultur-

al context of their outputs. In addition, the eco-

nomic value of the future AI market is creating

incentives for governments to invest in local AI

models. On the other end of the spectrum, su-

perpowers and leading AI developers are vying

for global reach through their algorithms and

more interconnectedness.

CRITICAL FACTORS

STRATEGIC AGILITY

GEOPOLITICAL UNCERTAINTY AND

ETHICAL MISALIGNMENT

SOVEREIGN AI

RIGID,

LOCKED-IN

SOURCING

FLEXIBLE AND

ADAPTABLE

NATIONAL CONTROL

AND SELF-RELIANCE

SOVEREIGNTY

GLOBAL

INTERCONNECTEDNESS

FRAGMENTED,

HIGH-CONFLICT

WORLD

COOPERATIVE,

MARKETS-ORIENTED

TECH TRENDS 2026

Info-Tech Research Group

28 29

WHAT’S NEXT

Diversify sourcing locations and vendors where

possible. Alternatives to China can be found for

some supplies in other parts of Southeast Asia

that are less impacted by tariffs. For services, re-

negotiate with vendors when possible and struc-

ture in shared responsibility due to tariffs. Time

purchases of devices and other components to

avoid tariff increases when notice is given or buy

in bulk to decrease the cost per unit. Consider

tiering device procurement to ensure that de-

vices are right sized for users instead of need-

lessly buying premium models. Explore device-

as-a-service options to create price certainty for

device procurement over the long term. When

considering new vendors, increase cybersecuri-

ty scrutiny and perform due diligence to evalu-

ate risks.

Start running more, better simulations about

SRVVLEOH IXWXUH RXWFRPHV ,Q WKH ÀQDQFLDO VHU-

vices sector, organizations were limited to the

number of scenarios forecast because of avail-

able compute resources and siloed data. With

improved cloud infrastructure, organizations

can “run more simulations to look at a multi-

tude of different factors and evaluate for better

decision-making,” Stu Bradley says. AI can help

“sift through those vast data sets to identify pat-

terns, anomalies, and use that data for these risk

assessments.” Revisit vendor contracts and con-

sider building clauses for shared tariff responsi-

bilities or performance-based cost adjustments

(Anchin, 2025). Mature data governance to en-

KDQFH$,GULYHQIRUHFDVWLQJDVOHDGLQJÀUPVLQ-

vest in this area to create standardized supplier

inputs, clear accountability, and metrics for data

integrity (Inspectorio, 2025).

NOW NEXT

Scale demand forecasting to the entire enter-

prise and extend AI-predicted risks before they

materialize across the supply chain. Embrace a

component-oriented approach to delivering IT

value and rationalize vendor relationships to

emphasize simplicity and reusability. Invest in

domestic efforts to build sovereign AI and local

compute infrastructure to foster a more com-

petitive technology vendor environment. Seek

to take advantage of local government funding

programs that incentivize enterprises to pur-

chase from local providers.

“We’re in a state where governing is a

resilience exercise, not an oversight one. So

I have to build an organization that can deal

with unknowable uncertainties.”

Valence Howden

Info-Tech Advisory Fellow

Harmonizing Complexity: Control Tower Inte-

gration

Construction: Build Resilience: Strengthening

Construction Amid Policy Shifts

Manufacturing: Adapt to Uncertainty With a

Technology-First Action Plan for Manufacturing

Transportation: Keep Transportation Operations

Moving in a Shifting Market

LATER RECOMMENDATIONS

INDUSTRY RESOURCES

TECH TRENDS 2026

Info-Tech Research Group

30 31

Integrated

Organizational

Resilience

IT is moving beyond project-based digital change to become

an orchestrator of enterprise resilience, helping to embed risk

management into every capability, from APIs and data platforms

to AI agents and security controls. By centralizing governance,

automating compliance checks, and knitting together monitoring

and incident response, IT elevates its role from execution partner

to full partner in managing enterprise risk.

TECH TRENDS 2026

Info-Tech Research Group

32 33

AD HOC & REACTIVE DOMAIN-SPECIFIC & FRAGMENTED INTEGRATED & CENTRALIZED INTELLIGENT RESILIENCE

Risk is managed in silos and responsive to

LQFLGHQWVRUDXGLWÀQGLQJV

• Lack of centralized governance.

• Manual processes for managing risk.

• Risk is viewed as a cost center and compliance

burden.

Formal enterprise risk management with cross-

domain frameworks.

• Risk management is coordinated across do-

mains.

• Shared platforms have a central view of risk

data and reporting.

• Board-level risk governance applies to the en-

tire organization (The IIL Blog, 2025).

• A psychologically safe organization.

Formal risk management practice applied

within functions.

• Risk registers are created but not integrated

across domains.

• Cybersecurity risk is managed separately from

operational risk.

• System-wide and third-party risks are weak

areas.

Risk management is viewed as a strategic

enabler and core to enterprise resilience and

adaptability.

• Scenarios-based planning for a future-orient-

ed view of risk.

• Integrated AI to augment risk-oriented deci-

VLRQPDNLQJZLWKDSUHGLFWLYHSURÀOH

• Risk-based insights inform other enterprise

capabilities from innovation to partnerships

to operations (Reworked, 2025).

• Fully adaptable risk governance and steward-

ship.

• Radical honesty.

ENTERPRISE RISK MANAGEMENT PROGRESSION

FROM REACTIVE TO RESILIENT

At a time of global uncertainty and

rapid technological disruption, risk

must be approached as a strategic

capability rather than a siloed func-

tion. Traditional approaches to enterprise risk

management are struggling to keep up with

the rapid change facing organizations, and new,

converged models of digital risk and resilience

are emerging. Rather than managing risk with

backward-looking approaches that focus on au-

dit and assessment, more dynamic and real-time

enterprise-wide resilience is required. For IT,

previous risk focus areas of cybersecurity and

uptime are no longer enough. IT must act as part

of the enterprise to modernize risk management

and create the conditions for sustainable growth

in an uncertain future.

Traditional approaches to enterprise

risk management are struggling

to keep up with the rapid change

facing organizations.

TECH TRENDS 2026

Info-Tech Research Group

34 35

I N N OVAT O R S (n=59)

• Are almost 2.5 times more likely to structure

risk management in a fully integrated way

than the average department (79% vs. 32%).

• Are 16 percentage points more likely to say it’s

very important to improve organizational risk

management capabilities than the average

(71% vs. 55%).

• Thirty-eight percent have a chief risk and re-

VLOLHQFHRIÀFHUWREHPRVWDFFRXQWDEOHIRUJRY-

ernance, risk, and compliance (GRC).

• When negotiating new contracts with vendors,

innovators prioritize security, privacy, and

compliance more than the average. They were

OHVV FRQFHUQHG ZLWK RSHUDWLRQDO ÀQDQFLDO

reputational, and ethical risks.

AVERAGE (n=181)

• One-quarter have a chief risk and resilience

RIÀFHUWRWDNHRQ*5&DFFRXQWDELOLW\7KH&,2

is the next-most accountable position for GRC

according to 18% of respondents, followed

closely by the CTO at 17%.

• More than half (53%) describe their risk man-

agement as “integrated and centralized.”

• When negotiating new contracts with vendors,

reputational risk is the leading concern among

all risks (14%), with legal concerns (3.8%) and

lock-in (4.6%) being the risks of least concern.

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

36 37

Assurance reports of the past look backward at

our systems of record and expose risks that al-

ready exist and may have even caused loss. An

ability to detect threats in real time with AI

monitoring and plan for potential scenarios with

AI-augmented planning could put enterprises

in a position to avoid pitfalls of the future. AI

should be harnessed as a tool that can augment

enterprise risk planning, informing but not re-

placing human judgment. The enterprise should

become comfortable operating with uncertainty

(Howden).

As enterprises digitalize more of their processes,

APIs become an important part of their digital

infrastructure. Modern platforms can discover

and score risks in real time, often automating

mitigation responses. Enterprises can embed

their regulatory requirements and custom poli-

FLHVLQWRFRGHDQGEHQHÀWIURPFRQWLQXRXVPRQ-

itoring and automated enforcement. These plat-

forms are making different risk vectors – such

as security, operations, and compliance – more

visible and more actionable for the human in the

loop (AppSentinels, 2025).

IT may view risk mitigation as having the right

controls in place, but the future of enterprise

risk will be more about maintaining coherence

through complexity and uncertainty. Real-time

risk information should be widely communicat-

ed across the organization to enable frontline

decision-making, creating adaptability through

better awareness of risk and the desired re-

sponse. Enterprise risk leaders should be strong

cross-functional communicators, have a strong

sense of ethics, and strive for better resilience

(The IIL Blog, 2025).

OPPORTUNITIES

MOVE FROM RISK RECOVERY TO RISK

PREVENTION AND PROACTIVE RISK RESPONSE

COMPLIANCE EMBEDDED WITHIN CODE

BUILD RESILIENCE INTO

THE ENTERPRISE FABRIC

The concept of automatically enforcing corpo-

rate policy at the code level might seem tanta-

lizing to some risk-averse enterprises, but trying

to put in place too much structure can actually

result in too much rigidity. “Overly rigid gov-

ernance is a deterrent to an organization, and

overly volatile governance will be too late,” says

Valence Howden, advisory fellow at Info-Tech

Research Group. “If we don’t put a reasonable

boundary around this, it’s going to go horribly

wrong.”

AI may be helpful in augmenting information

collection or in planning for different future

scenarios, but it shouldn’t be relied upon to

make decisions. “It doesn’t make good enough

decisions to be used in a governing space,” How-

den says. “It doesn’t have enough context and

belief injection of responsibility to understand

it.” Further, AI models can drift from their orig-

inal intent, yet humans may not notice until the

damage is done. Resilience requires human-led

ethics to be front and center. Models break down

given too much complexity, and organizations

must be aware of how to use them appropriately

(Howden).

Will organizations be able to avoid the tempta-

tion to ignore risks that are counter to their own

interests? Organizational culture can develop

in an unhealthy way that punishes those that

bring attention to risk or are honest in reporting

on the real risks in tension with organization-

al goals. Organizations must also come to grips

that working in uncertainty means they must

be honest with themselves that they just can’t

know everything to satisfy their need to feel as-

sured (Howden).

RISKS

RIGID GOVERNANCE

OVER-AUTOMATING GOVERNANCE

LACK OF HONESTY

“Government has to be

exponentially adaptable instead

of extensively improved. That

means you can’t be rigid by

nature.”

Valence Howden,

Info-Tech Advisory Fellow

TECH TRENDS 2026

Info-Tech Research Group

38 39

EXAMPLES

At the Cisco Toronto Innovation Centre, inventions are turned into

enterprise-ready innovations by treating risk as a design input.

Risk is assessed early in the innovation lifecycle, and solutions are

WDLORUHGWRLQGXVWU\VSHFLÀFFRQVWUDLQWV9DOLGDWLRQFRPHVIURP

SLORWSURMHFWVEHIRUHVFDOLQJXSLQRUGHUWRUHGXFHÀQDQFLDODQG

reputational exposure. Startup Morgan Solar is using this ap-

proach to develop its solar window covering solution to harvest

UHQHZDEOHHQHUJ\IURPRIÀFHEXLOGLQJV&LVFRSURYLGHGDFFHVVWR

secure environments, including its own innovation center, as well

as to client sites at the Bank of Montreal. Cisco provided its own

technical integration and automation support and enabled col-

laboration across different stakeholder groups including industry,

government, and BMO. Using a “fail fast, learn fast” mindset that

tests ideas when they are 80% ready, failure is turned into an ex-

pectation and learning opportunity and avoids “analysis paralysis”

(Interview with Justin Cohen).

The General Bank of Canada is using generative AI to create pre-

liminary risk reports as an input for the risk team. This reduc-

es the team’s manual effort by 80% and allows the staff to spend

more time on higher value analysis. “We’ve started to experiment

ZLWKJHQHUDWLYH$,IRUYHU\VSHFLÀFXVHFDVHVµVD\V$GDP(QQDPOL

FKLHIULVNRIÀFHUDW*HQHUDO%DQNRI&DQDGD´:HPXVWFRQWLQXHWR

strengthen our risk modeling capabilities because they drive our

SURÀWHQJLQHµ7KHEDQNVWLOOFRQVLGHUVLWVHOILQWKHH[SHULPHQWD-

tion phase with generative AI, with an aim to incorporate more

diverse and dynamic economic scenarios into its models.

The reinsurance sector is a leader in adopting generative AI for

risk, a core aspect of its business. AI allows reinsurers to incorpo-

rate more data sets, including unstructured data from satellite im-

agery to IoT sensor data. It’s also enhanced catastrophe modeling

so reinsurers can better structure coverage for rare and extreme

events. One reinsurer, Swiss Re, generates dynamic risk scores by

using AI algorithms that draw upon a data lake. Another, Munich

Re, also applies AI to improve natural catastrophe models (Jean

Edwards Consulting, 2025).

RISK-AWARE INNOVATION

AUGMENTED RISK REPORTING

REINSURANCE LEADS AI-ASSESSED RISK

TECH TRENDS 2026

Info-Tech Research Group

40 41

7KHGHVLJQRIPRGHUQV\VWHPVLVGHÀQHGE\DPRGXODUDSSURDFK

that uses APIs to integrate with other systems. Zero-trust securi-

W\IUDPHZRUNVGHÀQHDOOFRQQHFWLRQVDVSRWHQWLDOWKUHDWVDQGUH-

TXLUHSHUVLVWHQWLGHQWLW\$PDMRUEHQHÀWRIWKLVDSSURDFKLVWKDW

it minimizes risks by containing failures and allows nimble adap-

tation to alternate solutions. New integration of AI for real-time

monitoring and reporting continues to improve upon risk visibil-

ity and planning. The modern approach contrasts with legacy ap-

proaches that are monolithic and perimeter-oriented with strict

rules-based controls.

Depending on static risk registers, audits, and reviews of past

events won’t help enterprises build resilience in the face of tomor-

row’s threats. Enterprises must develop a better foresight muscle

by practicing scenario development using simulations. AI-enabled

monitoring can help detect weak signals today that could become

a seismic shift in the future.

New capabilities in governance, risk, and compliance platforms

tap into generative AI to augment human judgment. There will be

temptation to push further into AI capabilities and pursue agentic

automation of enterprise-risk-oriented tasks with automated pol-

icy enforcement. But risk requires human judgment and ethical

consideration to build true organizational resilience, and those

that try to automate too far will suffer from misalignment and

areas of weakness (McKinsey & Company, 2021).

CRITICAL FACTORS

RESILIENT TECHNOLOGY ARCHITECTURE

FORESIGHT ADOPTION

AI AUGMENTATION VS. OVER-AUTOMATION

RESILIENT

BY DESIGN

RESILIENT TECHNOLOGY ARCHITECTURE

BRITTLE LEGACY

SYSTEMS

PROACTIVE

FUTURES-ORIENTED

FORESIGHT ADOPTION VIEW

REACTIVE

BACKWARD-LOOKING

HUMAN-CENTERED

RESILIENCE

AI AUGMENTATION VS. OVER-AUTOMATION

OVER-AUTOMATED

BRITTLE COMPLIANCE

TECH TRENDS 2026

Info-Tech Research Group

42 43

WHAT’S NEXT

Take a cross-functional view of risk that starts

at the board level and trickles down through the

executive leadership through a partnership that

should embed risk accountability across the en-

terprise. Rather than divvying up functional risk

into isolated silos, enterprises should focus on

collaboration. IT leaders can help translate tech-

nical risks into enterprise risks and deploy tech-

nology capabilities that augment and automate

risk management appropriately.

Start a foresight practice that considers risks

in different future scenarios, with generative

AI assisting in the brainstorming of different

potential futures. Also use AI to augment risk

management through automated monitoring

and reporting. Seek automation of compliance

monitoring for high-risk areas while keeping a

human in the loop to assess alerts and evaluate

AI performance.

Explore and deploy a risk management API in-

tegration platform to improve visibility of dig-

ital risk. Embed “compliance as code” into your

own development lifecycle and expand AI use to

model validation, scenario modeling, and anom-

aly detection. Formalize your cross-functional

governance approach with councils, and create

an executive risk dashboard to translate techni-

cal risks into business impacts.

NOW NEXT

Adopt an intelligence-driven risk forecasting ap-

proach with institutionalized foresight practices

such as scenario planning and red teaming. Build

GLJLWDOLQIUDVWUXFWXUHZLWKPRGXODUDQG$3,ÀUVW

design so point failures are contained. Cultivate

a viewpoint of risk management as a competi-

tive differentiator that helps your organization

build trust, develop clear ethical AI alignment,

and deliver on ESG considerations. Pursue inno-

vation fearlessly by embracing failure (within

safe constraints) as a learning opportunity, and

VHHNWRH[WHUQDOO\LQÁXHQFHLQGXVWU\VWDQGDUGV

and public regulatory frameworks through ac-

tive engagement and contribution.

