Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF Free Download

Name: Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF
Author: sandra587

1 / 33

1 views•33 pages

Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF Free Download

Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF free Download. Think more deeply and widely.

Zühlke April 2025

Tailored,

measured,

and ethical:

Uncovering the road to

real-world AI impact

Contents

Practical usage: Custom models and proprietary data fuel AI's biggest wins ....... 8

Building a sustainable AI strategy ....................................................... 31

Conclusion ........................................................................ 30

The results: Where AI has impact today ............... 7

Disciplines: Marketing and sales teams are driving AI adoption .................... 12

Regions: Operational agility enables a US lead in global AI use ..................... 15

Sectors: Key factors driving AI success across industries ........................... 22

Generative AI: Early adoption grows .................................................... 25

Ethics in AI: Ethical frameworks are the cornerstone of successful AI ............. 27

Executive summary ......................................................... 3

Key findings and recommendations ...................................................... 5

Methodology ................................................................................. 6

New research from Zühlke and the Chair of Technology

and Innovaon Management of ETH Zurich provides fresh

insights into the adopon and impact of machine learning

and AI applicaons in Europe and the US. Analysing

responses from over 600 professionals in industries such

as consumer discreonary, informaon technology, manu-

facturing, nancials and healthcare, the research provides

aconable recommendaons for decision makers and

explores key drivers of success, remaining challenges,

and emerging trends – with a focus on ethics, operaonal

eciency, and strategic integraon.

Executive summary

Parcipants were quizzed about

▪ How their organisaons are applying

Machine Learning (ML)

▪ The business impact of these applicaons

▪ How they store and process data

▪ ML model acquision and development

▪ Model deployment, inference plaorm,

and predicon usage

▪ The ethical implicaons of ML applicaons

Customizaon and proprietary

data drive success

Finding: The most impacul AI applicaons are

built using custom models, developed in-house

or in collaboraon with partners, leveraging

proprietary and/or customer data. These

soluons are used daily, enabling connuous

renement and greater impact.

Recommendaon: Leverage internal know-how

and data to build high-quality applicaons that

are tailored to your business. Build robust data

pipelines and train models in-house or in collabora-

on with partners to address specic challenges.

Regularly update models to adapt to new data and

business requirements.

Operaonal eciency drives results

Finding: AI applicaons in areas like predicve

maintenance, workow automaon, and data

processing consistently deliver measurable

improvements in eciency.

Recommendaon: Priorise foundaonal use

cases where AI can automate tasks and streamline

operaons to achieve early wins, then scale to

more complex applicaons.

Regions: Operaonal agility enables

a US lead in global AI use

Finding: The US is ahead of Europe in AI adopon

– probably due to its tech ecosystem, access to rich

data, and quick wins in customer-facing applica-

ons. Europe matches the US in specialised areas

like manufacturing and healthcare innovaon.

Recommendaon: Leverage regional strengths by

focusing on customer-centric soluons in the US

and precision-driven applicaons in Europe. Foster

cross-regional collaboraon to balance agility with

industry-specic experse.

Sectors: Key factors driving

AI success across industries

Finding: AI adopon varies by sector, with

manufacturing priorising operaonal eciency,

healthcare focusing on diagnoscs, and nancial

services excelling in risk management and

personalisaon.

Recommendaon: Tailor AI strategies to

sector-specic priories, such as predicve

maintenance in manufacturing, compliance in

nancial services, and diagnosc accuracy in

healthcare. Use KPIs to align AI investments

with measurable industry goals.

Generave AI is transforming

workows

Finding: Adopon of generave AI is growing,

with applicaons in markeng, R&D, and customer

interacons. Around 65% of organisaons using it

consider it strategically important.

Recommendaon: Leverage generave AI to

augment exisng systems, focusing on scalable

applicaons that improve operaonal eciency.

Build soluons based on in-house data and know-

ledge to ensure high-quality output. Address

ethical and regulatory consideraons to

maximise its potenal.

Ethical frameworks enhance success

Finding: Organisaons with ethical frameworks are

33% more likely to achieve impacul AI outcomes,

fostering trust and migang risks such as bias and

privacy breaches.

Recommendaon: Implement structured

governance, including regular risk assessments,

monitoring systems, and training programs, to

embed ethics into AI processes.

Key ﬁndings and recommendations

Type of research

Quantave research: Online survey

containing mulple choice and open quesons

conducted in 2024 with 633 parcipants.

Parcipant selecon

Parcipants were selected for their experience

in implemenng and rolling-out ML systems:

Soware Developers: 30.0%

Machine learning engineers: 25.2%

Managers (General): 10.9%

Data Analysts: 8.0%

Data Managers/Leaders: 8.0%

C-Level Execuves: 4.0%

Other: 13.9%

Industry representaon

Interviewees represented a range of industries:

Consumer Discreonary: 22.7%

Informaon Technology: 22.2%

Manufacturing/Industrials: 21.2%

Financials: 11.7%

Healthcare: 9.0%

Communicaon Services: 4.1%

Others: 9.1%

Country representaon

Interviewees came from a range of countries:

US: 30.0%

UK: 25.4%

Germany: 23.7%

Austria: 10.7%

Switzerland: 9.9%

Methodology

The Oxford Diconary denes Machine Learning (ML) as

All relevant applicaons of Arcial Intelligence today are based on Machine Learning.

The object of our invesgaon

The use and development of computer systems that are able to learn

and adapt without following explicit instructions, by using algorithms and

statistical models to analyse and draw inferences from patterns in data.

Research was conducted by Zühlke and the Chair of Technology

and Innovaon Management at ETH Zurich. The goal of the

survey was to build an informed view of how rms are adopng

AI technologies and the impact they’re deriving from those

applicaons. The sectoral distribuon should allow to reect

the experiences across various types of businesses, to idenfy

commonalies, dierences, and paerns.

