2025 Research Report

A people-first skills pilot, exploring

AI adoption in the workplace

05
Contents
03 Introducing AI Works
04 Foreword
05 Key findings
06 Executive summary
Watch AI Works in action
10 Recommendations Bringing hands-on AI training
12 The opportunity and the challenge to workplaces across the UK
16 Purpose and methodology
Watch on YouTube
21 Findings and impact
32 Education Sector Pilot
44 Union Members Pilot
56 Small & Medium Businesses Pilot

Public First worked with Google on the AI Works initiative, ensuring high standards of
research integrity and collaboration throughout. As a company partner of the Market
Research Society and member of the British Polling Council, Public First adheres to rigorous
professional and ethical standards. Fieldwork was conducted between August 2024 and
March 2025, with ethical considerations prioritised at every stage. Informed consent was
obtained from all participants, and those from the qualitative research were assured that
their identities would remain confidential in all external reports and publications.
Introducing AI Works
An overview of the key findings,
research methodology, and
policy recommendations

Research conducted by
05
Foreword
AI has the potential to transform the way we work, tackling “I thought it was something really complicated, and something
the UK’s productivity puzzle and unlocking economic growth. you needed a coding language to use. But I learned that it’s
actually quite simple — we all are now using it.”
Fully realising AI’s economic benefits requires widespread adoption ­— Union member, three-months post-training
of the technology. We need to ensure that workers across the
country, in organisations of all types and sizes, have access to The UK could unlock £400 billion in economic growth through AI,
AI-powered tools and the skills required to use them. But we’re but half of that depends on adoption. From previous technological
already seeing an AI adoption gap emerging, with women and waves, we know uptake isn’t guaranteed.
older workers especially at risk of falling behind the curve.
Already, we see that women over 55 are four times less likely to
That’s why Google launched AI Works — partnering with use AI than men under 35, and smaller businesses lag behind
Multi-Academy Trusts (MATs), educators, small and medium-sized larger firms. That is bad for growth and for job satisfaction:
businesses (SMBs), and a union — to begin to understand AI usage AI can save time and reduce admin.
barriers and the most effective ways to supercharge AI adoption.
In these pilots we were surprised, and encouraged, that after a few
This report shares the simple, scalable steps — such as empowering hours’ training, adoption rocketed. After three months, participants
teams to use AI and building daily habits — that double AI use and were still using AI in a range of ways, and often daily. We think that is
turn AI sceptics into power users. because using AI tools is easy and because the reward — saved
time — is immediate.
And AI Works is just the beginning. Our ambition is for these pilots to be
a catalyst that encourages the UK’s policymakers and business leaders The training helped many of the people least likely to adopt AI
to act now. We can’t afford to leave AI adoption to chance, and we’re — older, more often women, and less confident users. There aren’t
excited to share insights that will enable us to be intentional about many levers that increase productivity and growth, and support
unlocking the untapped potential of this transformative technology. those most likely to be left behind. These pilots suggest training
is an investment worth making.
­— Debbie Weinstein, President of EMEA, Google
­— Rachel Wolf, Founding Partner, Public First

04
KEY FINDINGS KEY FINDINGS

Key AI habits are easy

to form. Just a few
findings hours of training
doubled daily AI
Permission to prompt
Workers needed reassurance
AI adoption increases
worker well-being
Scaling AI skills training
for national adoption usage, with usage
remaining high
that the advantage AI gives Pilot participants report that Across all three cohorts,
them is legitimate, fair, and AI upskilling has a significant we noticed that interactive
comparable to using the impact on worker well- sessions that encouraged
internet or a search engine being. Workers in each hands-on participation,
to complete a task more
efficiently. This can be best
achieved by giving workers
sector estimate AI is now
saving them time equivalent
to over 122 hours a year,
practical demonstrations
showcasing real and relevant
applications and tailored
even months after
explicit permission to use
the technology and setting
clear policies on AI tools.
exceeding modelled
estimates of 100 hours.
content designed for each
sector's specific needs made
the training particularly
the pilots because
AI-powered tools
Upskilling significantly effective. Sessions on AI
Habit formation leads narrows AI adoption gaps prompting and refinement
to experimentation Interventions are key to were especially well received.

are rewarding and

More than 80% of closing the AI adoption gap,
participants were surprised particularly among women The positive relationship
by AI’s capabilities. After three and older workers. Before between optimism and use

easy to use.
months, most were not only training, just 17% of women As AI usage increased
using AI for basic tasks, but over the age of 55+ in our among workers, so did
also independently seeking cohorts used AI weekly, and optimism about the
prompting tips, watching only 9% used it daily. Three technology’s impact. This
videos, and reading articles to months later, 56% were suggests that improved
uncover new features and using it weekly, and 29% workplace adoption is likely
innovative use cases. had made it a daily habit. to have a positive impact on
overall public sentiment
about AI and its evolving
role in society.