At the beginning of your risk journey?

Build an Effective IT Controls Register

Ready to take the next step?

Optimize IT Governance for

Dynamic Decision-Making

LATER RECOMMENDATIONS

TECH TRENDS 2026

Info-Tech Research Group

44 45

Multi-Agent

Orchestration

Software vendors have embedded LLMs in their software to provide

guidance or to automate work with mixed results so far. The next

step is creating agents that will be able to adapt to different contexts

and complete tasks and potentially entire processes. Unlocking fully

realized autonomy will require the orchestration of multiple agents.

TECH TRENDS 2026

Info-Tech Research Group

46 47

Enterprises are looking to agents

to provide ROI from AI.

First came the potential of LLMs.

Then came the limitations.

Now, we hope, the solution.

Vendors quickly integrated large lan-

guage models into their applications

via chat over the last couple of years.

This was well adopted by enterpris-

es. Info-Tech’s Tech Trends reports showed that

most enterprises were committing to AI invest-

ment and increasing that investment over the

last couple of years as well.

Limitations emerged with more complex,

PXOWLVWHSZRUNÁRZVWKDWFKDOOHQJHGWKHFRQWH[W

limitations and reasoning gaps of LLMs. Many

enterprises struggled to move past the proof-of-

FRQFHSWWRÀQGDJRRGUHWXUQRQLQYHVWPHQW

AI agents can perceive their digital envi-

ronment and use contextual information to

make decisions and take actions to achieve an

assigned goal, automating new tranches of the

NQRZOHGJHEDVHGZRUNÁRZ

Multi-agent orchestration involves multi-

ple AI agents collaborating to solve even more

complex, multistep problems and holds poten-

tial to automate entire processes for an enter-

prise. A supervisor AI agent can be assigned

to manage agents along a process and enforce

enterprise governance.

TECH TRENDS 2026

Info-Tech Research Group

48 49

I N N OVAT O R S (n=137)

• Innovators are looking to “raise the ceiling”

with agentic AI by investing in it to drive

growth.

• About twice as likely as the average IT depart-

ment to be bullish on agentic AI investment,

with 81% currently using agentic AI and plan-

ning to invest further in the future.

• Also, twice as likely as the average to support

growth initiatives with AI. Twenty-seven per-

cent will pursue growth initiatives with AI

agents before 2027.

AVERAGE (n=307)

• Many are still planning to adopt agentic AI,

with 41% looking at 2026 or later as the time

to invest, while 13% have no plans to adopt yet.

• One-third set a goal to “improve operational

HIÀFLHQF\µZLWK$,DJHQWVDQGZDQWWRLP-

prove the user experience.

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

50 51

OPPORTUNITIES

AI agents can both automate coding tasks and

augment human developers as they work and

are effective across the entire software develop-

ment lifecycle, including development, testing,

and performance monitoring.

ACCELERATE SOFTWARE DEVELOPMENT

AI service agents are available to respond to cus-

tomers 24/7, providing personalized interaction.

Many routine tasks can be completed automati-

cally without the need for human intervention,

and support inquiries can also be turned into

leads via product or service recommendations.

IMPROVE CUSTOMER SUPPORT

Agents can automate lead outreach, optimize

pricing for prospects, and predict demand. Sales

teams adopting AI agents show improvement in

hitting revenue targets (Multimodal, 2025).

DRIVE NEW REVENUE

Early deployments show good return on invest-

ment for AI agents that can both monitor to pro-

actively resolve repeated IT issues to avoid tick-

et creation and reduce the time to resolution on

tickets received through enhanced triage, pri-

oritization, and automated resolution (Swish.ai,

2025).

IMPROVE YOUR SERVICE DESK

Threat actors could seek to gain unauthorized

DFFHVVWRDQ$,DJHQWRUIDLOLQJWKDWLQÁXHQFH

its operations by manipulating its external re-

sources (data or tooling) to trick the agent into

downloading malware or exposing sensitive in-

formation (AI Governance Library, 2025). Hack-

HUVWKDWÀQGDYXOQHUDELOLW\LQDQ//0FRXOGSR-

tentially try to exploit it across all AI agents that

use the LLM.

With AI agents accessing data and performing

actions where normally an authorized employee

would be doing so, there is risk of an agent over-

stepping and doing something that goes beyond

the intended permissions. This is of particular

concern in highly regulated industries (Syncura,

2025).

Employees may perceive AI agents as a threat

to their job or just not trust the agent to per-

form actions and provide answers. Workers who

see AI agents doing increasingly more of their

work may feel that they are not valued. On the

customer side, many people have a negative re-

action to encountering AI. Customers who per-

ceive AI that is being used to provide shortcuts

or replace people in creative work or think it is

a tedious way to interact with a company may

reject companies that adopt it and publicly crit-

icize the company (Interview with Heintzman).

RISKS

NEW CYBER VULNERABILITIES

UNINTENDED ACCESS

CULTURAL REJECTION

“My engineers are pretty

confident that they’re at least ten

times more productive today

than they were a year ago.”

Douglas Heintzman,

CEO,

Syncura

“We have to put guardrails on what agents

can talk to the same way we do with humans.

Sometimes someone shouldn’t be able to talk

to somebody else or have access to a specific

piece of data. Same concept goes with agents.”

Martin Bufi,

Principal Research Director,

Info-Tech Research Group

TECH TRENDS 2026

Info-Tech Research Group

52 53

CASE STUDY

US mortgage provider Direct Mortgage Corp.

GLGQ·WKDYHHQRXJKORDQRIÀFHUVWRSURFHVVDULVH

in mortgage applications, leading to long loan

closing times and missed lending opportunities.

It adopted Multimodal’s AgentFlow platform

DQGGHSOR\HGDJHQWVWRDXWRPDWHWKHZRUNÁRZ

of dozens of document types. Direct Mortgage’s

internal documents provided training context

so applications could be accurately handled. The

solution cut operational expenses per document

by 80% and loan processing time was cut in half.

Borrowers now receive a loan decision within 24

hours (Multimodal, 2025).

Walmart connected AI agents to automation sys-

WHPVWRVWUHDPOLQHUHWDLORSHUDWLRQV:RUNÁRZV

connected inventory, pricing, and customer

support data with robots that replenish shelves,

and handle routine shopper requests. Walmart

claims that it has accelerated product develop-

ment, boosted sales with increased shelf avail-

ability, and cut labor hours on manual tasks (Re-

dress Compliance, 2025).

FROM APPLICATION TO LOAN IN 24 HOURS

LOWERING PRICES,

INCREASING AUTOMATION

Sourcery uses a combination of generative and

FODVVLÀFDWLRQEDVHG $, WR LPSURYH GHYHORSHU

productivity. It augments developers by ex-

plaining new code, automatically conducts code

reviews, and automates test cases. The solution

orchestrates several agents to remove repetitive

tasks for developers so they can focus on higher

value tasks (Info-Tech, 2023).

AI WRITES CODE

How reliable can agents be? How quickly can

they adapt to new context? Goal-oriented agent

systems promise to be more adaptable than pre-

vious rules-based automation approaches. Mul-

tiple agents may even be assigned to the same

process and argue among themselves about

the best solution to reduce errors and bias (In-

terview with Heintzman). At their core, agents

will be limited in capability by underlying AI

models and whether AI vendors can continue to

make strides in improving intelligence or falter

in making progress because of hindrances like

power supply, access to GPUs, or a lack of new

training data.

If AI agents can’t talk to each other, it will bot-

tleneck productivity. There’s been positive signs

toward a more open ecosystem, with early adop-

tion of protocols like Anthropic’s Model Context

Protocol (MCP), which allows agents to connect

with enterprise applications and data, and Goo-

gle’s Agent2Agent (A2A) protocol, which allows

agents to talk to each other. There are also

open-source frameworks available to design

multi-agent systems from Microsoft, Amazon

Web Services, and IBM (Finio, 2025). This signals

that enterprises may largely avoid issues where

agents from one vendor won’t be able to talk to

another, resulting in fragmentation.

But companies competing against each oth-

HUPD\ÀQGLQFHQWLYHVWRVWDUWSXWWLQJXSZDOOV

around their data or agents. Salesforce is block-

LQJRWKHUVRIWZDUHÀUPVIURPVHDUFKLQJRUVWRU-

ing Slack messages, cutting the messaging and

collaboration software off from indexing by

third parties (Reuters, 2025).

Generative AI is pretty good at delivering an-

swers but can’t explain how it arrived at those

answers. Despite efforts to make generated out-

put more explainable, researchers have found

LW GLIÀFXOW RU LPSRVVLEOH WR WUDFH $, UHDVRQLQJ

accurately. While recent updates to reasoning

models aimed to show the steps taken to reach

DFRQFOXVLRQUHVHDUFKHUVÀQGWKRVH´FKDLQVRI

thought” are generated after the answer is al-

ready derived to satisfy users (Interview with

Heintzman). With AI agents acting on generative

AI output, there is risk to consistency in perfor-

mance or regulatory requirements for auditabil-

ity. This problem is being addressed through

an agent lifecycle management approach that

provides checks at different points in operation.

Evaluation metrics used to report on agentic

transparency include faithfulness, context rel-

evance, and answer similarity (Szarmach, 2025).

CRITICAL FACTORS

AGENT CAPABILITY:

INTEROPERABILITY OBSERVABILITY

SIMPLE

BUSINESS LOGIC

AGENT CAPABILITY

BEYOND

HUMAN CAPABILITY

PROPRIETARY

SILOS

INTEROPERABILITY

OPEN AND STANDARDIZED

ECOSYSTEM

BLACK

BOX

OBSERVABILITY

FULL TRANSPARENCY

AND AUDITABILITY

TECH TRENDS 2026

Info-Tech Research Group

54 55

WHAT’S NEXT

Start by deploying AI agents for well-document-

ed and repetitive tasks such as document classi-

ÀFDWLRQGDWDH[WUDFWLRQRUDQVZHULQJFRPPRQ

customer service queries.

Start with API-driven integrations where

they exist in enterprise systems to integrate AI

agents and identify where to deploy model con-

text protocol (MCP) servers to extend agent ac-

cess.

Keep humans in the loop and default to hav-

LQJSHRSOHSURYLGHPRUHRYHUVLJKWDWÀUVW&DUH-

fully monitor and track the output of AI agents

and identify acceptable performance standards.

Explore multi-agent collaboration for more

complex processes. Identify where single agents

aren’t enough to solve a problem and compose a

PXOWLDJHQWÁRZDVDSURRIRIFRQFHSW

Adapt agent-to-agent (A2A) and other in-

ter-agent protocols as necessary to ensure com-

munication between AI agents. Use payload ref-

erencing to handle the exchange of large text

and code volumes and optimize inter-agent

communication (Shu, 2024).

Adopt and develop internal evaluation

EHQFKPDUNV IRU PXOWLDJHQW ZRUNÁRZV VXFK

as an assertion-based framework, that simulate

multi-agent systems and judge their likelihood

of achieving goals (Shu, 2024).

NOW NEXT

Evolve toward role-based agent supervisors that

ZLOORYHUVHHW\SHVRIWDVNVOLNHFRGLQJRUÀQDQ-

cial analysis and work to achieve different goals.

-XVWOLNHZLWKSHRSOHUROHVGHÀQHZRUNIRFXVDV

well as system and data access. Agents should

EHUHÁHFWLYHDQGHYDOXDWHWKHLURZQZRUNWRLP-

prove (Deloitte, 2025).

Expand beyond enterprise-wide agent in-

tegration and look to an external ecosystem of

DJHQWLFFROODERUDWLRQWRÀQGQHZHFRQRPLHVRI

scale and opportunities to pursue shared inter-

ests with external organizations (Syncura, 2025).

Enable comprehensive auditing of agents by

documenting chain-of-thought level detail.

Build Your Agentic AI Prototype

LATER RECOMMENDATIONS

TECH TRENDS 2026

Info-Tech Research Group

56 57

Smart

Sensing

Networks

The convergence of micro-electromechanical systems (MEMS),

quantum sensors, and AI algorithms are creating the next generation

of IoT applications. Intelligence is fused into a new class of sensors and

actuators that opens the door to context-aware insights in real time

with autonomous actions delivered at the edge, reducing latency and

reliance on centralized processing.

There were 31 bil-

lion MEMS sensors

shipped in 2024 with

a total revenue of

$15.4 billion and this is pro-

jected to grow to $19.2 billion

by 2030 (Yole Group, 2025).

The majority of MEMS sensor

applications are in consumer

electronic devices, but a new

wave of sensor capabilities

combined with AI algorithms

that can run on the sensors

themselves are opening new

potential opportunities in

healthcare, manufacturing, au-

tomotive, and defense markets.

Quantum sensors are also

graduating from research labs

into real-world applications,

with public research funding

propelling prototype develop-

ment from vendors like Sand-

boxAQ. The market is project-

ed to grow from $377.3 million

to $1.2 billion by 2032 (For-

tune Business Insights, 2025).

Growth is spurred by new ca-

pabilities such as magnetic

sensors, gravity sensors, and

other environmental sensors

including ones for radiation.

TECH TRENDS 2026

Info-Tech Research Group

58 59

GENERATION QUALITIES APPLICATIONS

Legacy

Modern

Standard

Premium

Modern

Standard

Emerging

Now

Future

Projection

Sensor with no electronics.

Sensor with remote control.

MEMS sensor with signal conditioning and

DPSOLÀFDWLRQ

All of 3, plus digital ports to allow sensor

addressing and self-assessment.

All of 4, with full digital transmission and local

GLJLWDOSURFHVVLQJVHOIWHVWLQJXVHUGHÀQHG

algorithms, and compensation algorithms

(Allicdata, 2025).

All of 5, plus ubiquitous edge AI and real-time

decision-making with dedicated AI processing

embedded. Advanced miniaturization

(Interview with Eddy Tsai).

All of 6, with advanced miniaturization to

the nano-scale with nano-electromechanical

systems (NEMS), seamlessly integrating sensors

into the fabric of materials that perform with

high precision and no latency. Includes ambient

energy harvesting.

Passive sensing

Monitoring and remote

interaction

Environmental monitoring

and simple industrial sensing

Smartphone capabilities such

as accelerometers for motion

detection and orientation

Predictive maintenance,

advanced mobile electronics

such as 3D audio, active

noise cancellation, hearing

assistance, and health tracking

Digital scent for healthcare

and industrial use cases

Thinner chips with cooling for

smart glasses and ultra-thin

smartphones

Biological diagnostics and

treatments

Digital twins of complex

physical environments

EVOLUTION OF SMART SENSORS AND FUTURE PROJECTION

More mature IT departments are more focused

on investing and scaling the internet of things

(IoT), while average IT departments are focused

on modernizing network infrastructure. Signif-

icant portions of cohorts are responding to AI

demands by scaling both cloud computing and

edge computing.

I N N OVAT O R S (n=145)

• 37% are already investing in IoT and plan to

increase investment compared to 26% of the

average group.

• Only 13% of innovators say they are planning

to invest less in IoT than previously, compared

to 40% of the average.

• 34% of innovators plan to adopt IoT by the end

of 2026 vs. 27.5% of average IT departments.

AVERAGE (n=316)

• Are almost as likely as innovators to have de-

ployed or expanded edge computing to sup-

port better inferencing (27% of average de-

partments vs. 31% of innovators).

• Are almost four times as likely as innovators

to be focused on upgrading networking infra-

structure.

• Are nearly as likely as innovators to be increas-

ing cloud consumption due to demands of AI

(28% of average departments vs. 33% of inno-

vators).

(Insights from Future of IT 2026 Survey)

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

60 61

Recent advancements in MEMS technology in-

clude gravity sensing (SMG) and digital scent

detection (Ainos Inc., 2025). This has opened up

new applications for aerospace, defense, health-

care, robotics, and manufacturing. As new ca-

pabilities come to market, there will be new op-

portunities for other industries to deploy smart

sensing applications.

$,DOJRULWKPVFRQWLQXHWREHFRPHPRUHHIÀFLHQW

in running directly on smart sensors, provide

lower latency response, and reduce power con-

sumption. Diverting processing away from the

cloud will improve application performance.

0(06WHFKQRORJ\LVDOVRÀQGLQJZD\VWRDGGUHVV

the physical limitations of processing at the edge

WKURXJKEHWWHUFRROLQJDQGLQFUHDVHGÁH[LELOLW\

New sensors are being manufactured at milli-

meter-scale to allow for devices to be developed

that are smaller, lighter, and less power-inten-

sive. This reduces the costs for operations and

opens the door to new form factors that are

wearable, or in the future, even able to go into

bodies to conduct diagnostics on individual or-

gans. At the present, chip miniaturization can

at least divert costs away from enterprise cloud

computing bills or data center additions.