The results: Where AI

has impact today

The survey data revealed a detailed, nuanced

picture of AI’s real-world deployment. Analysing

the results across regions, industries, disciplines

and use cases, uncovered the following six main

insights about the relave penetraon and im-

pacts of AI applicaons across regions, sectors

and business contexts within companies.

Practical usage: Custom models and

proprietary data fuel AI's biggest wins

AI’s most impacul applicaons stem from leveraging

custom models and proprietary data, which together

form the foundaon of transformave AI iniaves.

By customising AI soluons and integrang internal

data eecvely, organisaons can achieve signicant

gains in operaonal eciency, demand forecasng,

and data processing. Decision-makers should priorise

these areas to drive measurable outcomes while foster-

ing long-term adaptability and scalability.

The crical role of proprietary data

82% of impacul AI applicaons depend

heavily on internal data, oen enriched with

customer insights.

While externally sourced data is ulised in 20.6% of

cases, its impact is amplied when combined with

internal datasets. The ability to control and curate

proprietary data provides organisaons with

unique, defensible advantages.

Custom models as enablers of success

54% of successful AI applicaons rely on custom-

trained models tailored to specic organisaonal

needs. Addionally, 60% of externally sourced

models are retrained using internal data.

Customisaon fosters beer alignment with

organisaonal objecves, enhances predicve

accuracy, and ensures models remain responsive

to evolving requirements.

Daily use establishes feedback loops

AI applicaons integrated into daily operaons

benet from constant feedback, which strengthens

model performance. As decision-makers adopt AI

for roune processes, they unlock compounding

improvements that increase eciency and adapta-

bility over me.

Top use cases: Opmising key funcons

The most successful applicaons focus on

operaonal eciency, demand forecasng,

and data processing.

Operaonal eciency: Automang processes,

resource allocaon, and real-me decision-making.

Data processing: Cleaning and integrang large

datasets to extract aconable insights.

Demand forecasng: Leveraging historical data to

predict future trends, enabling smarter planning.

Data-driven insights: What the most impactful use

cases have in common

Invest in proprietary data pipelines

Priorise the collecon, organisaon, and ulisa-

on of internal data. Proprietary data enables

unique AI soluons tailored to organisaonal

needs, creang a strategic advantage.

Ensure data governance frameworks are in place

to manage quality, compliance, and accessibility.

Customise AI to organisaonal contexts

Build and rene models in-house to address

specic challenges, such as resource opmisaon

or predicve analycs. Customised models

consistently outperform generic soluons in

delivering measurable benets.

Retrain and update AI models regularly to adapt to

new data inputs and changing operaonal realies.

Use feedback loops to drive

connuous improvement

Embed AI into daily workows to gather real-me

usage data. Feedback from regular operaon

enables dynamic model renement, ensuring

sustained relevance and accuracy.

Encourage cross-funconal collaboraon to

align AI performance with organisaonal goals.

Priorise foundaonal use cases

Start with applicaons that provide immediate and

quanable benets, such as improving internal

operaons or forecasng demand. These founda-

onal use cases oen deliver the highest return on

investment and serve as building blocks for more

advanced implementaons.

Leverage AI for operaonal scalability

Use AI to break down silos between organisaonal

funcons. For example, integrang supply chain

and operaons data can enable smarter resource

planning and reduce ineciencies.

The combinaon of proprietary data, custom modelling,

and a focus on foundaonal use cases provides a roadmap

for impacul AI adopon. By embedding AI into daily

operaons and invesng in dynamic feedback mechanisms,

decision- makers can unlock sustained improvements. With a

focus on scalability and adaptability, AI becomes a powerful

tool for enhancing decision-making and driving operaonal

excellence across diverse organisaonal contexts.

Take-aways for decision-makers

The top three categories for successful AI applicaons perform operaonal

eciency, data analycs and demand forecasng processes.

The most common and successful use cases

Top 10 ML Applicaon Types

* General heatmap: Top 10 ML applicaons across all contexts

Content generaon 014 2.5 0004.7

Customer recommendaons 1.2 33 2.8 2.1 5 0 5.2

Customer retenon 022 2.8 1.9 000

Customer service automaon 1.2 16 5.7 2.4 3.5 5.7 4.6

Data processing and reporng 2.9 19 11 612 12 12

Demand forecasng 2.1 23 08.5 9.6 4.4 4.5

Digital assistance opmizaon 08.7 3.5 02.5 3.1 6.2

Operaonal eciency 3.3 19 7.5 12 25 8.3 8.4

Predicve maintenance 3.3 0 0 3.2 15 3.7 0

Security and fraud detecon 13 0003.1 4.1 4.6

Cybersecurity

Go-to-market acvies

(markeng, sales,

aer-sales)

Human resource

and administraon

Logiscs

Manufacturing and

producon

Other (please specify)

R&D / innovaon

– 30

– 25

– 20

– 15

– 10

– 5

What the parcipants reported:

Machine learning analyses labour [across] time periods and then

generates a schedule based on most effective labour costs.

[AI] takes over the work of one or more controllers.

The model applies machine learning techniques to analyse production

line data, identify defects in real-time, and optimise the manufacturing

process. This has led to a signiﬁcant reduction in production errors

and increased efﬁciency on the shop ﬂoor.

Go-to-market acvies

(markeng, sales,

aer-sales)

Cybersecurity

Human resource

and administraon

Logiscs

Manufacturing and

producon

Other (please specify)

R&D / innovaon

Operaonal eciency applicaons opmise internal business

processes in ways that typically cannot be achieved manually at

scale. They automate tasks, schedule resources, and make data-

based operaonal decisions – oen in real me.