05
XX EXECUTIVE SUMMARY

Executive
summary

01.
The UK stands to gain £400 billion from
AI-driven growth, but only if workers from
all corners of the economy embrace the
technology. £200 billion of predicted gains
are dependent on the workforce adopting
AI and using it productively

At present, a third of UK • There are big demographic

workers report being regular divides. Women over 55 are four
users of generative AI at work. times less likely to use AI than
men under 35.
• Just over three in 10 workers
have already embraced AI. Increasing adoption among
Thirty-four percent of workers these groups is imperative,
use generative AI at work, with both to avoid the creation of a
three-quarters of those using it new “digital divide” and to take
multiple times a week and full advantage of the economic
nearly half using it daily. opportunity. We’ve learned that
most adoption is currently
• But the remaining 66% of worker-led rather than being
workers we spoke to did not directed by leadership. This
report using generative AI tools shows the fundamental appeal of
at work, and 57% reported not the technology, but leaders must
using a generative AI tool at all do more to move the UK from
in the past 12 months. organic exploration towards
widespread adoption.

06
EXECUTIVE SUMMARY

AI Works: accelerating AI sector impacted if not transformed.1

adoption across the UK Given AI’s extraordinary economic
potential, this long-tail pattern of
The current pace of AI adoption in adoption risks delaying
UK workplaces presents a clear productivity and compromising
challenge — the technology arrived long-term growth.
with significant fanfare and has since
demonstrated a wide range of use To learn more about what can be
cases. But while much has been written done to drive adoption, Google
about AI’s potential, little attention has worked with Public First and four
been paid to the question of how to partner organisations to pilot AI
increase AI adoption. training programmes in three key
sectors of the workforce: education,
We know from previous cycles of SMBs, and trade union members.
innovation that new technology is Together, these sectors represent a
typically embraced by a small substantial cross section of the UK
cohort of early adopters before workforce, selected to encompass
permeating the broader market both private and public entities,
more slowly. Almost two-thirds of skilled trades, and professional
jobs across the UK could be occupations across a number of
augmented by AI, with every demographic segments.

This ensured the research covered AI prior to the training, delivered

Almost two-thirds
a range of workplace environments, between 2.5 and 5 hours of bespoke
worker skill levels, and degrees of support, then measured attitudes
familiarity with AI, making it a strong again. To capture the impact on
test case for how AI adoption can be behaviour, we followed up after three

of jobs across encouraged throughout the broader

economy. The goal was to uncover the
most effective ways to accelerate AI
months to see if usage levels increased.

the UK could be
upskilling and adoption across the UK.

To do this, for each group we

augmented by AI
measured attitudes, confidence
levels, barriers, usage levels, 1. Public First, Public First supports new AI skills
appetites, and trust levels concerning pilot programme from Google, July 2024.

07
EXECUTIVE SUMMARY EXECUTIVE SUMMARY

Providing AI training Women’s daily AI usage Frontline workers understand the 17% to 56% weekly), though the
is effective challenges they face better than sample was small.
increased from 18% to