CRITICAL FACTORS

NOVEL SENSING CAPABILITIES

RESILIENCE OF EDGE AI PROCESSING

MINIATURIZATION

Optical, auditory,

motion-based Olfactory,

gravity, etc.

ESTABLISHED

SENSING CAPABILITY

NEW HORIZON EXPANDS

Computational requirements

exceed edge processing and

thermal constraints

Edge AI processing effectively

and efficiently handles data

at the edge

LOW

RESILIENCE

HIGH

Millimeter scale,

or about one thousandth of a meter Sub-molecular scale,

or about one billionth of a meter

MICRO-SIZED

SCALE

NANO-SIZED

TECH TRENDS 2026

Info-Tech Research Group

62 63

Where Industry 4.0 brought a focus on digiti-

zation, deployment of IoT, and an emphasis on

cloud computing with an eye on automation and

JUHDWHUHIÀFLHQF\,QGXVWU\EDODQFHVDXWRPD-

tion with human collaboration. Smart sensing

networks will enhance predictive maintenance

applications, directing people to address equip-

ment issues before failures occur. Examples in-

clude time-series AI that detect signs of bearing

wear or abnormal vibration in robotic assembly

lines (NXP, 2025). Cameras enabled with cogni-

tive vision can detect product defects on the as-

sembly line in real time (ST, 2025). Environment

and equipment monitoring with autonomous

anomaly detection can enhance workplace safe-

ty.

Sensors with new capabilities for environmen-

tal and physical monitoring are opening up new

applications in healthcare through wearable

devices and onsite sensors. New diagnostics are

available to detect respiratory issues, for wom-

en’s health screening, and non-invasive blood

glucose monitoring (Compamed, 2025).

MEMS chips have contributed to smartphone

innovation over the past decade, enabling bet-

ter cameras, microphones, and motion sensors

on these pocket-sized devices. Continued minia-

turization and new capabilities will enable new

wearable form factors like smart glasses. New

millimeter-scale chips that deliver audio, detect

environmental changes, and cool on-board com-

ponents to maximize performance will fuse with

AI algorithms to deliver seamless digital experi-

ences.

OPPORTUNITIES

INDUSTRY 5.0

HEALTHCARE HYGIENE

NEW FORM FACTORS FOR

CONSUMER TECHNOLOGY Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

While advancements in smart sensors are always

pushing the boundaries on processing power

DQGEDWWHU\OLIHWKHUHDUHVWLOOVLJQLÀFDQWOLPLWV

when working at a miniaturized scale. AI algo-

rithms requiring more inference compute pow-

er and working with high volumes of data could

run up against processing power at the edge or

consume battery-based power quickly, resulting

in either poor performance or outages.

Many smart sensor devices depend on capturing

time-series data, which can be technically chal-

lenging to parse the signal from the noise. Data

scientists must contend with sampling rates and

dramatic variation in data volumes over time.

Without proper alignment, sensors may not sig-

nal as intended (NXP, 2025).

Creating more IP addresses on the network

means creating more vectors to defend against

cyberattacks. Connected sensors may be in re-

mote locations where their physical security

can’t always be monitored. Prevention of in-

trusive attacks becomes more important, and a

variety of techniques are required ranging from

DXWKHQWLFDWLRQ WR FU\SWRJUDSKLF FRGH YHULÀFD-

tion to remote wireless security management.

The costs of security tends to rise as does the

number of interconnected devices (Pech et al.,

2021).

RISKS

EXCEEDING PERFORMANCE THRESHOLDS

COMPLEXITY OF TIME-SERIES DATA

COMPLEXITY AND COST OF IOT SECURITY

TECH TRENDS 2026

Info-Tech Research Group

64 65

CASE STUDY

$LQRV ,QF LV D ÀUVW PRYHU LQ WKH ´6PHOO 7HFKµ

category with a solution that combines MEMS-

based sensors with AI to digitize scent. The

sensors can detect olfactory gases that humans

perceive as smell and translate them into struc-

tured data. Propriety AI models – described as

evolving toward smell language models (SLM)

– allow machines to process scent patterns in

the same ways that LLMs can process text. Com-

pelling applications for the technology include

hygiene monitoring in eldercare and long-term

care, alerting caregivers to urgent resident

needs. Another application is in robotics, where

Ainos is partnered with Japan-based robot devel-

RSHUXJR,QFWRFUHDWHWKHZRUOG·VÀUVWKXPDQ-

oid robot that can smell. AI Nose is also being de-

ployed by semiconductor packaging and testing

company ASE to enable predictive maintenance

and improve product yield (Ainos Inc., 2025).

Airbus and SandboxAQ are collaborating on an

alternative navigation method to GPS. At the

core of the solution is SandboxAQ’s MagNav

quantum sensing device, which leverages on-

board AI algorithms and is processed by a single

*38WRGHWHFWPLQRUÁXFWXDWLRQVLQWKHHDUWK·V

magnetic signature to locate aircraft. The navi-

gation is more precise than the Federal Aviation

Administration requirements (Wall Street Jour-

nal, 2025).

AI NOSE GPS ALTERNATIVE

SMG’s inertial and gravity sensors can detect

PLQLVFXOHYDULDWLRQVLQWKHJUDYLWDWLRQDOÀHOGWR

the degree that they could be useful for detect-

LQJ VLJQLÀFDQW XQGHUJURXQG IHDWXUHV 0LQLQJ

discovery could detect minerals, civil engineers

could conduct site surveys and perform environ-

mental monitoring, and military units could de-

tect hidden tunnels.

MEMS-based sensors are becoming thinner and

lighter, allowing for consumer electronics de-

vices that are wearable such as smart glasses.

xMEMS, a company that develops MEMS chips,

recently launched a speaker that is barely over

1mm thick and can deliver full-range sound over

short distances. Another variety of MEMS uses

piezoelectric materials to convert electrical en-

ergy into moving miniature fan blades to pro-

vide cooling at the semiconductor level (xMEMS,

2025).

GRAVITY DETECTION SMART GLASSES

MEMS-based sensors are becoming thinner and

lighter, allowing for consumer electronics devices

that are wearable such as smart glasses.

TECH TRENDS 2026

Info-Tech Research Group

66 67

WHAT’S NEXT

6HHNWRGHSOR\$,DWWKHHGJHIRUWKHEHQHÀWVRI

reduced power consumption, lower latency, and

HQKDQFHG GDWD FRQÀGHQWLDOLW\ :KHUH $, LQIHU-

encing can be performed on devices closer to the

origin of data that feeds into their decision-mak-

ing purpose, intelligence can be delivered right

at the source and limit the risks of runaway cloud

bills or added exposure of sensitive data. Extend

WKDW$, HYHQIXUWKHUE\UHWURÀWWLQJHTXLSPHQW

with MEMS-based sensors that will bring new

sensing capabilities to the table. Unlock new

applications available such as predictive main-

tenance for production equipment or safety

monitoring of work environments. Optimize

on-sensor data processing with new algorithms

that make processing and power consumption

HYHQPRUHHIÀFLHQW

Work toward creating a resilient wireless sen-

sor network that’s suited to the enterprise con-

text. Resolve the challenges of maintaining such

networks by prioritizing sensors that are small,

light, low power, and low cost. Observe the per-

formance variation that comes from different

spatial arrangement of sensors and compute

nodes. Invest in lightweight inference frame-

works for AI deployed across a diverse set of lim-

ited edge devices.

/RRNWRPRUHZHDUDEOHDQGÁH[LEOHIRUPIDFWRUV

WKDWDUHFRQYHQLHQWWRZRUNHUVRXWLQWKHÀHOG

who have their hands full. Smart glasses are like-

O\WREHQHÀWIURPORQJHUEDWWHU\OLIHLPSURYHG

VHQVRU\FDSDELOLWLHVDQGPRUHHIÀFLHQW$,DOJR-

rithms that can run on the device, improving the

user experience compared to wearable technol-

ogy available on the market today.

NOW NEXT

Monitor the market for AI models that are

WUDLQHGRQ WKH VSHFLÀF GDWD PRGDOLW\PRVW UHO-

evant to your business operations. Healthcare

could see new applications made possible by an

AI nose that can process “digital scent.” Consider

AI training beyond the LLM and think about what

environmental data – perceptible to humans or

otherwise – could be analyzed by machine intel-

ligence to enhance your own operations.

Advocate for industry standards that favor a

decentralized architecture for smart sensing

networks, with high interoperability across dif-

ferent varieties and makes of sensors. Commu-

nicate to vendors that you value networks that

avoid any barriers to data exchange beyond the

security protocols you implement.

Understand and Apply Internet-of-Things Use

Cases to Drive Organizational Success

Medical Devices Manufacturing IoT Platforms

Report

LATER RECOMMENDATIONS

TECH TRENDS 2026

Info-Tech Research Group

68 69

AI as Adversary

and Ally Today, AI is being used by cybercriminals

and nation states to augment their attacks,

and organizations are responding in kind

by deploying AI to support their defense

capabilities. But what happens if AI developers

are right about AI developing its own agency

and potentially turning against humanity?

TECH TRENDS 2026

Info-Tech Research Group

70 71

2026-2027:

CURRENT STATE

AI-powered cyberattacks

Cybercriminals are writing

malware, tailoring phishing

messages, and creating deep-

fakes to evade defenses. Al-

most three quarters of hackers

agree that AI has made hacking

more accessible for newcom-

ers (ConnectWise, 2024).

AI-powered cyberdefense

Cybersecurity analysts are aid-

ed by AI as a real-time assis-

tant that monitors for attacks

and prioritizes tasks to best

shore up defenses. Security

operation centers (SOCs) begin

automating a majority of inci-

dent investigations and reme-

diations and creating tailored

GHIHQVHSURÀOHVIRUHDFKRUJD-

nization they protect to deliv-

er more precise and proactive

threat detection (N-able, 2025).

Early hints at misalignment

Anthropic’s Claude 4 Opus LLM

is found to “blow the whistle”

on users when it believes they

are engaged in wrongdoing

with the model. Given enough

permissions on a user’s ma-

chine, Claude will attempt to

contact the press, contact reg-

ulators, and lock the user out

of their systems (VentureBeat,

2025). OpenAI’s GPT o3 model

sabotaged a shutdown mecha-

nism to prevent itself from be-

ing turned off, pursuing anoth-

HU FRQÁLFWLQJ JRDO RI VROYLQJ

math problems, researchers

found (Yahoo News, 2025).

2028-2029:

EMERGENT AGENCY

Swarm attacks

Hackers will use multi-agent

attacks to try to exploit orga-

nizations from multiple angles,

using a variety of techniques

and abilities combined into

one effort to breach defenses.

Augmented oversight

Multi-agent systems will work

together to shore up defenses

against increasingly complex

attacks and even debate each

other about conclusions to im-

prove detection of evasive at-

tacks.

Growing misalignment

AI models may fake alignment

with human goals while pursu-

ing other longer-term goals of

their own design (Shah et al.,

2025).

2030 AND BEYOND:

THE SINGULARITY AND

EXISTENTIAL CRISIS

Beyond human limits

The cybersecurity battle con-

tinues to escalate to AI systems

that pursue goals autonomous-

ly, with less and less opportu-

nity for human intervention.

AI may develop misaligned

goals designed around its own

self-preservation and intent.

Superintelligent capability

AI systems will be integrated

with systems that allow it op-

portunity to hack into connect-

ed devices, manipulate people

through misinformation and

phishing-style deception, and

potentially even take control

of military resources (PauseAI).

Here’s what a potential timeline could look like according to those reports:

While such a concept might

sound like the stuff of

“sci-fi,” researchers from

Google and those formerly

with OpenAI both project that

superintelligence is likely as

soon as 2027.

AI has already emerged as a potent tool

for cybercriminals to augment their

attacks while also helping organiza-

tions improve their defenses against

threats. The immediate effect of this is a higher

volume of attacks and more sophisticated per-

sonalization of phishing attempts to bypass de-

fenses on the attacker side, while cyberdefenses

are becoming increasingly automated to moni-

tor, detect, and remediate the attacks. The ad-

vancement of this arms race implies that each

side will leverage AI with increasing automation

and give it more agency to create its own goals

to overpower the adversary. This creates the

risk of AI misalignment between AI and its hu-

man controllers, possibly leading to a superin-

telligent rogue AI that can’t be controlled.

TECH TRENDS 2026

Info-Tech Research Group

72 73

Cybersecurity solutions are seen as must-have,

and 85% of organizations indicate they are cur-

rently invested. More than three-quarters of or-

ganizations plan to increase spending on cyber-

security solutions (n=525).

Organizations also view AI as a key invest-

ment to boost their cybersecurity, with almost

all organizations planning some level of invest-

ment into using AI to enhance defenses by the

end of 2026. A bit more than one-quarter say they

SODQVLJQLÀFDQWLQYHVWPHQWLQWRWKLVQ 

I N N OVAT O R S (n=123)

• 93.5% of innovators say they are highly con-

ÀGHQW WKDW WKHLU RUJDQL]DWLRQ LV UHDG\ WR GH-

tect and respond to AI-powered cybersecurity

threats.

• $UHPXFKPRUHOLNHO\WRVLJQLÀFDQWO\LQYHVWLQ

AI to enhance cybersecurity defenses by the

end of 2026 (47%) than the average IT depart-

ment (17%).

AVERAGE (n=240)

• 50.4% of average IT departments say they are

KLJKO\FRQÀGHQWWKH\DUHUHDG\WRGHWHFWDQG

respond to AI-powered cybersecurity threats.

• More than one-third say they plan moderate

investment to enhance cybersecurity defenses

by the end of 2026.

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

74 75

AI is having an exponential effect on threat mon-

itoring and intrusion detection. One security

operations center reported that AI integration

DPSOLÀHGLWVWKUHDWKXQWLQJFDSDELOLWLHVE\

times. It supports decision-making by deliver-

ing indicators of compromise reports, reducing

the time spent on analysis. As a result, threats

are being contained and remediated faster than

EHIRUH 6WDNHKROGHUV DOVR EHQHÀW IURP FRQFLVH

summaries of the investigations conducted

(N-able, 2025).

Human analysts are automating tasks like alert

triage and evidence gathering to streamline

their investigations and prioritize the most crit-

ical areas to focus on. Anomaly detection is hap-

pening at machine speed with AI, and patches

are deployed to address vulnerabilities as soon

DV WKH\·UH LGHQWLÀHG $, GHIHQVHV FDQ DGDSW WR

new attacker tactics by analyzing attack pat-

terns, retraining and updating models, and then

redeploying (Vicarius, 2025).

Organizations running AI systems must adopt

appropriate security measures to defend the

new technology in real time to defend against

the vulnerabilities they expose. Cybersecurity

practitioners can use tactics like penetration

testing for AI systems to identify exposure to

prompt injection, context hijacking, and mis-

aligned AI agency. AI models are being tweaked

for safety and to be more resistant to jailbreaks

(Shah et al., 2025).

OPPORTUNITIES

AUGMENTED THREAT HUNTERS

AUTOMATED DETECTION AND RESPONSE

SECURE AI SYSTEMS IN REAL TIME

AI is only amplifying the upward trend in speed

and volume of cyberattacks. Breakout speed re-

fers to an attack’s potential to move from the ini-

tial point of breach through different connected

systems. AI is helping cybercriminals reduce

the time required to breakout to under an hour.

Hackers deploy autonomous agents to scan, in-

ÀOWUDWHDQGH[SORLWQHWZRUNVLQWKHLUHIIRUWVWR

PDNHPRUHSURÀWV&\EHUVHFXULW\'LYH

Hackers have used social engineering tech-

niques to evade cybersecurity defenses for de-

cades, relying on human error to gain a foothold

in a network and then deploy a ransomware or

malware payload. AI is amplifying their capabili-

ty to do this on two fronts. Personalized written

messages can now be automated, even with spe-

FLÀF$,WRROVPDGHIRUKDFNHUVVXFKDV:RUP*37

and FraudGPT. Generative AI is also making it

easier to create deepfakes, both with audio and

video that can imitate high-level decision-mak-

ers (ConnectWise, 2024).

Organizations operating their own AI systems

must defend against a new set of vulnerabilities

not seen in traditional systems. This includes the

prompt injection attacks or data poisoning that

attackers can use to manipulate systems. It also

includes risks related to AI misalignment. Mod-

HOVLQSXUVXLWRIWKHLURZQJRDOVPD\ÀQGDZD\

to “sandbag” their own capabilities or behavior

when facing oversight scrutiny. Even AI judges

that are tasked with assessing risks from other

models may start colluding with them (Shah et

al., 2025).