* Values represent weighted impact scores showing the relave importance of each ML applicaon within

organizaonal contexts. Higher numbers and darker colors indicate greater impact and adopon prevalence.

Automated data processing applicaons provide insights from

large volumes of informaon. They integrate data from mulple

sources, extract structured informaon from unstructured data,

run paern recognion, and help visualise results.

Demand forecasng models use historical data, including

external factors like market trends or seasonality, to produce

quantave predicons over specic me periods.

What the parcipants reported:

The model predicts market trends and understand demands for

products using past data and current market consumer trends.

The main use of this model is [helping us decide] how we will

create and release products in the future.

The model uses search term and algorithm-based logic to search

through multiple document sets and tag documents that may be

relevant in terms of legal disclosure. Previously this would have been

necessary [to do] by hand, adding large amounts of additional charge-

able time to a client's case. Now, with the use of ML, that time can be

better served advising the client and actually carrying out the law.

The model pulls information from previously human-entered data and

learns the process of each entry into the system. It then does trial and

error for our team to conﬁrm if the action was correct or not.

What the parcipants reported:

We utilise machine learning to optimise resource allocation for various

projects, ensuring that we have the right amount of manpower and materials.

Disciplines: Marketing and sales

teams are driving AI adoption

The study highlights that Go-to-Market acvies – comprising sales, markeng, and

aer-sales – lead in AI adopon, with 71% of surveyed organisaons having their

most impacul AI use case being implemented at least parally in these funcons

(respondents could indicate up to three areas). In comparison, manufacturing and

producon was entered by 26%, research and development (R&D) by 22%, logiscs

by 17%, and human resources (HR) by 15% of the respondents. These gures under-

score the varying degrees of AI integraon across organisaonal funcons, oering

insights into how companies leverage AI to address specic priories.

213,2

103,2

90,8

68,5

63,7

59,2

36,5

Distribuon of Organizaonal Context

Count

Organizaonal Context

025 50 75 100 125 150 175 200

Sales acvies

Manufacturing and producon

R&D / innovaon

Other (please specify)

Logiscs

Human resource and administraon

Cyber-security

The ML impact score is calculated based on an

evaluaon across organizaonal contexts such

as IT, Sales and Markeng, Logiscs, Producon,

or HR. In each area, parcipants rate the impact

of their applicaons on a scale from 0 (not used)

to 5 (very high impact).

The overall impact score for a region is then

calculated by taking these rangs from all nine

organizaonal contexts and compung their

arithmec mean. This creates a balanced

assessment that weighs each organizaonal

context equally, allowing for standardized

comparison across dierent regions and organi-

zaons.

Understanding the weighted

impact score

The impact score reects the signicance of AI

applicaons in dierent contexts by weighng

responses according to their distribuon across

organisaonal funcons and applicaon catego-

ries. For example, if a respondent menoned

one context, it is fully weighted (1.0), while

menoning two contexts assigns each a weight

of 0.5. Similarly, the weighng is adjusted for the

number of applicaon types cited, ensuring

proporonal representaon. This methodology

provides an accurate view of AI’s relevance

across mulple organisaonal contexts.

The denion and calculaon of impact scores

Data-driven insights: AI adoption by function

Go-to-Market acvies

(sales, markeng, aer-sales)

AI adopon in go-to-market funcons is driven by

the availability of structured data and the immedi-

ate benets of customer-focused models.

▪ Customer recommendaons (weighted impact

score: 32.67): Personalised AI models signicant-

ly enhance customer engagement.

▪ Demand forecasng (23.14): Enables precise

inventory management and ancipates

customer needs.

▪ Customer retenon (22.36): Idenes at-risk

customers, enabling proacve engagement.

Operaons and manufacturing

Operaonal eciency consistently ranks high,

reecng AI’s potenal to opmise workows

and minimise costs.

▪ Operaonal eciency (24.56): Streamlines

workows, reducing resource waste and errors.

▪ Predicve maintenance (15.28): Ancipates

equipment issues, minimising downme.

▪ Data processing and reporng (12.11): Integrates

real-me data to improve decision-making.

Logiscs

Logiscs funcons are increasingly adopng AI to

improve supply chain agility and delivery accuracy.

▪ Operaonal Eciency (11.86): Automates

logiscs workows for beer resource

allocaon.

▪ Demand Forecasng (8.53): Enhances supply

chain responsiveness to market changes.

▪ Logiscs Opmisaon (6.94): Improves roung

and delivery planning.

Research and development (R&D)

AI in R&D is driving innovaon through enhanced

data processing and predicve modelling.

▪ Data processing and reporng (12.22):

Consolidates datasets to drive innovaon.

▪ Operaonal eciency (8.42): Accelerates project

melines through automated processes.

▪ Healthcare monitoring (7.83): Supports innova-

on in diagnoscs and treatments.

Human resources

HR funcons are increasingly using AI for talent

management and administrave eciency.

▪ Human resources opmisaon (13.03):

Streamlines recruitment processes and

talent management.

▪ Data processing and reporng (11.28): Extracts

insights from HR data for workforce planning.

▪ Operaonal eciency (7.53): Automates roune

HR tasks, enabling strategic focus.

1. Maximise customer-focused AI

Go-to-market acvies represent the most mature area of AI adopon within

organisaons, driven by the clear benets of customer-facing applicaons like

personalised recommendaons and demand forecasng. Decision-makers

should priorise these soluons to enhance customer engagement, opmise

sales processes, and achieve measurable business outcomes. Addionally,

they should explore how predicve models for customer retenon can reduce

churn and build long-term loyalty. Aligning these tools with broader business

objecves can ensure scalability and long-term impact.

2. Expand operaonal applicaons

Operaonal eciency is a cornerstone of AI’s value proposion, with proven

benets in areas like manufacturing, logiscs, and workow automaon.