45
anyone. As they encounter these

%
Our research reveals that just a problems daily, they are also well- The impact was clear when we
small amount of training yields major placed to find practical solutions. asked respondents what had stuck
increases in AI adoption. Our research Through our AI training, workers not with them from the training three
found, at a basic level, just a few hours only learned how to use AI but also months on:
of training tripled the current usage of discovered ways to apply it to real
AI among trade union members and workplace challenges. The training “Exploring new tools and not being
doubled the current uptake with SMBs created the conditions for AI afraid of trying out new AI tools. Not
and teachers — unlocking substantial adoption to grow organically, as to be afraid and give up too easily.”
growth for the UK. people experimented with different — Female, Teacher, 61
use cases and shared their learning
Modelling by Public First shows that with colleagues. “I’m using Gemini within Gmail to
just a few hours of training delivers summarise complicated email
a 10:1 return on investment. By Demographic groups often threads and to help draft and polish
contrast, previous initiatives such as underrepresented in technology emails. I’ve used Gemini within
the Lifetime Skills Guarantee only adoption, specifically older Sheets to write a formula to split a
returned investment at a rate of 1:2.2 participants (those over 55) and spreadsheet into separate tabs so
women, saw their usage transformed. that each school is on its own tab.”
— Female, IT Manager, 61
• Women’s daily AI usage increased
from 18% to 45%, and weekly Post-training analysis revealed no
usage increased from 36% to 75%. significant difference in AI attitudes
or usage based on format, whether
• Those over the age of 55 in-person or through online sessions. This
increased their daily usage from suggests that well-structured, relevant
13% to 35%, and their weekly training content is more important
usage from 23% to 62%. than the mode of delivery. As a result,
AI training can be scaled efficiently
• T here was a near-tripling of daily without compromising impact,
usage for women aged over 55 making widespread adoption more
(from 9% to 29% daily, and from achievable across different sectors.

2. Department for Education, Measuring the Net Present Value of Further Education in England
2018/19, May 2021; HM Treasury, The Green Book: appraisal and evaluation in central
government, 2022; Office for National Statistics, Population estimates for the UK, England,
Wales, Scotland, and Northern Ireland: mid-2022, 2024.

08
XX EXECUTIVE SUMMARY

Shifting perceptions of AI

After training, all three cohorts using AI tripled. In addition to

reported feeling greater optimism boosting productivity, participants
about the impact technology will also reported meaningful well-being
have on society, increasing by 22 benefits: across each sector, workers
percentage pts among our education estimated that AI saved them the
cohort, 13pts among our SMB cohort, equivalent of 122 hours per year,
and 9pts among our union cohort. which exceeded our modelled
There was also a positive relationship estimates by 22%.
between optimism and use, with
increased daily use of AI tools All in all, we found that training
correlating with increased optimism increased participants’ confidence,
about the technology’s impact. understanding, and trust in the
technology, replacing negative
The training also fundamentally assumptions about how AI will impact
changed the way participants the future of work with a vision for
thought about AI. At the start of how these tools can enable more
training, most participants said that effective performance, freeing them
AI tools were primarily for writing to focus on higher-value tasks.
emails or summarising documents
and beyond this struggled to imagine
its relevance to their work. Post-
training, more than 80% said they
were surprised by AI’s capabilities and
after three months, a majority (70%)
of participants had started to
innovate, independently discovering
how AI tools could be used for new
use cases not taught in the training.
For example, one SMB participant
described how they used AI for
brainstorming by simulating feedback Post-training,
from different roles in their business.
more than 80%
This all added up to extremely high
usage among our cohort after
said they were
training. The number of education
workers using AI daily doubled and
surprised by AI’s
the number of trade union members capabilities

09
EXECUTIVE SUMMARY EXECUTIVE SUMMARY

What we recommend

Supporting adoption across the economy

As identified in the AI Opportunities Action Plan,

the Industrial Strategy and Skills England are key to
supporting AI adoption and skills development across
the economy.

• Businesses need support to equip workers with AI

tools, skills, and clear guidelines. As identified in the
AI Opportunities Action Plan, the Industrial Strategy will
need to set out how AI adoption can be best supported
in key industries. Our research shows that workers are
looking for explicit permission and guardrails on what
they can, and are encouraged to, use AI for. Businesses
and public sector leaders should develop practical and
positively framed policies that outline the AI tools
available, and where it is permitted and appropriate to
use them.

• Skills England should support a new accreditation

system for modular training that would recognises short,
effective training modules. As part of a new National
Skills Service, a model of micro credits would mean that
workers can take free or paid training courses that
have an immediate impact and a route into larger
qualifications. For the Government to deliver a lifelong
skills programme that is ready for AI, Skills England
should work to identify short modules of AI-related
training, make provision for these to be recognised,
and for funding to be available to workers to access
these courses via the new Growth and Skills Levy.