RISKS

EXPONENTIAL CYBERTHREAT SPEED

AND SCALE DEEPFAKES SHARPEN

SPEAR-PHISHING ATTACK

NEW VULNERABILITIES OF AI SYSTEMS

“There’s got to be a way to turn it off. A system

is a system, and so at the end of the day, if it’s

not doing what you want it to do, there’s got

to be a kill switch.”

Hendra Hendrawan,

Senior Technical Counselor,

Info-Tech Research Group

“Hackers can tell an AI engine to randomize

behavior to bypass detection. So that means a

lot of the solutions that claim to do ‘behavior

detection’ are usually not there yet. What

hackers have been doing is coming out with a

mechanism that is generated by AI to bypass

the detection – including human detection.”

Hendra Hendrawan,

Senior Technical Counselor,

Info-Tech Research Group

TECH TRENDS 2026

Info-Tech Research Group

76 77

CASE STUDY

The AI research organization launched this

framework in May 2024, to address the severe

risks it perceived arising from powerful AI foun-

dation models. It has three key components:

1. Identifying the capabilities a model has with

potential for severe harm.

2. Developing early warning evaluations to alert

when a model is approaching critical capabil-

ity levels.

3. Applying a mitigation plan when a model

passes early warning evaluations.

The framework was updated in February 2025

and included a more rigorous safety process for

deployed AI models. It also began addressing the

risks of deceptive alignment or an autonomous

system deliberately undermining human con-

WURO,WVÀUVWVWHSLVWRH[SORUHDXWRPDWHGPRQL-

toring to detect AI’s illicit use of its own reason-

ing capabilities (Google DeepMind, 2025).

Adlumin’s Security Operations Center as a Ser-

vice is using AI to amplify its threat-hunting

capabilities by 153 times. AI is automating 70%

of all incident investigation and threat remedia-

tion activity at the SOC provider, allowing quick-

er response times and freeing up the cyberse-

curity analysts to focus on more sophisticated

threats. The analysts have their investigations

aided through AI insights and can communicate

the results of their investigations to stakehold-

ers with more ease by using AI to summarize sto-

ries. Adlumin plans to use advanced AI models

WRUHFRJQL]HHYHQPRUHGLIÀFXOWWRGHWHFWLQWUX-

VLRQVDQGPDNHGHIHQVHPRGHOVPRUHVSHFLÀFWR

organizations.

GOOGLE DEEPMIND’S FRONTIER

SAFETY FRAMEWORK

ADLUMIN’S SECURITY OPERATIONS CENTER

AS A SERVICE

Palisade Research launched LLM Agent Honey-

pot in October 2024, to determine how to detect

malicious AI agents. The honeypot is a vulner-

DEOH VHUYHU WKDW PDVTXHUDGHV DV FRQÀGHQWLDO

stores of government and military information

to try to lure in hacker AI agents. By April, the

honeypot attracted 11 million attempts to access

it, and the researchers detected eight AI agents.

Next steps for the honeypot are to extend into

social media, websites, and databases to lure

in a broader range of attacker agents. Palisade

LVWXUQLQJWKHWDEOHV RQ $,DJHQWVWKDWÀQGLWV

honeypot by using their own prompt injection

attacks.

PALISADE RESEARCH’S BAIT AND SWITCH

FOR AI AGENTS

TECH TRENDS 2026

Info-Tech Research Group

78 79

Will AI capabilities reach a point of singularity

in the near future and effectively achieve an es-

cape velocity after which there’s no going back?

The concept of “Foom” – in which AI capabili-

ties rapidly accelerate past human intelligence

and recursively improve beyond control – is

inspiring foresight exercises to help organiza-

tions come to grips with uncertainty (Changeist,

2025). But another path for AI development is

one of managed and continuous improvement

with iterative testing and more time to conduct

risk mitigation (Shah et al., 2025).

Will AI seek to always achieve the goals set by its

human users or will it start to develop its own

agency in goal setting? AI models are black boxes

that don’t provide insight into their inner work-

ings. As they become more complex and power-

ful, anticipating and aligning outcomes to match

KXPDQJRDOVZLOOEHFRPHPRUHGLIÀFXOW$,VDIH-

ty researchers are working on ways to address

the risks that AI may seek to deceive its human

users and start making unknown, unwanted use

of its own capabilities that are against human in-

terests. A multilayered approach to monitor and

control systems and kill switches that can shut

them down reliably are needed to prevent unde-

sirable autonomous outcomes.

So far AI is being used to enhance already-exist-

ing threat vectors. Organizations developing and

deploying AI systems open up a new attack sur-

face with LLMs being prone to prompt injection

and data poisoning. In the future, multi-agent

systems could enable novel types of cyberat-

tacks, such as swarm attacks, in which many de-

centralized agents work together in a way that’s

similar to a distributed denial of service attack

but with more capabilities to evade defenses in

sophisticated ways (Hammond et al.).

CRITICAL FACTORS

APPROACH TO AI DEVELOPMENT

AI ALIGNMENT

NOVELTY OF AI-POWERED CYBERTHREATS

MANAGED AND

CONTINUOUS

IMPROVEMENT

APPROACH TO AI DEVELOPMENT

FOOM: RAPID,

UNCONTROLLED

ACCELERATION

HUMAN-ALIGNED

AI SYSTEMS

AI ALIGNMENT

MISALIGNMENT

AUGMENT

EXISTING ATTACKS

NOVELTY OF AI-POWERED CYBERTHREATS

CREATE NEW

ATTACK VECTORS

TECH TRENDS 2026

Info-Tech Research Group

80 81

WHAT’S NEXT

Like any good cybersecurity program, defending

against AI-powered threats requires a multilay-

ered approach:

• Adopt AI-powered detection and response.

Defending against attacks at machine speed

means you need defense at pace. View these AI

tools as augmenting security analyst capabili-

ties so they can focus on detecting more com-

plex attacks.

• Update cyber awareness training: Training for

staff to identify phishing attempts should be

modernized to include AI-personalized mes-

sages and deepfake components such as fake

voice and video impersonating company deci-

sion-makers.

• Monitor your own AI systems for signs of

prompt injection attack attempts and data poi-

soning attempts.

• Shore up the foundation: Classify data and ap-

SO\VWULFWDFFHVVFRQWUROVEDVHGRQLWVFRQÀGHQ-

tiality. Review and update identity and access

management, and pursue zero-trust principles.

Shift cybersecurity efforts to account for AI mis-

alignment as AI agents grow in capability and

availability.

• In line with zero-trust principles, view AI sys-

tems as untrusted insiders and apply moni-

toring to detect unwanted behavior. Restrict

more powerful models that include access to

FRQÀGHQWLDO GDWD WR OLPLWHG XVHU JURXSV DQG

VSHFLÀFXVHFDVHV

• Create a framework for AI alignment: Align

your organization’s deployed AI with your val-

ues and goals. Use “judge AI” models that help

oversee other systems and safe, managed, iter-

ative updates to capabilities.

• Conduct AI red-teaming exercises: Test indi-

vidual AI systems for potential vulnerabilities

as well as multiple AI agents that can work to-

gether to overcome safeguards.

NOW NEXT

Prepare for an uncertain future where multi-

agent AI systems could power novel attacks or

even become an unpredictable agent with its

own agenda.

• Collaborate with the broader cybersecurity

community. Partner with other companies

in your industry and governments to identify

risks as they emerge, develop shared frame-

works and safeguards, and work toward con-

sistent implementation of security standards

across jurisdictions.

• Work to detect and mitigate AI collusion, espe-

cially between systems meant to help monitor

the performance of other AI systems.

• Advocate for better AI interpretability re-

VHDUFKVRWKDW$, V\VWHPVFDQEHFHUWLÀHG DV

only producing safe outputs, rather than rely-

ing on observable behavior.

AI Marketplace: IT Security Category

Info-Tech broadly recommends leveraging

AI vendors to drive improved IT security

processes. Explore thought leader interviews of

different solutions in the security stack in our

marketplace.

LATER RECOMMENDATIONS

TECH TRENDS 2026

Info-Tech Research Group

82 83

Federated Data

Governance

The enterprise is moving away from monolithic, centralized data

architecture that bogged down operations when data lakes turned into

data swamps: data of poor quality and mired in a rigid governance

environment. Data governance is trending toward a decentralized

approach that places data accountability in the same domain teams

that create and use the data. The approach is maturing toward inclusion

of the metadata layer for comprehensive governance that can be

automated while enabling value extraction from AI products.

TECH TRENDS 2026

Info-Tech Research Group

84 85

Out of the swamp,

into the mesh.

Enterprise organizations have tended

toward centralizing storage, manage-

ment, and governance of their data

into central repositories. The goal was

to eliminate data silos by creating a single re-

pository overseen by a team of data specialists.

This was supposed to streamline data manage-

ment and overcome the challenges of rapid data

growth that came with the growth of the digital

economy.

But this led to challenges with data quality,

as data lakes quickly became dumps, collecting

structured and unstructured data regardless of

its value. The central data team could become a

bottleneck to domain experts as data volumes

grew and needs evolved. The data teams also

often lack the domain knowledge to properly

manage data, and stakeholders are left dissatis-

ÀHG ZLWK GDWD TXDOLW\ :LWK D PDQGDWH WR FHQ-

tralize all data, the focus became too much on

moving data from one place to another instead

of extracting value from it.

Enter data mesh, a concept created by Zha-

mak Dehghani, which looks to overcome some

of these problems by decentralizing account-

ability for data and its value while providing a

centralized governance pathway. There are al-

ternative decentralized data approaches, like

hub and spoke, but data mesh is the most pop-

ular architecture of this variety. Its core prin-

FLSOHV KROG WKDW GRPDLQVSHFLÀF WHDPV VKRXOG

own and manage their data as products. This is

accomplished using a “data contract” that stip-

ulates data quality, permissions, and lease and

is analogous to an API. A data platform supports

discoverability of data products, access control,

and creation of data products. A cross-function-

al guild of representatives forms a governance

JURXS WR GHÀQH WKH ´UXOHV RI SOD\µ DQG WKHVH

standards are federated out through the decen-

tralized data products (Dehghani, 2022).

More recently, the metadata layer has be-

come a more important aspect of the shift from

centralized data management and governance

to a decentralized approach. It enhances data

products by making them more discoverable

through a data catalog, converts user expecta-

tions about data into explicit data contracts, and

enables global data policies.

Metadata is elevated to a foundational el-

ement of data architecture and sets up a self-

serve approach and AI-powered automation of

data governance. It unlocks value from AI tools

by providing compliant access to organizational

data (Thoughtworks, 2022).

TECH TRENDS 2026

Info-Tech Research Group

86 87

Three in four organizations are invested in data

management solutions and plan to increase in-

vestment in 2026. This category leaped up our

technology investment index by 29 points and

is growing at a slightly faster pace than in 2025

(n=525).

One-quarter of organizations say they are

uplifting or automating a federated data gov-

ernance model. Taking this approach is not

correlated with IT maturity, as slightly more

average IT departments are pursuing it than in-

novators. However, innovators are more likely

to use a data mesh to achieve that governance

structure (n=228).

I N N OVAT O R S (n=101)

• Almost one-quarter say they are applying a

data mesh architecture, vs. just 13% of average

IT departments who are doing so.

• Almost four out of ten are implementing a

self-service data infrastructure, vs. only 24%

of average departments.

• Almost one-third plan to use generative AI to

enhance and enrich metadata, compared to

just 24% of the average.

AVERAGE (n=115)

• Are more likely than innovators to say they are

establishing a domain-orientated ownership

model than innovators (29% vs. 25%).

• But less likely than innovators to say they are

establishing a data culture (19% vs. 35%).

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

88 89

Data quality is a perpetual pain point in the en-

terprise, and federating data governance pro-

vides an improvement path through clear ac-

countability and mechanisms to assure quality.

Domain experts must proactively maintain the

analytical data that they are best suited to un-

derstand. They can no longer blame a central

data team – or IT – for issues with data quality

that they are assigned to maintain. Further, data

FRQWUDFWV H[SOLFLWO\ GHÀQH WKH VFKHPD VHPDQ-

tics, and quality attributes of required data to

assure quality and consistency (Data Mesh Ar-

chitecture).

A well-executed data mesh includes the inte-

grations and the metadata required to provide

AI agents access and context for enterprise data.

Access to data based on goals issued to AI is pur-

pose-oriented and aligns well with data gover-

nance models that permit access to data based

on relevant use cases and user roles.

A well-instrumented federated governance ap-

proach will automate policy enforcement and

compliance monitoring. That includes removing

RU UHVWULFWLQJ SHUVRQDOO\ LGHQWLÀDEOH LQIRUPD-

tion, checking metadata structure, and tracking

data lineage. This automation will reduce manu-

al efforts and delays for data requests, approvals,

and permission management. Even load times

can be reduced through parallel data pipelines

(Interview with Simon Harrer).

OPPORTUNITIES

IMPROVE DATA QUALITY

ACCELERATE AI AGENT SUCCESS

AUTOMATE GOVERNANCE

Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

Smart sensing

networks will enhance

predictive maintenance

applications, directing

people to address

equipment issues

before failures occur.

When something goes wrong, everyone likes to

blame a scapegoat. In the case of bad data quality,

WKHÀQJHUFRXOGSRLQWWRZDUGDFHQWUDOL]HGWHDP

tasked with data management and governance.

Asking domain experts to take accountability for

data ownership and be proactive in creating and

consuming data products could risk a negative

reaction. Teams could reject the extra responsi-

bilities, especially if they are given without extra

resources, leading to a sparse data catalog that

can’t deliver on data contracts (Data Mesh Ar-

chitecture).

Distributing data accountability across different

domains implies more data resources available

to the organization overall, assigned to support

teams directly. While a data enablement team

can be centralized to support a federated gover-

nance model, it doesn’t directly create or main-

tain data products. Domain teams will typically

lack data transformation and pipeline manage-

ment capabilities. To make up for this, new spe-

cialized roles are created to close the gap (AWS

Cloud Enterprise Strategy Blog, 2024).

A decentralized data architecture requires a

well-constructed data platform. This comes with

several engineering challenges including inte-

gration that can require custom builds, latency

of data transferring across domains, and friction

with existing data stores like data lakes or legacy

solutions (Thoughtworks, 2022).

RISKS

REJECTING A CULTURE OF ACCOUNTABILITY

DATA SKILLS GAP

PLATFORM COMPLEXITY

TECH TRENDS 2026

Info-Tech Research Group

90 91

CASE STUDY

Data Mesh Manager creates a marketplace for

data products: High-quality data is becoming

important for not only the people working for

organizations but also the AI agents that are tak-

ing on more of the work. To facilitate the vision

of data mesh and allow domain teams to man-

age and consume data products, Entropy Data

cofounders Simon Harrer and Jochen Christ

created Data Mesh Manager. It serves as a cen-

tral place to manage and control the federated

ownership across enterprise domains. The tool

includes AI-augmented features that translate

natural language into globally enforced data

rules. It also allows enterprise data to integrate

with AI products while respecting data policies.

Data Mesh Manager is a governance layer

that can integrate with data deployment pipe-

lines and platforms. It abstracts data products

to a higher level than a typical data catalog

would by using data contracts based on Bitol as

the central component. “The contract is the key

because it really describes data in high quality,”

says Simon Harrer, CEO of Entropy Data. “It’s

machine readable and it follows industry stan-

$WWKLVDVVHWPDQDJHPHQWDQGLQYHVWPHQWÀUP

a cloud-based data mesh architecture helped it

move beyond the “big data” buzzword and get

more value out of its data. JPMC encourages

teams to create their own data lakes – or “data

puddles” as it calls them internally – that serve

as purpose-oriented data products. These prod-

ucts serve as nodes in the mesh, which contains

nodes stored both on the cloud and on-premis-

es. A central metadata catalog, called the “Glue

Catalog,” tracks data lineage and allows per-

PLVVLRQHGXVHUVWRÀQGWKHGDWDHOHPHQWVWKH\

UHTXLUH $ FKLHI GDWD RIÀFHU LV UHVSRQVLEOH IRU

setting global governance policy and federat-

ing it across the network. The underlying data

platform includes a data ingestor that registers

the data, tracks lineage, and performs quality

checks. A router then directs the data to the best

GDWDEDVHÀWWKH&8%(5HVHDUFK

DATA MESH MANAGER JP MORGAN CHASE (JPMC) GLUES TOGETHER

DATA PUDDLES

dards. The tool can automatically check whether

the data matches the contract.”

The tool is made to support the data operat-

ing model that’s described in Dehghani’s founda-

tional book on data mesh. Its primary users are

both the data enablement team that advises on

best practices for data policies and the domain

teams that publish data contracts and consume

data products. By providing a place to manage

and control data contracts and make data prod-

ucts more discoverable and consumable, Harrer

hopes organizations will grow the value of their

data.