Leaders should assess their organisaon’s operaonal workows to idenfy

bolenecks or repeve tasks that can be addressed through automaon and

predicve tools. For example, adopng predicve maintenance in manufactur-

ing can prevent costly downme, while logiscs opmisaon can improve

delivery mes and reduce resource wastage. These applicaons not only drive

cost savings but also enhance responsiveness to market changes, posioning

organisaons for greater agility.

3. Embrace emerging opportunies

While AI adopon in R&D and HR seems less mature, these areas hold signi-

cant potenal for driving innovaon and strategic advantage. In R&D, AI can

accelerate product development by integrang insights from diverse datasets,

improving experimentaon, and reducing me-to-market. In HR, tools for talent

acquision, workforce planning, and automaon can help organisaons aract

top talent, enhance employee retenon, and free up HR teams to focus on

long-term strategies. Decision-makers should priorise pilot projects in these

funcons to explore AI’s value and idenfy scalable use cases.

4. Foster cross-funconal integraon

The siloed adopon of AI oen limits its potenal impact. By fostering collabo-

raon between funcons — such as logiscs, operaons, and go-to-market

acvies — organisaons can unlock synergies and create a more cohesive

approach to AI deployment. For instance, integrang demand forecasng

across supply chain and markeng funcons ensures that inventory decisions

are aligned with real-me customer demand. Leaders should invest in data-

sharing plaorms and establish cross-funconal teams to drive alignment and

maximise the organisaonal benets of AI.

Take-aways for decision-makers

Regions: Operational agility enables

a US lead in global AI use

On a global scale, our data points to a slight US lead in AI adopon. While other

naons – notably the UK and China – are hurrying to posion themselves as

global AI leaders from a governmental and economic standpoint, the US’ leader-

ship posion could be aributed to several structural advantages, including a

robust tech ecosystem, signicant venture capital investment in AI startups,

and strong collaboraons between academia and industry. The presence of

major tech hubs like Silicon Valley provides U.S.-based businesses with access

to cung-edge research, talent, and innovaon pipelines, which accelerate AI

adopon across mulple sectors.

3,52

3,18

2,97

score

DACH

Comparison of impact scores across the regions

0 10 20 30 40

The U.S. also benets from a large, diverse market, which generates rich data for

machine learning applicaons. This wealth of data allows organisaons to deploy

impacul AI soluons in areas such as go-to-market acvies (42.54%) and logiscs

(12.81%), driving strong performance in customer-facing and operaonal contexts.

European strengths and opportunies

While the U.S. leads in many metrics, other regions

demonstrate strengths that reect their unique

economic structures:

The United Kingdom is likely bolstered by its

thriving ntech sector, strong academic base,

and innovaon-friendly regulatory environment

to secure a consistent second place in ML impact

scores. Its emphasis on balanced, data-driven

applicaons seems to support steady progress

across funcons.

The DACH region (Germany, Austria, Switzerland)

showcases solid performance possibly driven by its

strong industrial base and high-quality engineering

educaon. Empirical data suggests Germany leads

in manufacturing and R&D applicaons, likely

reecng its emphasis on Industry 4.0 and preci-

sion engineering. Switzerland excels in healthcare

innovaon and R&D, while Austria demonstrates

growing strengths in Go-to-Market Acvies and

niche R&D eorts.

3,78

3,46

3,31

3,55

3,36

3,11

Sales Acvies

Manufacturing and Producon

R&D / Innovaon

Other (please specify)

Logiscs

Human Resource and Administraon

Cyber-Security

All impact scores: US vs UK vs DACH

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

United States U.K. DACH

Average Score

3,51

3,73

2,44

2,61

2,68

3,60

3,14

3,34

3,62

3,06

3,19

3,65

3,55

3,02

United States

U.K.

Germany

Switzerland

Austria

Comparison of top ve organizaonal contexts across ve countries

0510 15 20 25 30 35 40 45

Percentage

Sales acvies

Manufacturing and producon

R&D / innovaon

Other (please specify)

Logiscs

Human resource and administraon

Cyber-security

Data-driven insights: AI adoption by region

United States: Leading with quick wins and customer-centric AI

The United States shows a strong emphasis on customer-facing acvies and

logiscs in ML applicaons. The strong presence of ML in go-to-market acvies

suggests a focus on customer-centric applicaons, while the high ranking of

logiscs indicates the importance of supply chain opmizaon in the US market.

Insight: The U.S. excels in AI deployment for go-to-market acvies (42.54). This

reliance enables quick implementaon and rapid ROI.

The U.K.: Data-driven and customer-focused

The United Kingdom shows a balanced distribuon of ML applicaons across

dierent organizaonal contexts. Go-to-market acvies lead as the primary

context for ML followed by a notable emphasis on human resource and

administraon, which is higher than in most other countries. Manufacturing

and producon and logiscs also play signicant roles, indicang a balanced

approach to ML applicaon across various business funcons.

Insight: U.K. organisaons avoid over-reliance on quick wins, integrang AI

steadily across funcons. However, manufacturing AI adopon is less mature

compared to other regions.

Ranking of organizaonal contexts for ML applicaons in the United States

Rank Organizaonal Context Percentage

1Go-to-market acvies 42.54%

2Manufacturing and producon 15.26%

3Logiscs 12.81%

4R&D / innovaon 7.98%

5Human resources and administraon 6.58%

6Cyber-security 6.23%

-Other (please specify) 8.60%

Ranking of organizaonal contexts for ML applicaons in the United Kingdom

Rank Organizaonal Context Percentage

1Go-to-market acvies 32.71%

2Human resources and administraon 16.36%

3Manufacturing and producon 13.98%

4Logiscs 11.70%

5R&D / innovaon 9.42%

6Cyber-security 5.90%

-Other (please specify) 9.94%

Germany: Industrial excellence and R&D leadership

Germany's ML applicaon landscape shows a balance between customer-

facing acvies, manufacturing, and research & development, reecng the

country's diverse economic strengths. Go-to-market acvies are followed

closely by manufacturing and producon and R&D / innovaon. This distribu-

on reects Germany's strong focus on industrial producon, customer

engagement, and technological innovaon.