10
EXECUTIVE SUMMARY

Embedding AI in the public sector It should focus on the hardest to

reach frontline service delivery
The public sector can lead the way on teams and back-office personnel, as
AI adoption. opposed to those in digital and data
roles. Lessons learned will not only
• T he Government should guarantee provide a generalisable blueprint for
AI training for all public sector driving adoption across frontline
workers. Ensuring access for NHS, services, but also help to set the
local government, and civil service Government up for success in its
employees will help to drive plan to recruit 2,000 tech
adoption among this key segment apprentices and deploy specialist
of the UK workforce. Training should tech teams in departments.
be interactive and role specific, and
include practical demonstrations to •A
 n AI leader should be appointed
build confidence and ensure in every government department
relevance. Follow-up support and Executive Committee, with
clear workplace policies should mandated, regular AI training for
reinforce learning and encourage all members, ensuring strategic
long-term AI adoption. oversight and informed
decision-making on AI adoption at
• To support this goal, the Government the highest levels of government.
should launch a larger-scale public The Government should assess whether ensuring AI is deployed where it can
sector AI adoption trial, in line with its •A
 I training should be integrated the existing Digital Fast Stream scheme improve efficiency and replace repetitive
test-and-learn approach. The goal of across all Civil Service Fast Stream sufficiently covers AI, or if a standalone tasks. The survey should measure AI
the trial would be to identify barriers schemes, making AI a core AI scheme is needed. As AI teams are integration across sectors, workforce
and solutions to unlock adoption of development requirement for introduced in each department, all Fast readiness, and alignment with the
available but under-utilised AI tools. future leaders. Streamers should be required to Government’s TechTrack goal of
complete a rotation in an AI team, increasing digital roles in government.
embedding AI expertise across the
civil service. •A
 ll government departments should
As technological Tracking progress
conduct regular AI adoption audits,
coordinated by the new Centre for

advances continue to be As technological advances continue to be

Digital Government, assessing where
AI can reduce manual workload and

made, tracking progress made, tracking progress on AI adoption

will be key to keeping pace.
enhance productivity. Audits should
track progress towards the one in 10

on AI adoption will be •A
 s part of its remit to identify skills gaps,
civil servants in tech and digital roles
target, and follow the wider ambitions

key to keeping pace Skills England should track progress with

an annual AI Skills and Adoption Survey,
for AI in government as set out in the
AI Opportunities Action Plan.

11
 OPPORTUNITY AND CHALLENGE

Introduction:
The opportunity

02.
and the challenge
Why AI’s economic potential remains largely
untapped and what’s at stake for the UK

AI presents the UK with a huge the UK, where a gap between

economic opportunity. Previous innovation and implementation
modelling by Public First in 2023 has repeatedly undermined
for Google finds that it could economic potential. UK
contribute £400 billion in annual productivity is below the average
economic growth by 2030.3 of other G7 nations, with digital
Subsequent analysis suggests that adoption a particular challenge
up to half of this potential depends for small businesses that are
on the workforce adopting AI and significantly less likely to adopt
using it productively. technology than businesses in
other European countries.4 Small
Slow adoption is not a new businesses that do embrace
phenomenon. History shows this innovation see transformative
pattern recurring worldwide effects: on average those that
through successive waves of introduced innovation in the past
technology. But the challenge has three years saw their revenue
been particularly pronounced in increase by 14.8%.5

3. Public First, Google’s Impact in the UK 2023, April 2024.

4. Be The Business, The UK’s Technology Moment, 2020.
5. FSB, The Tech Tonic, Aug. 2023.

12
OPPORTUNITY AND CHALLENGE

AI represents a
once-in-a-generation
chance to grow the
economy for everyone

In its December 2024 call for

evidence on technology adoption,
the Government highlighted that
the UK does relatively poorly for
knowledge diffusion and knowledge
absorption.6 UK companies also invest
less in technology adoption than
their international peers.7 This
adoption lag results in suppressed
growth, stagnant productivity, and
ultimately lower wages and fewer
opportunities for people. AI represents
a once-in-a-generation chance to 6. Department for Science, Innovation and
grow the economy for everyone Technology, Technology Adoption Review,
— so concerted action is needed to Dec. 2024.
7. TechUK, Joint letter to the Chancellor of the
break the old cycle of slow adoption Exchequer: Scaling digital adoption to
and missed opportunities. strengthen the UK economy Oct. 2023.