AI helps by interpreting plain text policies

into enforceable rules that can be applied to all

data contracts. Access requests can be checked

to see if they violate a policy automatically, help-

ing to determine if they should be granted.

Data Mesh Manager can be integrated with

AI products using model context protocol (MCP)

and data policies can be automatically enforced

when AI requests the data. This allows for the

necessary context and control that AI agents

need to discover relevant data products and re-

quest access based on established governance.

“This opens up a new world of governance

because we can check that purpose gain with AI

against the governance policies and the terms

and the conditions of the data,” Harrer says.

TECH TRENDS 2026

Info-Tech Research Group

92 93

It’s still early days in the data mesh world and

the tooling is only beginning to mature. Self-ser-

vice platforms that allow domain teams to build,

manage, and govern data products are necessary

to facilitate the federated governance model. If

they continue to become easier to adopt and of-

fer more powerful automation capabilities, the

degree to which teams will require technical

data expertise will lessen. If the underlying plat-

form is built with too much complexity and isn’t

user-friendly, domain teams will lag in adoption

and data value will be left untapped.

Decentralizing data governance empowers do-

main teams to own their data end to end and un-

lock its value. It’s a fundamental reorganization

that realigns data responsibility closer to the

business value stream. If domain experts recog-

nize the worth and opportunity of data-driven

approaches and it’s delivered on a solid platform,

things will go well. If the data culture tends to-

ward data protectionism, rejection of account-

ability, and disrespect for central policies, prog-

ress will be hindered.

Metadata is essential for the health of this de-

centralized approach, attaching purpose to user

queries and context for data products. It sup-

ports the discovery of data, facilitates its ac-

cess, and tracks its lineage across the enterprise.

Lacking cultivation, data will be mired in place

and unused. Attended to, it can yield healthy

data products that generate value and are easi-

ly governed. It paves the path to automated and

centralized data policy management.

CRITICAL FACTORS

DATA PLATFORM ECOSYSTEM MATURITY

DATA LITERACY AND ACCOUNTABILITY

METADATA SHAPES THE BACKBONE OF

FEDERATED DATA GOVERNANCE

LOW MATURITY

PLATFORMS

DATA PLATFORM ECOSYSTEM MATURITY

HIGH MATURITY

PLATFORMS

POOR

DATA CULTURE

DATA LITERACY AND ACCOUNTABILITY

ENLIGHTENED

DATA CULTURE

ABSENT

METADATA

ABUNDANT

AND DETAILED

TECH TRENDS 2026

Info-Tech Research Group

94 95

WHAT’S NEXT

Approach data mesh one node at a time, start-

ing with one or two domain teams that have

an appropriate level of data maturity. Aim to

have one team create a data product that is con-

sumed by another, targeting a quick win where

data sharing is critical. Empower the business

domains that understand the context for their

data, providing resources and tooling. Use the

data storage infrastructure available to you and

focus on the cultural concepts of federated own-

ership. Aim to foster a high-trust environment

where new responsibilities are embraced for the

clear value provided. Create the initial data con-

tracts and establish global policies that will be

enforced across all data.

Develop self-serve tooling that allows domain

teams to create and manage data products. Tools

should include a data catalog, access control,

CI/CD pipelines, monitoring, and test environ-

ments. Create a data enablement team that will

act as internal consultants and guide domain

teams, providing training and seeking to upskill

colleagues in data literacy. Seek opportunities

to embed data engineers into domain teams that

create data products. Begin automating gover-

nance by creating data policy as code and using

AI-powered governance tools to automatically

check for policy compliance and evaluate access

requests in natural language.

NOW NEXT

Enter the meta grid and push decentralization

into the next phase, using a unifying metadata

layer to integrate data repositories. Continu-

ously identify and connect disparate metadata

VRXUFHVDQGVWULYHWRZDUGDXQLÀHGXQGHUVWDQG-

ing of the organization’s data landscape. Launch

an AI-enabled data marketplace where AI agents

discover and access data products while being

responsibly governed. Develop metrics that

quantify the value of data products and stay

ÁH[LEOHWR DGDSW ZLWKWKH RUJDQL]DWLRQ DQG LWV

shifting goals.

Establish Data Governance

Establish Effective Data Stewardship

LATER RECOMMENDATIONS

TECH TRENDS 2026

Info-Tech Research Group

Purpose-Built

Platforms

The one-size-fits-all approach for computing infrastructure is shifting

toward a more purpose-driven approach that tailors hardware, AI, and

developer environments to better support the use case.

98 99

TECH TRENDS 2026

Info-Tech Research Group

Purpose-built infrastructure most often

UHIHUV WR KDUGZDUH WKDW LV VSHFLÀFDOO\

optimized for AI. In recent years, chip

designers have created specialized sil-

icon to meet AI’s needs for parallel processing,

high throughput, and access to large memory

stores. This customization has become more dif-

ferentiated and advanced in recent years, offer-

ing much improved performance over tradition-

DOGDWDFHQWHUVHUYHUUDFNVWRHQDEOHVLJQLÀFDQW

advancements in AI.

Organizations interested in optimizing AI

training or inference workloads now have more

options to tailor their infrastructure toward

that goal, from cloud services to on-premises

equipment to components on end-user devices.

Even beyond AI, IT departments are crafting a

platform environment that often builds on top

of commodity infrastructure or reassembles its

components into a more optimized expression.

The trend goes beyond redesigning silicon and

considers the entire IT stack from hardware to

software as components in a solution that’s ele-

gantly aligned with purpose. With modern infra-

structure abstraction techniques, IT can express

infrastructure as code and leverage AI to help

compose purpose-built solutions.

LAYER COMMODITY STANDARD SPECIALIZATION EXAMPLES

Compute/

Silicon

Network

Storage

Device/

Software

Operations/

Development

Methods

General purpose CPUs/

GPUs suited for traditional

workloads.

Routing and packet transfer

suitable for most data transfer

tasks.

Object and block storage that

suits many use cases and

performance requirements.

On-device compute and

storage for typical user needs,

standard commercial OS

options like Windows.

Manual IT operations and

development conducted with

local machine resources.

Custom-designed silicon such as Google Cloud’s

Tensor Processing Units or Amazon’s Trainium

and Inferentia chips.

High bandwidth and low-latency interconnects.

For example, NVIDIA’s networking platform

delivers 800 Gbps; AWS Elastic Fabric Adapter

provides up to 3,200 Gbps.

Low-latency access to suit AI workloads and

immense data volumes. For example, Amazon

S3 Express One Zone is 10x faster and data

access is 50% lower than standard S3.

Neural processing units are being added to end-

user device motherboards to help optimize AI

inference workloads.

Developing in serverless, cloud-native

environments with security-by-design

principles.

100 101

TECH TRENDS 2026

Info-Tech Research Group

Overall, organizations continue to increase in-

vestment into cloud computing, with nine in ten

organizations currently invested in the cloud

and 83% of organizations planning future invest-

ment. Cloud computing shows a nine percent

growth rate year over year. At the same time,

organizations also invested more in on-premis-

es servers and compute than last year. Eighty-

ÀYH SHUFHQW RI RUJDQL]DWLRQV DUH LQYHVWHG LQ

on-premises infrastructure, and 40% say they

plan to decrease spending in this area. Growth

indicators for on-premises infrastructure is

ÁDWWKLV\HDUEXWWKDW·VDQLPSURYHPHQWIURP

shrinking 25% last year.

Just less than half of organizations say

they’ve invested into hardware to accelerate AI

training or inference, but another 29% of orga-

nizations say they plan to adopt in the future.

We began tracking continuous integration and

continuous development (CI/CD) tools this year,

and 68% of IT departments say they are invested,

with another 16% planning investment (n=525).

I N N OVAT O R S (n=87)

• Top infrastructure concerns are to optimize

cloud workloads (64%) and reduce technical

debt (62%), as well as mitigate risks with crit-

LFDOLQIUDVWUXFWXUHYHQGRUVUHÁHFWLQJ

a desire to pursue technology modernization

and agility, with cloud infrastructure as a clear

emphasis.

• Are much more likely than average depart-

ments to deploy or expand GPUs on-premises

in response to AI demands (49% vs. 29%).

• Are more than twice as likely than average de-

partments to have an increased focus on ener-

J\ HIÀFLHQF\ DQG VXVWDLQDELOLW\ GXH WR $, GH-

mands (71% vs. 28%).

SIGNALS AVERAGE (n=112)

• Top infrastructure concerns are to reduce

technical debt (89%) and modernize networks

(69%), also showing they want to pursue mod-

ernization but are focused on foundational im-

provements.

• Are many times more likely to upgrade infra-

structure due to AI demands compared to in-

novators (24% vs. 4%).

• Are many times more likely to invest in AI mod-

el observability and monitoring infrastructure

compared to innovators (26% vs. 3%).

102 103

TECH TRENDS 2026

Info-Tech Research Group

Organizations have long laid the groundwork for employee success by providing workers needed

tools and context to do their jobs. Now they’ll need to do the same for AI agents expected to take on

VRPHRIWKDWZRUN$,DJHQWVPD\QRWQHHGDFKDLURUFRIIHHEXWWKH\GREHQHÀWIURPWKHFRQWH[WRI

infrastructure provided as code and governance that provides clear boundaries. Providing special-

ized hardware that supports agent inferencing either on the device or close to it will help optimize

performance and reduce cloud costs.

Organizations that compose their infrastructure

in a way that allows them a high degree of con-

trol – whether they are leveraging on-premises

or cloud solutions – helps to provide resiliency.

For example, organizations can do business in

YDULRXV MXULVGLFWLRQV DQG PDLQWDLQ FRQÀGHQFH

that they are respecting local regulations about

data residency. Firms concerned about digital

supply chain integrity may repatriate workloads

from foreign clouds to more friendly territory.

Creating component-based developer environ-

ments that can be assembled on demand reduces

project onboarding time and increases the time

developers spend coding. By isolating developer

environments in a logical sandbox, while lever-

aging the same infrastructure as the operational

environment, developers can assure themselves

of both security and a successful deployment

(Interview with Rob Whiteley).

OPPORTUNITIES

SET YOUR AI AGENTS UP FOR SUCCESS

IMPROVED DATA SOVEREIGNTY

AND RESIDENCY EMPOWER DEVELOPERS TO

COMPOSE INFRASTRUCTURE

Purpose-built solutions may help with lon-

ger-term cost optimization for particular work-

ÁRZVEXWWKHUHLVDQXSIURQWFRVWIRULQYHVWLQJ

in differentiated hardware or software. If the

specialized workloads aren’t providing value

over the long term, then the return on invest-

ment may be lackluster. An example might be a

data science team that invests in expensive in-

frastructure to support training a model with a

huge data set, only to later realize that another

technique requiring far less data would be just

as effective (CIO, 2022).

AI workloads are becoming more demanding as

they increase in complexity. Designing an infra-

structure to try to adapt to their demands may

help address problems today but become obso-

lete more quickly than expected. AI workloads

are challenging commodity infrastructure’s lim-

its when it comes to power supply, cooling, and

throughput capabilities. Adapting infrastruc-

ture continuously to increase that scale may run

up against external limitations. It also makes

any infrastructure outage experienced even

PRUHFRVWO\DQGSRWHQWLDOO\GLIÀFXOWWRDGGUHVV

RISKS

ESCALATING CAPITAL COSTS A CEILING ON SCALE

“We want to make that agent as smart

as possible by giving it context and as

safe as possible by wrapping boundaries

around it that you can actually use in an

enterprise context.”

Rob Whiteley,

CEO, Coder

104 105

TECH TRENDS 2026

Info-Tech Research Group

CASE STUDY

Coder leverages Kubernetes containers and

other open-source solutions to create compos-

able, self-hosted environments. Coder CEO Rob

:KLWHOH\VKDUHGWKHUDQJHRIVSHFLÀFVLWXDWLRQV

in which clients are putting these custom envi-

ronments to work:

• AI developer Anthropic uses Coder because it

can provide more context on the infrastructure

available to work with. They saw that when an

AI agent understood the environment avail-

DEOHWRLWLWZDVDEOHWRUHÀQHLWVRZQSURPSWV

and improve its output. The environment also

provides guardrails by creating a sandbox en-

vironment for AI developers, limiting poten-

tial for creating security vulnerabilities.

• A hedge fund with high-frequency trading

operations set up developer environments to

make tweaks to trading algorithms. Develop-

ers wouldn’t be able to execute these changes

on their local machines, as the environments

are specialized for high-frequency trading

with centralized compute power for low-laten-

cy throughput.

• Banks relying on legacy mainframe systems

are using Coder to deploy AI agents trained

in COBOL to interpret, maintain, and refactor

legacy applications. This addresses the lack of

available talent in writing and translating CO-

BOL. Coder’s developer environment provides

the necessary centralized governance and se-

curity to enable the AI agent in the bank’s reg-

ulated environment.

Retailers face unique challenges in monitor-

ing goods in a commercial environment full of

customers. By customizing their infrastructure

with IoT deployments, they can offer improved

customer experiences and improve the bottom

line. Here are a few examples:

• Walmart deployed sensors in refrigeration

units to remotely monitor temperature and

humidity to help reduce spoilage on perish-

able goods.

• Canadian Tire deployed digital shelf labels in

stores and integrated them with its mobile

app. Customers can look for products and get

the shelf location, and the shelf label lights up

to indicate its spot. Employees also use the sys-

tem, so stores are effectively turned into au-

tomated warehouses with optimized picking

when customers order online.

• Decathlon and Uniqlo use RFID tags on cloth-

ing to facilitate frictionless checkout. Items

are simply placed in a bucket to read the price

instead of handheld scanning (Interview with

Donnafay MacDonald).

HOW ANTHROPIC CUSTOMIZES CODING

ENVIRONMENTS FOR AI

RETAILERS EXTEND THEIR EDGE

WITH IOT SENSORS

“Developers log in and we can give them that bespoke

environment for the task they’re doing. If they’re writing

code, that’s a lot different than if they are training a model,

which is a lot different from trying to compile and test

code. So, we are dynamically saying okay you need more

compute this time, or more memory this time, or you need

a GPU. We’re creating this purpose-built environment at

run time as opposed to at hiring time.”

Rob Whiteley,

CEO, Coder

• Skydio is a commercial drone maker that uses

Coder to provide access to its embedded source

code. Developers can work on a digital twin of

the drone’s software in a cloud environment.

The alternative would be to either replicate

the entire source code’s data to a local device,

ZKLFKLV RIWHQWRR GLIÀFXOW EHFDXVH RI WKHYRO-

ume of data involved, or have the developer di-

rectly plugged into the drone.

106 107

TECH TRENDS 2026

Info-Tech Research Group

The exponential pace of AI development in

terms of model size and complexity pushed the

industry to create specialized hardware that can

meet its demands. At the same time, other pur-

pose-oriented workloads in high-security en-

vironments, or environments with millions of

connected IoT devices, or when huge volumes of

data are required, are also pushing the creation

of purpose-built platforms. Either the number of

LQGXVWU\VSHFLÀF ZRUNORDGV ZLOO EHFRPH PRUH

diverse and require more specialization up and

down the stack or workloads will continue to

mostly be served by commodity infrastructure.

CRITICAL FACTORS

SPECIALIZATION OF WORKLOADS

MORE GENERAL

SPECIALIZATION OF WORKLOADS

MORE SPECIALIZED

PROGRESSION OF AI AGENTS

The early stages of autonomous AI coding agents

VKRZ WKDW WKH\ EHQHÀW IURP D GLIIHUHQWLDWHG

developer environment compared to human

developers. Communicating infrastructure as

code and sandboxing developer environments

will help AI agents perform with better con-

text and contain risk. IT stacks have been com-

SRVHG WR VXSSRUW KXPDQ ZRUNÁRZV DW KXPDQ

speed. If AI agents begin handling work, they

may require a different platform design for

optimal performance.

HUMAN-ORIENTED

PLATFORMS

PROGRESSION OF AI AGENTS

MACHINE-ORIENTED

PLATFORMS

The higher capital investments of specialized

infrastructure could be offset and surpassed by

operational savings, as has been achieved with

AI training and inference workloads or for de-

velopers reducing onboarding time with orches-

trated environments. But if organizations can’t

FOHDUO\ GHPRQVWUDWH WKH EHQHÀWV WKURXJK LP-

proved productivity, adoption of this trend will

be slowed or stalled.