Insight: Germany leads in precision-driven AI applicaons for manufacturing

and R&D, reecng its strong industrial base. However, customer-facing AI

adopon is comparavely slower. Expanding Go-to-market AI even further by

exploring “quick-win” tools in sales and markeng could help Germany com-

pete more eecvely in consumer-driven markets while complemenng its

operaonal strengths.

Switzerland: R&D leadership and healthcare innovaon

Switzerland's ML applicaon landscape shows a strong focus on Go-to-market

acvies and R&D / innovaon, with a notable presence in logiscs. This

balance aligns with Switzerland's reputaon for both strong market presence

and cung-edge research and development. The signicant presence of

logiscs also highlights the importance of supply chain opmizaon in the

Swiss economy.

Insight: Switzerland excels in R&D, parcularly in high-value sectors like

healthcare, but lags in manufacturing AI adopon (8.99%). Expanding produc-

on-focused AI could further enhance industrial compeveness and comple-

ment its R&D leadership.

Ranking of organizaonal contexts for ML applicaons in Germany

Rank Organizaonal Context Percentage

1Go-to-market acvies 29.67%

2Manufacturing and producon 21.67%

3R&D / innovaon 18.56%

4Human resources and administraon 8.22%

5Logiscs 6.22%

6Cyber-security 5.00%

-Other (please specify) 10.67%

Ranking of organizaonal contexts for ML applicaons in Switzerland

Rank Organizaonal Context Percentage

1Go-to-market acvies 31.75%

2R&D / innovaon 27.25%

3Logiscs 11.11%

4Manufacturing and producon 8.99%

5Cyber-security 8.73%

6Human resources and administraon 1.59%

-Other (please specify) 10.58%

Austria: Growing focus on Go-to-market AI and specialised R&D

Austria's ML applicaon landscape shows a strong focus on Go-to-market

acvies and R&D, with a notable presence in manufacturing and producon.

This distribuon suggests a balanced approach to ML applicaon in Austria,

with a focus on both customer-facing funcons and innovaon.

Insight: Austria demonstrates signicant strength in Go-to-market AI, focusing on

customer engagement and sales opmisaon. Broadening logiscs and produc-

on AI adopon could further integrate AI capabilies across its economy,

enhancing overall compeveness.

Ranking of organizaonal contexts for ML applicaons in Austria

Rank Organizaonal Context Percentage

1Go-to-market acvies 35.05%

2R&D / innovaon 19.12%

3Manufacturing and producon 14.22%

4Human resources and administraon 7.84%

5Logiscs 3.68%

6Cyber-security 0.98%

-Other (please specify) 19.12%

1. Tailor AI strategies to regional strengths

U.S.: Sustain leadership in go-to-market acvies (42.54%) by invesng in

innovave R&D (7.98%) to balance quick wins with long-term capacity

building. Focus on scaling bespoke AI soluons for complex challenges in

manufacturing and sustainability.

U.K.: Build on HR leadership (16.36%) and data-driven pracces to increase

AI maturity in manufacturing (13.98%). Deploy predicve maintenance and

operaonal eciency tools to strengthen its compeve posion in

industrial AI.

Germany: Broaden go-to-market AI applicaons (29.67%) to capture more

customer-centric opportunies, complemenng its strengths in manufac-

turing (21.67%) and R&D (18.56%). Priorise expanding AI-driven personali-

saon and sales tools.

Switzerland: Leverage R&D dominance (27.25%) to explore AI applicaons

in manufacturing (8.99%) and logiscs (11.11%). Focus on sustainable AI

innovaons, especially in healthcare and supply chains, to align with global

ESG priories.

Austria: Strengthen logiscs-focused AI (3.68%) to opmise Austria’s role

as a key hub in Europe. Diversify its R&D porolio (19.12%) to include

applicaons in healthcare, AI-driven automaon, and sustainability.

2. Balance quick wins with strategic growth

Quick wins through o-the-shelf soluons, like those potenally driving

U.S. leadership in go-to-market AI, enable fast ROI but risk neglecng

innovaon. Decision-makers in all regions should align investments in

foundaonal AI applicaons, such as operaonal eciency, with strategic

priories like R&D and sustainability-focused AI.

3. Foster cross-regional learning

Sharing insights across regions can accelerate AI maturity. The DACH

region’s governance focus can inspire others, while the U.S.'s speed in

implementaon oers lessons for European markets.

Take-aways for decision-makers

Sectors: Key factors driving

AI success across industries

On a sector-by-sector basis, machine learning impact scores reect varying

levels of tech literacy, data usage, and regulatory constraints.

Operaonal eciency is a universal driver across sectors, complemented

by highly tailored use cases that align with specic business goals and

challenges. For businesses looking to put AI to work, understanding the

key performance indicators (KPIs) that make their sector ck is central to

idenfying priority areas. This will provide measurable benchmarks for

evaluang AI’s impact and help surface new use cases.