13
OPPORTUNITY AND CHALLENGE

Percentage who have used generative AI tools at work

Our research identified demographic

disparities as a primary barrier to
80%
achieving widespread AI adoption.
Women report lower AI utilisation rates
than men across sectors, and that gap
widens with age. Without addressing
these imbalances, the full economic 70%
potential of AI technologies will remain
unrealised and workforce skills gaps
will get wider.
60%

50%

Women report 40%

lower AI utilisation 30%

Public First, landscape survey, Feb. 28, 2025–March 7, 2025.

rates than men 20%

across sectors,
and that gap
10%

widens with age Women

under 35
Men
under 35
Women
35-49
Men
35-49
Women
55+
Men
55+

14
 OPPORTUNITY AND CHALLENGE

34 %
Percentage using generative AI at work by income

80%

70%

60%
of people said they use
generative AI tools at work
50%

40% Socioeconomic factors also have a documents (59%). Our expectation

large impact on adoption rates. is that once the baseline rate of
High-income professionals adoption increases, engagement
demonstrate significantly higher AI levels will rise and this will prompt
30% utilisation. Further to this, there were innovation and an expanded range
clear disparities between sectors. of use cases.

Public First, landscape survey, Feb. 28, 2025–March 7, 2025.

20% Of course, adoption isn’t just about But how do we encourage worker
using a new technology once — it’s adoption? While there are plenty of
also frequency of usage. Our survey studies looking at AI’s potential
shows that once workers start using impact, surprisingly little has been
10% generative AI tools, they use them written about how to increase usage
often. Thirty-four percent of people outside the core of early adopters.
said they use generative AI tools at This report addresses that critical
work, with three-quarters of those knowledge gap with findings from
using them at least a few times a Google’s AI Works pilot programme.
£10,000--£14,999

£15,000--£19,999

£20,000--£24,999

£25,000--£29,999

£30,000--£34,999

£35,000--£39,999

£40,000--£44,999

£45,000--£49,999

£50,000--£59,999

£60,000--£69,999

£70,000--£79,999

£80,000--£89,999

£90,000--£99,999

£100,000 or more

week and half using them daily.

However, this activity is typically
restricted to a small number of use
cases: the most common being
assisting with communications or
writing (59%) and summarising longer

15
XX PURPOSE AND METHODOLOGY

AI Works:
Purpose and

03.
methodology
Identifying AI adoption barriers and
testing targeted training in three
key sectors of the UK economy

The AI Works pilot programme was designed to:

• Identify barriers to workplace AI adoption through
comprehensive research.
•D  evelop and test hypotheses about adoption
challenges in three sectors through tailored training.
•U  nderstand what works in AI training by identifying
factors behind successful adoption that can be
replicated or scaled for widespread impact.
• T rack longer-term behavioural changes via follow-up
impact surveys.

From this structured approach, we aimed to quantify

potential increases in AI usage and productivity, while
gathering insights into which training approaches and policy
interventions are most likely to fuel widespread adoption.

16
PURPOSE AND METHODOLOGY PURPOSE AND METHODOLOGY

How AI Works and challenges, and effective members (where unions have
was developed training approaches. Our diverse historically played a key role in
selection of interviewees also supporting workers through
Expert interviews provided insights into cross-industry technological change). For
Public First conducted 18 interviews opportunities and barriers. each, we developed targeted
with experts and representatives hypotheses:
spanning trade unions, employers, Hypotheses
and both public and private sector Based on our initial research, • For education workers:
experts. we identified three sectors with demonstrating how AI tools work
distinct adoption challenges: and behave is key to building trust
This approach enabled focused teachers (facing high workload), and giving users confidence —
discussions on key themes: AI’s SMBs (with known barriers to tech particularly in high-stakes
workplace impact, adoption benefits adoption), and trade union environments.

AI Works project timeline

April-July 24 July 24 Aug.-Sept. 24 Sept. 24 Dec. 24 Feb. 25 March 25

Training

Baseline Sector Landscape Pre-training Post-training Three-month Landscape

research hypotheses analysis analysis research impact analysis analysis follow-up
Expert interviews Initial research Landscape poll + Worker survey Worker survey Worker survey Second
+ literature review drove focus on expert interviews and focus groups + focus groups + focus groups landscape poll
three core sectors + literature review across three core across t hree core across three core
Baselining the state of Measuring how
AI usage in the UK Baselining the state of sectors areas areas attitudes to and usage
Education
Unions AI usage in the UK 
Assessing AI attitudes Understanding worker Measuring how of Al changed between
SMBs before training intentions about AI use intentions translated August 2024 and
immediately into usage March 2025
post-training

17
PURPOSE AND METHODOLOGY PURPOSE AND METHODOLOGY

• For SMB workers: creating new AI-related work habits

will help empower and upskill owners and employees.
• For trade union members: presenting role-specific AI
applications will accelerate upskilling and adoption
among unionised workers.