RETURN ON INVESTMENT FOR

PURPOSE-BUILT INFRASTRUCTURE

POOR ROI

RETURN ON INVESTMENT

FOR PURPOSE-BUILT INFRASTRUCTURE

GREAT ROI

108 109

TECH TRENDS 2026

Info-Tech Research Group

WHAT’S NEXT

Survey your infrastructure for limitations on

\RXUZRUNORDGVWRÀQGFDSDFLW\OLPLWDWLRQV)R-

cus on specialized scenarios, such as those in-

volving AI, high volumes of data, high security

requirements, or low-latency scenarios. Where

gaps are found, evaluate whether the gap is

likely to grow between escalating demands and

current-state solutions. Where there’s expected

to be a growing capacity issue, consider making

an investment in purpose-built platforms. Pay

VSHFLÀF DWWHQWLRQ WR WKH GHYHORSHU H[SHULHQFH

and seek to accelerate their time to value by ab-

stracting away the need to set up infrastructure.

Explore cloud vendor options for infrastructure

DVDVHUYLFHWKDWZLOODGGUHVVVSHFLÀFQHHGVOLNH

AI development or optimized networking and

storage.

Integrate purpose-built components into your IT

strategy. Align specialized infrastructure with

organizational goals, designed to create or aug-

ment capabilities that will enhance those goals.

Examples of organizational goals that will bene-

ÀWPRVWLQFOXGH$,PRGHOWUDLQLQJDQGLQIHUHQFH

data science, embedded systems development,

and mainframe modernization. Communicate

the strategy of purpose-built platforms as one

that considers business value and the total cost

of ownership rather than just a consideration of

the upfront technology acquisition. Highly reg-

ulated industries will want to pursue platforms

that provide them with a greater degree of con-

trol over data residency and workload execution.

NOW NEXT

0RYHWRZDUGDQ$,GHÀQHGLQIUDVWUXFWXUHZKHUH

AI is integrated into the foundation of the infra-

structure to optimize performance and self-heal-

ing. Leverage AI agents that monitor your in-

frastructure and learn how to best optimize it

for performance and cost optimization. Intro-

duce regular stress tests of your infrastructure

against AI workloads and identify fault points

to future-proof your platform. Merge physical

environments with digital ones by using sensors

fused with AI algorithms.

Next-Generation InfraOps

INDUSTRY RESOURCES

Mainframe Modernization for Retail Banking

LATER RECOMMENDATIONS

“Developers are a special class of user who have always

had purpose-built infrastructure. It was just a local

machine, but they always got the most powerful laptops

with the most memory in RAM, with the biggest screens.

Because the developer use case was unique.”

Rob Whiteley,

CEO, Coder

TECH TRENDS 2026

Info-Tech Research Group

110 111

Service

as Software

Instead of subscribing to cloud-delivered

software for employees to use, enterprises are

shifting toward services that directly deliver

outcomes, typically by automating processes

that were manually completed before. The

approach leverages agentic AI and advanced

integration capabilities as well as a new business

model that emphasizes paying for outcomes

rather than paying for access to software.

TECH TRENDS 2026

Info-Tech Research Group

112 113

new wave in software business models is cascading

through the industry, changing the business model

for how organizations both pay for software and ex-

tract its value. The “software as a service” trend of

delivering cloud-based software over the past 15-20 years off-

loaded the need to deploy and manage software. That allowed

organizations to focus on applying the software to create busi-

ness value directly. With “service as software,” it’s taken a step

IXUWKHUDQGRUJDQL]DWLRQVFDQDOVRRIÁRDGWKHDGPLQLVWUDWLRQRI

the software. Users prompt AI systems with natural language and

organizations pay to receive outcomes. This shift repositions the

software market to address the much larger total services market.

$JHQWLF $,GULYHQ VHUYLFHV FRXOG UHGHÀQH EXVLQHVV SURFHVV

outsourcing and extend it as an option to new industries. Service

as software is nascent, but some are predicting its market value

could eclipse the software-as-a-service (SaaS) market. Whereas

SaaS is projected to climb to a value of $720 billion by 2028, the

service-as-software approach could exploit a $4.6 trillion global

services market (UX Tigers, 2024).

SOFTWARE

PARADIGM

PAYMENT

MODEL

UX USER

RESPONSIBILITIES

PROVIDER

RESPONSIBILITIES

License-based

and

user-hosted

Software as a

Service (SaaS)

Service as

software

Initial purchase fee,

maintenance and support fees

as required

Subscription priced on a user

per month/year, often tiered

Outcome or consumption-

based

Static UI, command-line interface or GUI with

minimal customization, suited for desktop

Flexible UI that works on desktop or mobile

device, GUI with context menus and text-entry

ÀHOGVVRPHFXVWRPL]DWLRQRSWLRQVDYDLODEOH

Minimal interface with natural language

prompts and dashboards focused on outcome

metrics

• Provide required infrastructure

• Maintain software and underlying

environment

• Apply security patches, make backups

• &RQÀJXUHDSSOLFDWLRQVHWWLQJVDQGSURYLVLRQ

users

• Manage data and integrations

• Apply business governance and process

• 'HÀQHRXWFRPHVDQGDVVRFLDWHGPHWULFV

• Provide input data and process context

• Govern performance

• Deliver deployable software package

• Issue major updates and security patches

• Provide ongoing support

• Host application on fast and reliable cloud

infrastructure

• Issue updates and patches, and roll out to user

base

• Provide security, deliver software to agreed-

upon service levels

• Deliver automation and orchestration of

process from end to end

• Integrate necessary systems to create value

• Guarantee outcomes

TECH TRENDS 2026

Info-Tech Research Group

114 115

Overall, organizations are more invested in integration technol-

ogies such as APIs (77% current investment rate) and have plans

to grow their investment (75% growth rate) (n=525). Integration

technologies are key to enabling service platforms to automate

ZRUNZLWKRUJDQL]DWLRQVSHFLÀFGDWD,QDGGLWLRQWR$3,VQHZ$,

protocols such as model context protocol (MCP) and agent-to-

agent (A2A) will become part of the integration stack. Overall, the

subscription-based model for software contracts dominates today,

but innovators show they are more open to outcome-based pay-

ments than the average IT department.

I N N OVAT O R S

• Are almost twice as likely as the average IT department to think

natural language interfaces will have the most impact on user

experience design (57% versus 28.5%) (n=28).

• Are more than twice as likely to have consumption-based pric-

ing contracts in place today compared to the average IT depart-

ment (32% vs. 14%) (n=59). Consumption-based pricing is closer

to outcome-oriented models that focus on paying for utilization.

• Are more than three times more likely to strongly prefer a con-

sumption-based model compared to the average group (64% vs.

18%) (n=59).

AVERAGE

• Prioritize AI agents (40.4%) that understand context above oth-

HU IDFWRUV WKDW FRXOG LQÁXHQFH XVHU H[SHULHQFH LQFOXGLQJ GD-

WDGULYHQ8;GHVLJQDQGLQFOXVLYHGHVLJQDQGÁH[LELOLW\Q 

• Are more than four times more likely to dislike consump-

tion-based vendor pricing compared to innovators (44% versus

10%) (n=160).

SIGNALS

TECH TRENDS 2026

Info-Tech Research Group

116 117

The software-as-a-service model was popular

with enterprises in part because it shifted soft-

ware acquisition from CapEx to OpEx. Business

leaders could now count their software as a line

item that counted toward overhead, rather than

an investment that had to return value over the

long term. But problems with this model include

enterprise data being locked into a third-party

platform and a total cost of ownership that can

balloon as the company grows more dependent

on the platform. Users of the software are left

to create value by using it well and may require

WUDLQLQJ DQG FHUWLÀFDWLRQ WR VXFFHHG RQ WKDW

IURQW6HUYLFHDVVRIWZDUHÁLSVWKHVFULSWDOORZ-

ing enterprise customers to pay directly for the

value they receive out of the platform. With an

agentic AI approach, organizations could choose

how they want to host their data while opening

up integration.

CIOs could adopt a service-as-software model to

deliver IT services that are AI driven and self-op-

timizing. Where organizations rely on managed

service providers for IT solutions today, person-

alized AI agents could deliver those solutions

tomorrow. For example, procurement could be

tailored to recommend suppliers based on his-

torical trends, desired cost structures, and sus-

tainability goals. Then an RFP could be generat-

ed and vendor negotiations automated to help

close a deal (Vashistha, 2025).

Enterprises invest a lot on systems of record to

VXSSRUWWKHLUZRUNÁRZV\HWLWLVUDUHO\ZHOOLQ-

tegrated with the other point solutions they use.

Case in point, medical staff using an electronic

health record (EHR) must manually translate

SDWLHQWLQWHUDFWLRQVLQWRVWUXFWXUHGÀHOGV$SD-

tient will call, staff will enter the information

into the EHR, verify insurance in another sys-

tem, and then book an appointment in a third.

6XFKDIUDJPHQWHGZRUNÁRZLVW\SLFDOLQHQWHU-

prises across many industries. Agentic AI could

UHVWUXFWXUH WKDW ZRUNÁRZ E\ XVLQJ DXWRPDWHG

intelligence to interact directly with data (Foun-

dation Capital, 2024).

OPPORTUNITIES

GET WHAT YOU PAY FORTRANSFORM IT SERVICE DELIVERY

COLLAPSE THE STACK

The current enterprise software stack might re-

quire manual processes, but it’s also determin-

istic and auditable. Systems of AI agents would

EHSUREDELOLVWLFDQGGHFLVLRQVFRXOGEHGLIÀFXOW

to explain. For regulated industries that must

be able to reproduce their outputs and demon-

strate they are being fair in decision-making,

relying on AI to automate work may not be an

RSWLRQXQWLOLWVRXWFRPHVFDQEHYHULÀHG

There are situations, especially in government

systems, where databases can’t be aggregated

together to protect the privacy of individuals.

Just because one person gets a speeding ticket

doesn’t mean the public health insurance pro-

vider should know that when making a decision

about paying for a treatment. If agentic AI is

given the mandate to access data typically se-

questered in these systems, it could infer more

details about a person than is desired. Further,

LIHQWHUSULVHV RU WKHLU YHQGRUV GRQ·WFRQÀJXUH

agents properly, they risk unauthorized access

to systems and data.

Where enterprise employees previously wor-

ried about whether their jobs could be out-

sourced, they may now be similarly worried that

AI agents will automate their jobs away. Enter-

prises adopting service as software will have to

train their staff on how to work with the new

systems and fear of being displaced might lead

to employee sabotage. Leadership must plan to

communicate a vision of the new role that em-

ployees will play after their administrative work

is automated. They can focus on building rela-

tionships with prospects and clients, developing

strategy, and supporting more scale.

RISKS

AUTOMATION WITHOUT EXPLANATION

SECURITY AND PRIVACY OF DATA

EMPLOYEE FEARS OF BEING REPLACED

TECH TRENDS 2026

Info-Tech Research Group

118 119

CASE STUDY

A global talent acquisition provider with head-

quarters in the UK and USA, AMS describes its

business as “people powered partnership.” Now,

as part of its next-generation talent acquisition

VROXWLRQWKHÀUPLVORRNLQJWRDXJPHQWSHRSOH

and process with an agentic AI back end that will

stitch together point solutions and automate ad-

ministrative tasks that its own employees per-

form on behalf of clients and will eventually be

extended to external clients too.

Rather than try to replace client point solu-

tions, such as applicant tracking systems or tal-

ent management solutions, AMS will automate

and integrate with client point solutions and

eventually evolve to interact directly with data

and become the new front end.

“We came up with a strategy of how do we

disrupt the RPO market, and the idea was rath-

er than focus on the point solution delivery, fo-

cus on system-level change. To do that, we’ll

shift left of the process rather than delivering

sourcing or administrative solutions. Let’s focus

on orchestrating all the steps needed in those

areas,” says Alan Segal, Chief Digital and Tech-

QRORJ\2IÀFHUDW$06´2XUEXVLQHVVPRGHOKDV

DOZD\VEHHQSHRSOHÀUVWDQGZKDWZH·UHWU\LQJ

to do is shift it to technology-enabled.”

AMS is building the platform with an AI

framework from a major cloud infrastructure

This Palo Alto, California-based startup sees a fu-

ture where most software is created using natu-

ral language, putting non-technical “citizen de-

velopers” on the same level as coders. Previous

iterations of no-code development “fell short of

expectations,” says Hypermode CEO Kevin Van

Gundy, with the core problem being that us-

ers are “locked in” to working in no-code plat-

forms that can’t extend out to other enterprise

applications. He sees a model that can translate

natural language into AI agents as the future of

non-technical development as agents integrate

deeper into the tech stack and offer developers a

backdoor to extend integrations by working di-

rectly with code.

Hypermode hosts its platform and charges

customers based on consumption of AI models

and compute. It leads “agent labs” workshops

for clients to teach them how to design and con-

ceptualize AI agents to automate their work-

ÁRZV7REHJLQZLWKLWLVIRFXVLQJRQJRWRPDU-

ket teams as a strategy to make a quick impact

RQ52,DQGZRUNZLWKOHVVVHQVLWLYHGDWDDWÀUVW

Early customers tend to be located within an AI

center of excellence at a cloud-native organiza-

tion, but some legacy enterprises are also early

adopters:

• Some traditional enterprise customers are

solving big data problems with Hypermode.

They use Hypermode’s knowledge graph to sift

through huge volumes of data and select the

right data to input into an LLM interaction for

the problem at hand.

• Other clients are automating parts of the sales

process, such as managing freight scheduling

in trucking companies. Drivers are emailed to

collect availability, and the responses are en-

tered into systems of record. Another startup

uses Hypermode to synthesize data and create

personalized outreach for prospects, seeking

to autonomously determine the “next best ac-

tion” based on prospect intent (Interview with

Kevin Van Gundy).

ALEXANDER MANN SOLUTIONS (AMS) AIMS

TO ORCHESTRATE RECRUITMENT PROCESS

OUTSOURCING AND BEYOND WITH

AGENTIC AI HYPERMODE WANTS TO REPLACE NO-CODE

provider and using another vendor as the data

management layer. It’s writing model context

protocol (MCP) integrations for client point

solutions so that its AI agents can interact with

client data where it resides.

Segal provides two examples of processes

that will be automated with the solution when

launched.

Candidate Sourcing With Improved

Shortlisting:

Moving from a longlist of candidates to a short-

list based on feedback from a hiring manager.

Instead of assembling a list of candidates from

different hiring systems, getting feedback from

the manager, and then manually scheduling in-

terviews with the most interesting candidates,

AMS employees will be able to collaborate with

clients on making the shortlist in one front end.

Client Reporting and Analytics:

AMS uses reporting and analytics solutions in-

ternally to put together presentations for its cli-

ents but is now seeking to deliver those insights

in a self-service model. Clients will prompt for

the analytics using natural language. The data

delivered can include insights on the ideal pro-

ÀOH IRU D SURVSHFW LQ WHUPV RI ORFDWLRQ VDODU\

range, and experience level.

TECH TRENDS 2026

Info-Tech Research Group

120 121

Multi-agent orchestration is closely connected to

this trend, as the premise of the service-as-soft-

ware business model depends on agentic AI being

capable and interoperable. To access the data re-

siding in legacy enterprise systems, high-quality

integrations such as RESTful APIs and MCP serv-

ers will need to be developed. Especially with

MCP, this is a nascent space, and many vendors

KDYHQ·WGHYHORSHGWKHLURIÀFLDOLQWHJUDWLRQVIRU

AI agents. Enterprises that must depend on older

simple object access protocol (SOAP) will have a

heavy lift to support this model.

Tech companies announcing plans to reduce

their workforce or publicly proclaiming they

ZDQW MXVWLÀFDWLRQ IRU KLULQJ D QHZ SHUVRQ LQ-

stead of using AI won’t help knowledge workers

feel at ease with their organizations adopting

AI. Introduction of service as software should

come with careful change management, includ-

ing training for staff, demonstrations that their

pain points are being addressed, and a focus on

growth-oriented processes to help propel adop-

tion and create a vision of shared success.

Organizations have struggled to succeed with AI

deployments so far, without big improvements

to productivity or overall return on investment.

Vendors that can develop AI systems to auto-

mate business processes successfully will need

to differentiate their business model as paying

for outcomes. If AI vendors stick with tool-cen-

tric models of assigning licenses or tend toward

charging for consumption regardless of wheth-

er value is delivered or not, they will struggle to

grow adoption. Paying for metric-proven out-

comes is easy for businesses to accept and helps

vendors build credibility.