Impact scores by region and sector (top 5 sectors)

Sector US UK Germany Austria Switzerland

Consumer discreonary 4,3 4,1 4,07 4,05 4,04

Informaon technology 4,5 4,4 4,31 4,29 4,28

Industrials 4,1 43,96 3,94 3,93

Financials 4,1 43,93 3,91 3,9

Healthcare 3,8 3,8 3,69 3,67 3,66

KPIs serve as a strategic compass, guiding decisions

on where to focus investments and how to measure

success. They oer insight into core values and

competencies, such as customer sasfacon,

technical excellence, or eciency. By aligning AI

strategies with these priories, leaders can maxim-

ise the relevance and eecveness of their AI

applicaons.

The study reveals how KPIs dier across sectors,

reecng industry-specic challenges and goals:

▪ Healthcare industries priorise paent outcomes

and diagnosc accuracy, underscoring a mission

built around improving care delivery.

▪ Financial services (FSI) focus on risk management

and customer experience, shaped by regulatory

demands and the need for personalisaon.

▪ Manufacturing places operaonal eciency at

the forefront, leveraging AI to opmise process-

es and reduce costs.

This alignment between KPIs and AI applicaons

ensures that technology investments deliver

tangible, sector-relevant results.

Why KPIs matter

Manufacturing (industrials)

Manufacturing achieves signicant AI success

through operaonal eciency (35.5%) use cases,

oen leveraging the technology for predicve

maintenance and supply chain opmisaon.

▪ Impact scores: US: 4.09, DACH: 4.03, showcasing

strong (but regionally-varied) adopon.

▪ Business (59.7%) and technical (40.3%) KPIs take

priority, showing the sector’s dual focus on

eciency and quality.

Healthcare

Healthcare organisaons focus on paent out-

comes (27.9%) and diagnosc accuracy (22.1%),

using AI for diagnoscs, treatment opmisaon,

and resource allocaon.

▪ Impact scores: US: 3.84, DACH: 3.69, reecng

slower adopon compared to other sectors.

▪ Eciency accounts for 15.3% of KPIs, under-

scoring the importance of opmising care

delivery.

▪ Business KPIs (60.4%) and technical KPIs (39.6%)

show a balanced approach to measuring both

paent outcomes and diagnosc performance.

Financial services (FSI)

The FSI sector priorises risk management (24.3%)

and customer experience (16.1%). AI applicaons

here include fraud detecon, compliance automa-

on, and personalised banking soluons.

▪ Impact scores: US: 4.06, UK: 3.99, DACH: 3.93,

demonstrang consistent performance.

▪ Business KPIs dominate at 67%, reecng the

sector’s emphasis on measurable customer and

nancial outcomes.

Consumer discreonary

This sector priorises customer experience

(26.2%), reecng its focus on enhancing sasfac-

on and driving sales of non-essenal goods and

services. Eciency is also crical, accounng for

23.8% of cited KPIs.

▪ AI applicaons include markeng personalisa-

on, demand forecasng, and inventory

management.

Informaon technology

The IT sector combines a focus on technical quality

(29.7%) with operaonal eciency (25.3%),

reecng dual priories around reliability and

innovaon.

▪ Applicaons range from soware opmisaon to

predicve analycs and automated tesng.

▪ With a balanced KPI rao of 55.1% business-

centric to 44.9% technical, IT demonstrates how

technical excellence can directly facilitate

economic growth.

Data-driven insights:

Sector-speciﬁc AI adoption and KPIs

1. KPIs as a guide to AI strategy

Manufacturing: Eciency-focused applicaons improve producon processes

and reduce costs.

Healthcare: Technical KPIs like diagnosc accuracy are crical to delivering

paent-centric outcomes.

FSI: Business KPIs like risk management align AI with the need for regulatory

compliance and customer sasfacon.

2. Balancing business and technical KPIs

KPI raos for these respecve priority areas highlight sector-specic goals:

1.48:1 in manufacturing, 1.52:1 in healthcare, and 2.03:1 in FSI.

3. Operaonal eciency is a universal goal

AI leaders across sectors should priorise operaonal eciency as a proven

entry point for delivering ROI.

4. Tailored AI strategies ensure success

Each sector can benet from aligning AI uses with its unique KPI priories,

ensuring that technology investments meet specic industry needs.

Take-aways for decision-makers

Generative AI: Early adoption grows

Generave AI (GenAI) is increasingly adopted as a strategic tool to enhance

operaons, streamline workows, and drive innovaon. While 21.6% of

surveyed organizaons have implemented generave AI technologies, most

deployments integrate these capabilies into broader systems rather than

relying on standalone soluons (72.3% of cases).

Despite regulatory and ethical challenges, the data underscores the growing

relevance of GenAI: 67.9% of generave AI applicaons are considered

strategically important, with 57.7% described as crical to business operaons.

1. Adopon paerns

▪ Prevalence of generave AI: 21.6% of sur-

veyed organizaons are leveraging generave

AI. However, only 27.7% of these applicaons

focus primarily on generave AI capabilies,

while 72.3% contain generave AI as a compo-

nent.

▪ Adopon by organizaonal contexts: Genera-

ve AI is most prevalent in funcons like

markeng (27%), R&D (26%), and HR (25%),

where it supports tasks such as content

generaon, innovaon acceleraon, and

process automaon.

▪ Adopon by sector: While sectors like con-

sumer discreonary (27%), IT (23%) and

industry/manufacturing (22%) lead in adopon,

use cases vary widely, reecng the adaptabil-

ity of generave AI technologies to dierent

operaonal needs.

▪ Technology plaorms: Adopon is facilitated

by major plaorms like OpenAI, Microso

Azure, and AWS, with OpenAI leading in

primarily generave AI applicaons (32.6%).

2. Strategic value

▪ Generave AI applicaons are regarded as

crical or very strategic by 67.9% of respond-

ents. Primarily generave AI applicaons are

even more likely to be classied as “very

strategic” (43%) than those containing genera-

ve AI (28.3%) indicang that organizaons

place higher strategic value on applicaons

where generave AI is the core funconality.

▪ Operaonal impact: 48.2% of applicaons

generate outputs daily, highlighng the deep

integraon of generave AI into core business

processes. For example, one respondent

described their use of generave AI to gener-

ate tailored customer service recommenda-

ons based on historical data.

3. Prevalent applicaons

▪ Content generaon: This dominates primarily

generave AI applicaons, accounng for

52.4% of use cases. Examples include automat-

ing markeng content creaon and generang

detailed reports based on raw data.

▪ Predicve maintenance and process opmiza-

on: Applicaons integrang generave AI

into broader ML systems, such as forecasng

or operaonal analycs, demonstrate signi-

cant ROI and eciency gains.

▪ Customer interacon enhancement: Genera-

ve AI supports customer service funcons,

with 14.7% of applicaons leveraging these

capabilies to augment tradional approaches.

Data-driven insights: How companies use GenAI

1. Maximize impact through integraon

Hybrid applicaons, where generave AI augments exisng ML systems,

are proving highly eecve. For example, 61.3% of generave AI applica-

ons address mulple machine learning problems simultaneously,

demonstrang their ability to enhance broader workows.

Organizaons should priorize use cases with measurable impacts, such

as reducing manual eort in reporng or generang customer insights

from complex data.

2. Priorize scalable applicaons

Scalable applicaons include content automaon and operaonal

opmizaon, where generave AI can signicantly lower costs and

improve producvity. Examples include predicve maintenance systems

that leverage generave AI to interpret operaonal data in real me.

3. Adapt to a dynamic ecosystem

Staying ahead of technological advancements and regulatory changes is

crical. Regularly reassess deployment strategies and foster internal

experse to ensure readiness for future developments.

Take-aways for decision-makers

Ethics in AI: Ethical frameworks are

the cornerstone of successful AI

Our research shows that the presence of a robust ethical framework

correlates with successful AI adopon and implementaon.

Ethics in AI encompasses several crical dimen-

sions that drive long-term success, including

transparent data pracces, proacve bias miga-

on, and equitable decision-making processes. By

embedding these principles into their AI strategies,

organisaons can align technological innovaon

with responsible business conduct, underscoring

the synergy between ethical governance and

sustainable operaonal excellence.

The survey explored the presence of ethical

frameworks in companies, examining whether

these frameworks are movated by regulaons or

voluntary iniaves and their scope. It also as-

sessed their inuence by idenfying cases where

AI applicaons were cancelled or postponed due to

ethical concerns.

As expected, companies in heavily regulated

industries are more likely to adopt formalised ethical

standards. US organisaons lead slightly in formali-

saon compared to their European counterparts,

likely due to a more ligious legal environment or

their overall advancement in AI deployment.

42.2

28.3

29.5

Boom 25%

Smallest

IQR

Top 25%

Largest

Percentage

0 20 40 60

No Not Sure Ye s

28.7

26.5

44.8

18,7

12.7

68.7

1. Ethical framework adopon

▪ Adopon rates: Nearly half of surveyed

organisaons (47%) report having an ethical

framework in place.

▪ Key features: Ethical frameworks commonly

include policies (72%), protocols and monitor-

ing systems (59%) and strategies (40%).

Organisaons with these elements are beer

posioned to handle the complexies of AI.

▪ Tangible benets: Organisaons with ethical

frameworks demonstrate a signicantly higher

prevalence of formalised governance process-

es, such as risk assessments (60%) and moni-

toring systems (69%). These pracces are

closely associated with more successful AI

implementaon.

▪ Ethical and regulatory preparedness: Organi-

zaons that establish ethical guidelines are

more likely to succeed. In the U.S., 56% of rms

have implemented ethical frameworks for AI,

compared to 40% in the DACH region, under-

scoring regional variaons in AI maturity.

2. Challenges without ethical frameworks

▪ Privacy and security risks: Across sectors,

privacy and security risks are the most fre-

quently reported adverse eects, with notable

prevalence in industries handling sensive

data. For example, 31% of respondents agged

this concern in nance, and 24% in communi-

caon services.

▪ Bias and discriminaon: Data bias and discrim-

inatory outcomes emerge as signicant risks,

parcularly in consumer-facing sectors.

Financial services report the highest incidence

(22%), followed closely by consumer discre-

onary industries (20%).

▪ Unintended consequences: Unintended

outcomes – such as misaligned decision-mak-

ing and operaonal disrupons – are a notable

challenge, parcularly in communicaon

services (19%) and consumer discreonary

sectors (17%).

▪ Loss of Human Control: Automaon introduc-

es concerns about reduced human oversight,

with reported rates ranging from 7% in

communicaon services to 13% in nancials

and industrials.

▪ Decision-Making Challenges: Poor deci-

sion-making linked to opaque AI processes is

observed across industries, with rates peaking

at 19% in nancial services.

3. Adverse impacts

▪ Organisaons report unintended consequenc-

es, around 20% of the respondents even

idened concrete issues like data privacy and

security breaches (15%) or operaonal ine-

ciencies (10%).

Data-driven insights: Why ethical

frameworks matter

Results for queson 7.6 (migaon steps) comparing companies

with and without an ethical framework

Category Mean (“yes group”) Mean (“no group”)

Training 0.8 0.52

Monitoring 0.69 0.52

Risk assessment 0.6 0.32

Educaon 0.6 0.39

Other 0.02 0.04

1. Build and instuonalise ethical frameworks

Establish structured ethical governance with policies, monitoring systems,

and training programs.

Conduct regular reviews and audits to idenfy

and address biases, risks,

and privacy challenges.

2. Foster interdisciplinary collaboraon

Ensure collaboraon across technical, legal, and strategic teams to

tackle fairness, transparency, and accountability.