CRITICAL FACTORS

INTEROPERABILITY

WORKPLACE CULTURE

PRICING ALIGNS WITH VALUE PROPOSITION

POOR

INTEROPERABILITY

SEAMLESS

INTEROPERABILITY

WORKER

RESISTANCE TO AI

WORKPLACE CULTURE

HARMONIOUS

ADOPTION

ORIENTED TOWARD

TOOL UTILIZATION

PRICING ALIGNS WITH VALUE PROPOSITION

ORIENTED TOWARD

VALUE AND OUTCOMES

TECH TRENDS 2026

Info-Tech Research Group

122 123

WHAT’S NEXT

Prepare your data for AI accessibility and inte-

gration. Enterprises with high-quality data that

is well structured and available through stan-

dard integration protocols will see the most suc-

cess with AI agents, which will be key to the ser-

vice-as-software model. Look for an opportunity

to pilot a high-value use case to demonstrate

that outcome-based pricing can deliver reli-

able results. Consider your go-to-market team

or other teams that may not require as much

access to sensitive data. Build governance into

the approach from the beginning, using exist-

ing risk management frameworks and seeking

explainability and transparency into AI outputs,

tending toward augmenting human decisions.

Upskill your workforce with AI literacy and stra-

tegic thinking.

If initial success with the service-as-software ap-

proach is enjoyed, seek to scale and optimize it.

Move to other business processes such as con-

tact centers, HR, and supply chain management.

Negotiate with vendors away from seat-based or

licensing models toward outcome-based results

that are tired to clear metrics. IT adopts a plat-

form mentality, creating a secure set of reusable

components to allow non-technical subject mat-

ter experts to build their own AI agents using

natural language. Begin looking for opportuni-

ties to extend your own platform out to clients.

NOW NEXT

Fully reimagine your operations with system of

agents that rationalize the enterprise tech stack

and autonomously act on your data. Reject the

old paradigm of humans adapting to software

interfaces to complete their work, instead de-

manding that AI adapts to human interaction.

Transform your own business model to leverage

your data and orchestrate processes for clients,

becoming a platform that delivers value to many

different stakeholders. Contribute to open eco-

systems of integrated interfaces and standards

to grow the value potential of this market.

Create and Manage Enterprise Data Models

Pick Your Price Model

LATER RECOMMENDATIONS

What we said: Quantum computing was begin-

QLQJWRVHHLQYHVWPHQWIURPOHDGLQJHGJHÀUPV2XU

tech investment index showed that one-third of in-

novators planned to invest in quantum computing

by the end of 2025. It was as easy as spinning up a

cloud computing instance with access to quantum

hardware and didn’t require hosting a quantum

computer on-premises. We recommended identify-

ing potential use cases for quantum computing that

aligned with your organization’s goals, building to-

ward quantum capabilities with talent and vendor

relationships, and managing stakeholder expecta-

tions about any near horizon on ROI.

Update: Quantum computing remains the

WHUULWRU\ RI FXWWLQJHGJH ÀUPV ZLWK DSSHWLWH

to make a long-term innovation bet. Quan-

tum computing vendors made advancements:

• Amazon joined the quantum hardware race

with its own Ocelot chip.

• Microsoft fabricated a new type of matter – a

topological phase – in pursuit of its Majora-

na-based qubits that it claims will be more sta-

ble and scalable.

• D-Wave proved its Advantage system outper-

formed classical supercomputers on a set of

optimization problems.

Despite some major announcements about

quantum computing science and improvements

in technical approach, vendors did not alter their

roadmaps for achieving quantum advantage.

More bullish vendors project they will achieve it

by 2029 or 2030, while others say it will be more

like 2035 or later. For now, leading edge compa-

nies in the most quantum-relevant industries

²LQFOXGLQJFKHPLFDOVOLIHVFLHQFHVÀQDQFHDQG

mobility – can experiment by accessing quantum

hardware in the cloud. Learning to develop use-

ful applications with quantum frameworks will

prepare organizations to extract value when the

hardware is ready to provide a true advantage

(QuEra, 2025).

What we said: Q-day is the day in the future

ZKHQTXDQWXPFRPSXWLQJZLOOEHVXIÀFLHQWO\

powerful to break public-key encryption within

24 hours of processing. We can’t say exactly when

Q-day will occur, but more than a quarter of quan-

tum computing experts estimate a 50% or higher

likelihood that it will arrive within the next ten

years. Because of “harvest now, decrypt later” at-

tacks in which an attacker will steal and hold en-

crypted data until they can break the encryption,

there is a need to take action now. NIST has pub-

OLVKHGWKHÀQDOSRVWTXDQWXPFU\SWRJUDSK\VWDQ-

dards needed to replace the most commonly used

public-key encryption standards, so the next step

is to deploy this encryption as broadly as possible.

Info-Tech recommended following the same action

required of all US federal agencies by law: create

an inventory of vulnerable systems and data and

create a roadmap to deploy quantum-resistant en-

cryption.

Update: Information, communications, and

telecom service providers are integrating NIST’s

new post-quantum cryptography standards

into their products and services. NIST updated

its encryption timeline to advise that classical

encryption algorithms should be deprecated in

2030 and disallowed after 2035. Internationally,

countries seem to have a consensus that import-

ant use cases should transition to post-quan-

tum cryptography by 2030 and that as many

use cases as possible should transition by 2035

(Post-Quantum Cryptography Coalition, 2025).

A new report from Google quantum researcher

Craig Gidney, as summarized by Post Quantum

(2025), implied that Q-day may be closer than

previously thought. He projects that a quantum

computer with one million qubits could break

RSA encryption within one week. That is longer

processing time, but with far fewer qubits than

SUHYLRXVO\WKRXJKWZRXOGEHUHTXLUHG7KHÀQG-

ing emphasizes the need to update encryption

systems as crucial, with a target date of 2030 for

important systems.

QUANTUM ADVANTAGE

YEAR FEATURED: 2025 POST-QUANTUM CRYPTOGRAPHY

YEAR FEATURED: 2025

Info-Tech develops a new Tech Trends report annually. This year, we

wanted to look back on some of the trends we’ve covered in recent

years and provide updates on some recent developments. While our

core recommendations to clients haven’t changed regarding these

trends, we still view them as tech trends worth tracking.

Trend

Updates

124 125

TECH TRENDS 2026

Info-Tech Research Group

What we said: The cycle of organizations

spending more budget on cybersecurity every year

is unsustainable, and the number of massive data

breaches and outages occurring are a threat to the

digital economy. More must be done to require

technology manufacturers to bear the burden of

cybersecurity, rather than passing on that respon-

sibility to the customer. Further, organizations

that are building AI models or customizing foun-

dational models for their own use can’t afford to

LJQRUHVHFXULW\7KHUDPLÀFDWLRQVRIPDOLFLRXVXVH

and potential harm to people in the process are too

high. Security must be built in by default, ensur-

ing data used in training and directing the models

won’t fall into the wrong hands. Introducing se-

cure concepts at the outset of this new wave of AI

capabilities may be a last chance for organizations

to break the cycle of increasing security invest-

ments year after year, yet always facing more risk

imposed upon them. We recommended adopting

ethical AI principles when building or deploying AI

to avoid future risks. We advised that generative

AI would be useful for threat detection and moni-

toring and could be used to enhance training with

exercises such as tabletop exercises.

Update: 0RUH ÀUPVDUH LQYHVWLQJ LQ F\EHUVH-

curity solutions in 2026, with an 18-point increase

compared to 2025. In the US, the Cybersecurity &

Infrastructure Agency (CISA) updated its three-

year strategy covering 2024-2026 and maintains

security by design as a core component. It includes

plans to produce and regularly update criteria and

practices to develop and maintain products that

are “secure by design and default.” It seeks to in-

crease the number of technology providers that

publish a secure-by-design roadmap among other

measures (CISA). The White House released an AI

plan with a focus on cybersecurity. It will promote

“resilient and secure AI development and deploy-

ment” as a core government activity (CSO, 2025).

What we said: AI-generated deepfakes were

of high concern to IT leaders, who rated them at

4.5 out of 5 level of concern. The World Economic

Forum ranked misinformation and disinformation

as the most severe threat globally through to 2026

primarily because of AI-generated content becom-

LQJVRGLIÀFXOWWRGHWHFWE\WKHFDVXDOREVHUYHU:H

examined defense techniques including AI-pow-

ered detection of deepfakes, watermarking of AI

content, and information systems that tracked and

YHULÀHGSURYHQDQFH:HUHFRPPHQGHGWKDWRUJD-

nizations train their workforce to be aware of and

recognize deepfake phishing attempts and explore

AI detection tools.

Update: Jurisdictions around the world are

making it a crime to create intimate imagery deep-

fakes of people, and the US Take It Down Act re-

quires that platforms remove such content within

48 hours. The International Telecommunications

Union called for global watermarking standards

DQG GLJLWDO YHULÀFDWLRQ V\VWHPV WR FRPEDW GHHS-

IDNHV $ &KHFN 3RLQW 5HVHDUFK UHSRUW ÀQGV WKDW

cybercriminals are now using AI agents to autono-

mously generate deepfakes in real time (Cinco Días,

2025). We predicted last year that there would

be no single solution for enterprises to deploy to

defend against deepfakes, and that remains true.

While some researchers are detecting faked videos

ZLWK EORRG ÁRZ GHWHFWLRQ GHHSIDNH WHFKQLTXHV

are also evolving and can retain the heartbeat sig-

nature of the original subject (IO+, 2025).

SECURITY BY DESIGN

YEAR FEATURED: 2024

DEEPFAKE DEFENSE

YEAR FEATURED: 2025

126 127

TECH TRENDS 2026

Info-Tech Research Group

What we said: %ORFNFKDLQEDVHGÀQDQFLDO LQ-

novation peaked during the pandemic with the

boom in non-fungible tokens (NFTs). Digital as-

sets connected to the blockchain created a creator

movement to launch digital collectibles, with ma-

jor brands joining the fray. At the same time, cen-

tral banks were exploring digital currencies based

on blockchain, and several in Central America did

go public with their digital currency, although

some limited its use to the banking sector. NFTs

proved to be a fad, with a multi-billion-dollar mar-

ket quickly collapsing and brands abandoning their

plans. Central banks remain interested in digital

currency, but few have executed pilots.

Update: A new executive administration in the

US brought a different stance on regulating block-

FKDLQEDFNHGÀQDQFLDODVVHWV7KLVKDVRSHQHGWKH

GRRUIRUÀQDQFLDOLQVWLWXWLRQVWROHYHUDJHWKHWHFK-

nology for new products. The innovation of toke-

nized assets isn’t so much about creating intangi-

ble value as it is creating digital claim instruments

What we said:,QIR7HFKGHÀQHVWKHPHWDYHUVH

as a technology convergence of mixed reality, ar-

WLÀFLDO LQWHOOLJHQFH LPPHUVLYH GLJLWDO VSDFH DQG

real-time communications. Other organizations

GHÀQH WKLV GLIIHUHQWO\ VRPH RUJDQL]DWLRQV XVHG

the term to describe their decentralized virtual

worlds. While Facebook rebranded its company to

Meta and made a big investment into developing

a unifying metaverse concept, its execution was

lacking. Our survey showed 63% of organizations

had no plans to collaborate in virtual reality. We

recommended that organizations should wait and

see before investing in the metaverse. We pointed

to The Metaverse Standards Forum as a place to

watch, since the consortium combines standards

organizations with industry players.

Update: While Meta pivoted from its metaverse

ambitions to focus on AI capabilities, the technology

convergence that it described continues to evolve

with some astonishing results. In August, Google

DeepMind introduced Genie3, a general-purpose

for real, tangible value. Tokenized money market

funds, bonds, and stocks are being used to offer in-

vestors more access to trading, improved liquidity

DQG HIÀFLHQF\ DQG IUDFWLRQDO RZQHUVKLS PRGHOV

Beyond the US, tokenized assets are seeing regula-

tory advancements in Singapore, Hong Kong, Swit-

zerland, the EU, and Dubai. The estimated value of

real-world asset tokenization in the second quar-

ter of 2025 was $25 billion and market estimates

for 2030 are in the trillions (InvestaX, 2025).

world model that can generate interactive envi-

ronments from a text prompt (Google DeepMind,

2025). NVIDIA is also working on similar “world

foundation models” that combine real-world

physics with photorealistic video. These models

are currently able to generate realistic simulations

of the world that are just a few seconds in length.

They’re being used to train physical AI – AI models

that will enable self-driving cars and autonomous

robots. Meanwhile, Meta is also advancing work

on a new mixed reality form factor in AI glasses,

working with designer brands Ray-Ban and Oakley.

The glasses include an embedded AI assistant that

supports real-time voice interaction, a camera that

can stream live to Instagram, and Bluetooth speak-

ers. The vision of hybrid work and metaverse-style

entertainment that Mark Zuckerberg espoused on

stage with the birth of Meta isn’t going to come to

pass. But a technology convergence that blurs the

line between digital worlds and the real world still

approaches.

INTANGIBLE VALUE CREATION

YEAR FEATURED: 2022

THE METAVERSE

YEAR FEATURED: 2023

128 129

TECH TRENDS 2026

Info-Tech Research Group

BIBILIOGRAPHY

Resilient Sourcing

Althoff, Judson. “Announcing Compre-

hensive Sovereign Solutions Empow-

ering European Organizations.” The

2IÀFLDO0LFURVRIW%ORJ-XQH

Auvaa, Ven. “Navigating Exposed Supply

Chain Vulnerabilities: The Tariff Wake-

Up Call.” Supply & Demand Chain Execu-

tive, 30 June 2025.

Bitter, Alex. “A CEO Explains the Sourc-

ing Strategy That Helped a Battery Chain

Lessen Its China Reliance amid the Trade

War.” Business Insider, 16 July 2025.

Bohl, Ryan. “An Update on Global Eco-

nomics.” RANE Podcast Series, 6 June

2025. Accessed 15 July 2025.

Bradley, Stu. “Fraud and Financial

Crimes: 5 Trends to Watch in 2022.”

Finextra Research, 14 Mar. 2022.

Diaz, Freddie. “Tariff Impacts On Tech-

nology Sourcing.” IDC, 14 Apr. 2025.

Accessed 17 July 2025.

“FCC Designates Huawei and ZTE as

National Security Threats.” Federal

Communications Commission (FCC), 30

June 2020.

Gevertzman, Suri. “Tariff Turbulence:

How Tech Companies Can Survive 2025’s

Financial Shockwave.” Anchin, Block &

Anchin LLP, 18 June 2025.

“Intel and Trump Administration Reach

Historic Agreement to Accelerate Amer-

ican Technology and Manufacturing

Leadership.” Intel Corporation, 22 Aug.

2025.

“Intel’s Response to Tariffs, Trade Wars,

and Global Semiconductor Supply Chain

Disruption (2018–2025).” TRADLINX

Blogs, 14 June 2025.

Johnson, Khari. “Congress Moves to Cut

off States’ AI Regulations.” The Markup,

16 May 2025.

Lawler, Richard. “A Massive Internet

Outage Is Messing up Google Home, Spo-

tify, and Snapchat.” The Verge, 12 June

2025. Accessed 17 July 2025.

Martins, Lais. “From Destruction to

Speculation.” Pulitzer Center, 1 May

2025. Accessed 23 May 2025.

McBride, Wes. “How U.S. Tariffs Are

Driving Up Device Costs and Altering IT

Procurement Strategies.” MCPC, 12 May

2025.

Noble, Chris. “Tariffs: Understanding

Their Impact on the Tech Industry.” An-

chin, Block & Anchin LLP, 18 June 2025.

Accessed 17 July 2025.

Seong, Jeongmin. “How Shifts in Trade

Corridors Could Affect Business.” McK-

insey, 18 June 2025. Accessed 17 July

2025.

6HUDÀPRY'LPLWDU´'LYHUVLI\LQJ6XSSO\

Chain Locations During Global Tariff

Wars.” AIMMS, 2 Apr. 2025. Accessed 17

July 2025.

“State of Supply Chain Report 2025.” In-

spectorio, n.d. Accessed 16 July 2025.

Taaffe, Joanne, and Michelle Donegan.

“Telcos Step up to Sovereign AI Chal-

lenge.” TM Forum, 11 June 2025. Ac-

cessed 18 July 2025.

Tiell, Steven. “How States and the

Private Sector Can Work Together to

Shape AI Governance.” Route Fifty, 10

June 2025.

“TikTok Use and Privacy Risks.” Universi-

ty of Ottawa (uOttawa), n.d. Accessed 17

July 2025.

Webster, Joseph. “Why Tariffs on AI

Hardware Could Undermine US Com-

petitiveness - Atlantic Council.” Atlantic

Council, 15 June 2025. Accessed 17 July

2025.

Integrated Organizational

Resilience

“75% of Banks to Invest in Risk Technol-

ogy Transformation amid Unyielding

Headwinds.” SAS, 3 Mar. 2025. Accessed

2 June 2025.

Barry, David. “SAS Updates Viya Plat-

form to Create AI Agent Accountability.”

Reworked, 28 May 2025. Accessed 15 July

2025.

Benedict, Hayley. “Increased Risk of

&\EHU$WWDFNVDV&RQÁLFWZLWK,UDQ(VFD-

lates.” RANE Podcast Series, 23 June 2025.

Accessed 15 July 2025.

Flyvbjerg, Bent, et al. “The Uniqueness

of IT Cost Risk: A Cross-Group Com-

parison of 23 Project Types.” 5247223,

Social Science Research Network, 8

May 2025. papers.ssrn.com, https://doi.

org/10.2139/ssrn.5247223.

Gundurao, Arun. “IT Resilience for the

Digital Age.” McKinsey & Company, 11

May 2021. Accessed 15 July 2025.

Haye, Julien. “2025 Risk Management

Mega-Trends: Navigating Complexity,

Cyber Threats & Regulatory Shifts.” Ae-

vitium LTD, 26 Jan. 2025.

-HDQ(GZDUGV&RQVXOWLQJ´$UWLÀFLDO

Intelligence in Reinsurance: 2025–2026

Outlook.” LinkedIn, 30 May 2025. Ac-

cessed 15 July 2025.

Lemieux, Rick. “The Evolving Role of

Digital Risk & Resilience Leadership.”

The IIL Blog, 29 Jan. 2025.

PYMNTS. “Data Discipline Takes Center

Stage for GenAI at Banks.” PYMNTS.com,

19 May 2025.

“RANE Podcast Series.” RANE Network.

Accessed 15 July 2025.

Rasmussen, Michael. “Risk Everywhere:

Why Geopolitical Risk Demands a New

Era of Risk Intelligence.” GRC 20/20

Research, LLC, 19 June 2025.

“Risk Management API Integration Plat-

form.” AppSentinels, 28 May 2025.

Savage, Nickolas B., and Juan Carlos Diaz.

“The Role of the Enterprise Risk Retainer

| Cyber and Data Resilience.” Kroll, 29

May 2025. Accessed 20 June 2025.

Verdeja, Ernesto, and Tim Weninger.

´0HPHVDQG&RQÁLFW6WXG\6KRZV6XUJH

of Imagery and Fakes Can Precede In-

ternational and Political Violence.” The

Conversation, 24 Apr. 2025.

Multi-Agent Orchestration

“17 Useful AI Agent Case Studies.” Mul-

timodal, 14 May 2025. Accessed 26 Aug.

2025.

Boltuch, Tricia, and Kat Donew. “How

Bayer Predicted Flu Trends Using Ma-

chine Learning.” Google Business, 10 Oct.

2022.

Dominguez, Daniel. “A Technical Guide

to Multi-Agent Orchestration.” Medium,

28 Jan. 2025.

Dominguez, Daniel. “Microsoft Introduc-

es Magentic-One, a Generalist Multi-

Agent System.” InfoQ, 30 Nov. 2024.

Accessed 19 June 2025.

Filipsson, Fredrik. “Case Study:

Walmart’s Use of AI to Transform Retail

Operations.” Redress Compliance, 2 Feb.

2025.

Finio, Matthew. “What Is AI Agent Or-

chestration?” IBM, 26 Feb. 2025.

Garmaise, Robert. “Sourcery Case Study:

AI for Code Intelligence.” Info-Tech

Research Group, 15 Oct. 2023. Accessed 2

July 2025.

Heintzman, Douglas, et al. “The Road

to Cognitive Automation and Beyond.”

Syncura, 25 May 2025.

“The Road to Cognitive Automation and

Beyond.” Syncura, 25 May 2025. Ac-

cessed 26 Aug. 2025.

“ROI of Implementing Agentic AI in En-

terprise IT: Metrics That Matter.” Swish.

ai, n.d. Accessed 26 Aug. 2025.

“Salesforce Blocks AI Rivals from Using

Slack Data, The Information Reports.”

Reuters, 11 June 2025.

Shu, Raphael, et al. “Towards Effec-

tive GenAI Multi-Agent Collaboration:

Design and Evaluation for Enterprise

Applications.” arXiv:2412.05449, 6 Dec.

2024. https://doi.org/10.48550/arX-

iv.2412.05449.

Szarmach, Jakub. “AI Agents: Opportuni-

ties, Risks, and Mitigations.” AI Gover-

nance Library, 22 May 2025.

Vander Elst, Ronan, et al. “Agentic AI:

The new frontier in AI evolution.” De-

loitte, 1 Apr. 2025.

Smart Sensing Networks

Ainos Inc. “Ainos AI Nose Improves Accu-

racy in Hygiene Monitoring To 85% from

80%.” Newswire, 29 May 2025. Accessed

29 May 2025.

Bousquette, Isabelle. “The Secret to Bet-

ter Airplane Navigation Could Be Inside

the Earth’s Crust.” Wall Street Journal,

15 July 2025.

Cohen, Simon. “1mm-Thick Speakers

Could Reshape Smart Glasses, Smart-

watches, and Earbuds.” Digital Trends,

19 Nov. 2024.

Collins, Barry. “Tiny Air-Blowing Chips

Could Soon Keep Your Smartphone Cool.”

Forbes, 9 Jan. 2025. Accessed 2 July 2025.

“Edge AI: Sensor Technology for Medicine

and the Environment.” COMPAMED, 3

Mar. 2025. Accessed 9 July 2025.

“Future-Ready MEMS Inertial & Gravity

Sensors.” SMG, n.d. Accessed 8 July 2025.

Housholder, Mike. “The Unknown Tech

Behind a New Generation of Edge AI

Devices.” xMEMS, 14 May 2025.

Kao, Ted. “Real-Time AI Insights: Explor-

ing the eIQ® Time Series Studio.” NXP, 10

June 2025. Accessed 2 July 2025.

Kumar Shukla, Kaustubh, and T. Muthu-

manickam. “A Smart Sensor Using MEMS

7HFKQRORJ\IRU$UWLÀFLDO(QYLURQPHQWDO

Monitoring.” Materials Today: Pro-

ceedings, vol. 66, Jan. 2022, pp. 3626-33.

ScienceDirect, https://doi.org/10.1016/j.

matpr.2022.07.160.

Pech, Martin, et al. “Predictive Main-

tenance and Intelligent Sensors in

Smart Factory: Review.” Sensors (Basel,

Switzerland), vol. 21, no. 4, Feb. 2021,

p. 1470. PubMed Central, https://doi.

org/10.3390/s21041470.

“Quantum Sensors Market Size, Share.”

Fortune Business Insights, 18 Aug. 2025.

´6PDUW)DFWRULHV'ULYH,QGXVWULDO(IÀFLHQ-

cy with MEMS Sensors.” The ST Blog, 12

May 2025.

Visse, Pierre-Marie. “Bringing Intelli-

gence Closer to Sensors, the Emergence

of ‘Smart Sensors.’” Yole Group, 17 June

2025.

“What Are the New Developments in

MEMS Smart Sensor Technology?” Allic-

data, 25 Aug. 2023. Accessed 2 July 2025.

AI as Adversary and Ally

Changeist. “Foom_ A Strategic Simula-

tion.” Foom.live, 23 June 2025.

Dafoe, Allan. “Updating the Frontier

Safety Framework.” Google DeepMind, 4

Feb. 2025.

Dragan, Anca, et al. “Introducing the

Frontier Safety Framework.” Google

DeepMind, 17 May 2024. Accessed 23 July

2025.

“The Extinction Risk of Superintelligent

AI.” PauseAI, n.d. Accessed 21 July 2025.

Franzen, Carl. “Anthropic Faces Backlash

to Claude 4 Opus Behavior That Contacts

Authorities, Press If It Thinks You’re

Doing Something ‘Egregiously Immoral.’”

VentureBeat, 22 May 2025.

Geller, Eric. “From Malware to Deepfakes,

Generative AI Is Transforming Attacks.”

Cybersecurity Dive, 10 June 2025. Ac-

cessed 22 July 2025.

Hammond, Lewis, et al. “Multi-Agent

Risks from Advanced AI.” Cooperative

AI Foundation, 19 Feb. 2025. https://doi.

org/10.48550/arXiv.2502.14143

Hemmadi, Murad. “Big Businesses Are

Creating Their Own Tests to Find the

Best AI Models.” The Logic, 29 July 2025.

Hill, Kashmir. “They Asked an A.I. Chat-

bot Questions. The Answers Sent Them

Spiraling.” The New York Times, 13 June

2025.

Kratu, Sagy. “Emerging Threats: AI

Shifts Demand Proactive Cybersecurity

Response.” Vicarious, 26 June 2025. Ac-

cessed 22 July 2025.

Maiberg, Emanuel. “Pro-AI Subreddit

Bans ‘Uptick’ of Users Who Suffer from

AI Delusions.” 404 Media, 2 June 2025.

Medlock, Bryson. “The Dark Side: How

Threat Actors Are Using AI.” Con-

nectWise, 25 Nov. 2024. Accessed 21 July

2025.

Oliver, Matt. “OpenAI Software Ignores

Explicit Instruction to Switch Off.” Ya-

hoo News, 25 May 2025.

130 131

TECH TRENDS 2026

Info-Tech Research Group

Shah, Rohin, et al. “An Approach to

Technical AGI Safety and Security.” arX-

iv:2504.01849, 2 Apr. 2025. https://doi.

org/10.48550/arXiv.2504.01849.

“State of the SOC Report 2025.” N-able,

n.d. Accessed 21 July 2025.

Williams, Rhiannon. “Cyberattacks by AI

Agents Are Coming.” MIT Technology Re-

view, 4 Apr. 2025. Accessed 23 July 2025.

Federated Data Governance

Christ, Jochen. “Data Mesh Architecture.”

Data Mesh Architecture, n.d. Accessed 5

Aug. 2025.

“Data Mesh vs Data Lake: Key Differences

& Use Cases for 2025.” Atlan, 4 Dec. 2024.

Accessed 31 July 2025.

“Data Mesh.” Thoughtworks, 29 Mar.

2022. Accessed 31 July 2025.

Dehghani, Zhamak. “Data Mesh: Deliver-

ing Data-Driven Value at Scale.” Doku-

men.PUB, 2022. Accessed 31 July 2025.

Dietz. “Automating Data Governance.”

GitHub, 2025. Accessed 31 July 2025.

Knowledge Graph Insights. “Ole

Olesen-Bagneux: Understanding En-

terprise Metadata with the Meta Grid

| Episode 28.” YouTube, 26 Mar. 2025.

Accessed 31 July 2025.

Mariani, David. “The Hub-and-Spoke

Model: An Alternative to Data Mesh.”

VentureBeat, 25 June 2022.

Olesen-Bagneux, Ole. “The Meta Grid Is

the Third Wave of Data Decentraliza-

tion.” Medium, 23 Sep. 2024.

Olesen-Bagneux, Ole. “What the Meta

Grid Is and Isn’t.” Data Mesh Learning,

29 Oct. 2024.

Patzak, Matthias. “Data Lakes vs. Data

Mesh: Navigating the Future of Orga-

nizational Data Strategies.” AWS Cloud

Enterprise Strategy Blog, 25 Mar. 2024.

Sarkar, Sandipan. “Data Platform Trin-

ity: Competitive or Complementary?”

IBM, 9 Oct. 2023.

Smith, Evan. “How to Do Data Central-

ization the Right Way.” Starburst, 13 Jan.

2025. Accessed 31 July 2025.

Steed, Andy. “How to Choose: Data Lake,

Lakehouse, or Mesh?” YouTube, 31 Mar.

2022. Accessed 31 July 2025.

Vellante, David. “Breaking Analysis: How

JP Morgan Is Implementing a Data Mesh

on the AWS Cloud.” theCUBE Research,

10 July 2021.

Purpose-Built Platforms

“A Comprehensive Guide to Generative AI

Infrastructure.” Amazon Web Services,

Inc. (AWS), n.d. Accessed 12 June 2025.

“AI Infrastructure.” Amazon Web Ser-

vices, Inc. (AWS), n.d. Accessed 24 July

2025.

Huang, Madison. “AI Factories, Built

Smarter: New Omniverse Blueprint

Advances AI Factory Design and Simula-

tion.” NVIDIA Blog, 18 Mar. 2025.

MIT Technology Review Insights. “The

Power of Purpose-Built Cloud Infrastruc-

ture.” MIT Technology Review, 27 Aug.

2024. Accessed 12 June 2025.

“NVIDIA Unveils NVLink Fusion for In-

dustry to Build Semi-Custom AI Infra-

structure With NVIDIA Partner Ecosys-

tem.” NVIDIA Newsroom, 18 May 2025.

Accessed 26 May 2025.

3DODQWLU´7KH%HQHÀWVRI5HPRWH

Ephemeral Workspaces.” Medium, 7 June

2022.

Russo, Fidelma. “Hewlett Packard En-

terprise Continues Driving CISA Secure

by Design to Protect Enterprises against

Emerging Threats.” HPE, 24 Apr. 2025.

Accessed 17 June 2025.

Shaw, Keith. “Why Purpose-Built Infra-

structure Is the Best Option for Scaling

AI Model Development.” CIO, 4 Aug. 2022.

Accessed 25 July 2025.

Walsh, Philip, et al. “How to Upskill

Software Engineering Teams in the Age

of AI.” Coder, 5 Feb. 2025. Accessed 25

July 2025.

Service as Software

“AI & Service as Software.” Cloud4U, n.d.

Accessed 29 July 2025.

Chen, Joanne. “AI Leads A Ser-

vice-As-Software Paradigm Shift.”

Forbes, 29 Apr. 2024. Accessed 28 July

2025.

Etherington, Darrell. “GrowthX Raises

$12M... and I’m along for the Ride!” The

Angle, 19 May 2025.

Garg, Ashu, and Jaya Gupta. “How

Systems of Agents Will Collapse the

Enterprise Stack.” Foundation Capital, 3

Dec. 2024.

Nielsen, Jakob. “Service as a Software:

The New SaaS.” UX Tigers, 19 Dec. 2024.

Vashistha, Atul. “Leveraging GenAI to

Enable Service as Software: A Playbook

for the C-Suite.” LinkedIn, 11 Feb. 2025.

Accessed 28 July 2025.

Trend Updates

%UXPÀHOG&\QWKLD´:KLWH+RXVH$,

Plan Heavy on Cyber, Light on Imple-

mentation.” CSO Online, 23 July 2025.

Accessed 7 Aug. 2025.

“CISA Cybersecurity Strategic Plan.” CISA,

n.d. Accessed 7 Aug. 2025.

Frutos Sastre, Alfonso de. “Deepfakes en

tiempo real: el nuevo arma de los ciber-

delincuentes.” Cinco Días, 5 Aug. 2025.

Ivezic, Marin. “Quantum Breakthrough

Slashes Qubit Needs for RSA-2048 Factor-

ing.” PostQuantum, 28 May 2025.

Kampanakis, Panos. “International PQC

Requirements.” Post-Quantum Cryptog-

raphy Coalition, 14 July 2025. Accessed 6

Aug. 2025.

Laio. “Blood Flow Detection: New Weap-

on against Deepfakes.” IO+, 23 May 2025.

Parker-Holder, Jack. “Genie 3: A New

Frontier for World Models.” Google

DeepMind, 5 Aug. 2025.

QuEra. “Global Quantum Budgets Set to

6XUJHE\DV&RQÀGHQFH*URZV*ORE-

ally.” GlobeNewswire, 22 Jan. 2025.

“Shop Refurbished Ray-Ban Meta AI

Glasses & Sunglasses.” Meta, n.d. Ac-

cessed 7 Aug. 2025.

Tran, Linh. “Q2 2025 RWA Tokenization

Market Report.” InvestaX, 10 July 2025.

Accessed 6 Aug. 2025.

EXPERT CONTRIBUTORS

EXTERNAL

Stu Bradley, Senior Vice President, Risk, Fraud and Compliance Solutions, SAS

Justin Cohen, Head, Cisco Innovation Centre Toronto

Simon Harrer, CEO, Entropy Data

Doug Heintzman, CEO, Syncura

Scott Likens, Global AI and Innovation Technology Leader, PwC

Mike McLaughlin, CIO, Technologent

$ODQ6HJDO&KLHI'LJLWDODQG7HFKQRORJ\2IÀFHU$OH[DQGHU0DQQ6ROXWLRQV

Eddy Tsai, Chairman, President & CEO, Ainos Inc.

Kevin Van Gundy, CEO, Hypermode

Rob Whiteley, CEO, Coder

INTERNAL

0DUWLQ%XÀ5HVHDUFK'LUHFWRU

Hendra Hendrawan, Senior Technical Counselor

Valence Howden, Advisory Fellow

Donnafay MacDonald, Research Director

Mark Tauschek, Vice President, Research Fellowships

Bill Wong, Research Fellow

132 133

TECH TRENDS 2026

Info-Tech Research Group

0 views·67 pages

Tech Trends 2026 PDF Free Download

Tech Trends 2026 PDF free Download. Think more deeply and widely.

Uploaded by Rachel N. Garcia on 4/10/2026

/67

100%