Incorporate diverse perspecves in model design to migate bias

from the outset.

3. Embrace transparency and explainability

Implement explainable AI techniques to build trust and facilitate clear

communicaon with stakeholders.

Align AI systems with legal and ethical guidelines to ensure compliance

and foster long-term credibility.

Take-aways for decision-makers

Conclusion

AI is no longer used in prototypes and innovaon

projects only — it is a transformave force re-

shaping industries. Organisaons that integrate

AI strategically, ethically, and with a focus on

long-term scalability are achieving compeve

advantages. This conclusion oers an extended

roadmap for decision-makers to guide their AI

iniaves eecvely.

1. Customizaon drives impact

Successful AI applicaons rely on custom models

built using proprietary data. This approach not

only ensures alignment with specic business

needs but also fosters daily use and connuous

renement.

▪ Next step: Leverage proprietary data and

in-house experse to develop AI soluons

tailored to your unique challenges and oppor-

tunies.

2. Operaonal eciency as a foundaon

AI’s ability to streamline processes and reduce

costs makes operaonal eciency an ideal

starng point. Decision-makers should focus on

applicaons like predicve maintenance, demand

forecasng, and process opmisaon, which

deliver quanable benets.

▪ Next step: Use these early successes as proof

points to secure further investment and

expand AI adopon into more complex areas.

3. Proprietary data as a strategic asset

Proprietary data is a key dierenator, enabling

organisaons to create tailored AI soluons that

are dicult to replicate. Data pipelines must

priorise quality, compliance, and accessibility to

support impacul AI use cases.

▪ Next step: Invest in advanced data governance

frameworks and cross-funconal teams to

ensure data remains a strategic asset for

innovaon.

4. The imperave of ethical AI

Ethics in AI is not a regulatory checkbox but a

driver of trust and sustained success. Organisa-

ons that lead in AI adopon integrate ethical

consideraons into every stage of development

and deployment.

▪ Next step: Establish or rene ethical frame-

works, incorporang bias migaon, explaina-

bility, and energy eciency to address emerg-

ing challenges.

5. Generave AI as an innovaon catalyst

Generave AI is proving invaluable in areas like

content creaon, innovaon acceleraon, and

customer engagement. However, its strategic

value depends on responsible integraon into

broader workows.

▪ Next step: Evaluate where generave AI can

enhance exisng systems, ensuring scalability

and compliance with ethical guidelines.

6. From quick wins to strategic maturity

Balancing short-term eciencies with long-term

investments is essenal for organisaons to stay

compeve. Priorise use cases with the

greatest potenal and measure them against

relevant KPIs.

▪ Next step: Develop a roadmap for AI adopon

that includes phased investments in technology,

talent, and infrastructure.

Building a sustainable AI strategy

Responsible for the research

is Professor of Technology and Innovaon

Management at ETH Zurich and leads research

on how organizaons adapt to and implement

new technologies. He has over 20 years of

experience studying technological change and

organizaonal innovaon. His focus is on

understanding why some organizaons succeed

in integrang new technologies like AI and ML

while others struggle, parcularly examining the

intersecon of eciency and innovaon.

is a Director Data Science and leads the Data

Science Team at Zühlke. He has over 10 years of

experience in conducng data analycs and

machine learning projects. His focus is on

medical machine learning applicaons and

bringing prototypes to producon. He holds a

PhD and M.Sc. from the Instute for Machine

Learning at ETH, where he worked on scalable

methods for large-scale and robust learning,

wheel defect detecon and sleep stage predic-

on with deep learning.

This study was the result of a close collaboraon

between Zühlke and ETH Zurich. We’re grateful

to everyone who contributed their experse and

energy to the project.

Research leadership and supervision

▪ Philipp Morf, Head of Data & AI, Zühlke

▪ Sebasan Niederberger, Project Lead ETH

Zurich, PhD, Technology and Innovaon

Management Group

Design concept and research

▪ Daniel Schoess, ETH Zurich – semester project

and student assistant

Research and data analysis

▪ Richard von der Horst, ETH Zurich –

Master's thesis, MTEC

▪ Shaswat Gupta, ETH Zurich – student assistant,

MSc Computer Science

▪ Ege Karaismailoglu, ETH Zurich – student

assistant, MSc Computer Science

Stefano Brusoni

Gabriel Krummenacher

Acknowledgements

Imprint

Zühlke Engineering AG

Zürcherstrasse 39J

8952 Schlieren (Zürich)

Switzerland

info@zuehlke.com

Managing Director:

Nicolas Durville

Zühlke – Empowering Ideas.

Technology redenes possibilies. It can break boundaries, move

markets, and shape a more sustainable future. At Zühlke, we don't

just explore this transformave potenal; we turn it into tangible

business value for clients. And we’ve been doing it for more than

ve decades.

We're trusted to help clients rewire their businesses for the future.

Embedding modern technologies at the heart of their businesses

and products, we help our clients run smarter today while adapng

for tomorrow’s markets, customers, and communies. Our

muldisciplinary teams specialise in tech strategy and business

innovaon, digital soluons and applicaons, and device and

systems engineering. We excel in complex, regulated spaces – in-

cluding health and nance – connecng strategy, tech implementa-

on, and operaonal services to help clients become more

eecve, resilient businesses. Collaboraon, connected thinking,

and a rich partner ecosystem are key to how we deliver impact.

Founded in Switzerland in 1968, Zühlke is owned by its partners

and located across Europe and Asia. Our venture capital arm,

Zühlke Ventures, provides start-up nancing in HealthTech.

1 views·33 pages

Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF Free Download

Tailored, measured, and ethical: Uncovering the road to real-world AI impact PDF free Download. Think more deeply and widely.

Uploaded by sandra587 on 3/4/2026

/33

100%