While focusing on these three sectors, the study was

designed to yield recommendations that would be
broadly applicable across the economy.

Based on our
initial research,
we identified
three sectors with
distinct adoption
challenges:
teachers, SMBs,
and trade union
members
18
PURPOSE AND METHODOLOGY PURPOSE AND METHODOLOGY

We designed
Landscape poll training cohorts demonstrated lower
A sample of 3,100 workers was confidence with AI and came from
conducted in August 2024 to demographics with historically
benchmark AI usage and attitudes, lower engagement rates compared

bespoke training
as well as current access to training. to the average for their sector.
A second landscape poll was
conducted in March 2025 to Training

programmes
investigate whether public attitudes Working closely with our partners,
beyond our selected cohorts had we designed bespoke training
changed in seven months. programmes addressing our
hypotheses and tailored to each

addressing our
Partnerships and recruitment cohort’s specific workplace context:
Google then partnered with three
sets of organisations to recruit • Education workers: 2.5-hour
participants and shape the training in-person workshops followed by

hypotheses and
programme. These partners helped up to six 45-minute optional online
tailor the training content to the sessions and 1:1 support over two
needs of their members, facilitated months. Content covered AI
engagement, provided ongoing fundamentals, limitations, prompt

tailored to each
support, and played a crucial role in customisation, and various AI tools.
embedding AI adoption within • Trade union members: Three
their respective sectors. The one-hour sessions delivered
organisations were: through a mix of online and optional

cohort’s specific
in-person formats, with 1:1 training
• In education, LEO Academy Trust available for participants. Training
and Lift Schools. covered AI fundamentals,

workplace context
• F or union members, Community responsible use, and effective
trade union. prompt writing. Participants learned
• For SMBs, Enterprise Nation. to use AI for tasks such as creating
articles, refining emails, and
Demographic data and baseline improving business documents.
engagement levels for participant • SMBs: Five one-hour webinar
groups are presented in subsequent sessions and four 30-minute
deep-dive chapters. But it’s worth lunchtime sessions, emphasising
noting at the outset that all of our peer learning through case
studies of AI integration in sales,
marketing, and operations.
Or a single full-day programme
(delivered in person).

19
PURPOSE AND METHODOLOGY PURPOSE AND METHODOLOGY

Evaluation
For each sector, Public First conducted:

• Before training, one to two focus groups and

a pre-training survey on attitudes and usage.
• Immediately post-training, one to two focus
groups and a post-training survey.
• Three months after training, two focus
groups and a final impact survey.

The research deliberately varied training

interventions across cohorts to address
sector-specific needs, granting rich insights
into real-world adoption patterns. Public First
ensured analytical rigour by conducting
statistical significance testing on adoption
changes and gathering qualitative data through
post-training surveys and interviews. While
not a formal academic study, this approach
offers valuable insights into effective
implementation strategies across diverse
workplace environments.

20
FINDINGS AND IMPACT FINDINGS AND IMPACT

Findings

04.
and impact
Brief training dramatically increases AI
usage, with lasting impact on adoption

In Public First’s initial landscape Most UK workers don’t currently

poll, 28% of workers reported see AI as relevant or useful
using generative AI at work, The landscape surveys suggest
increasing to 34% six months that perceived relevance is the
later. This pattern of modest biggest barrier to adoption,
usage and modest growth held rather than issues of capability
across our targeted sectors, with or complexity. Throughout the
approximately one-third of study, worker confidence about
education, SMB, and union AI remained high (above 80%).
workers engaging with However, a majority of
generative AI at work. respondents considered AI to be
Our training participants irrelevant to their roles. In other
reflected these low usage words, most UK workers already
patterns. Before training, only think they can use AI, they just
9% of our union cohort used don’t see the point. The
generative AI daily, compared to qualitative data suggests
19% in our education cohort and a two-fold challenge:
29% in our SMB cohort.
•M
 any workers can’t see
how to apply AI in their
day-to-day work.
• Others have tried AI tools
once or twice, but gave up
after failing to see a benefit.

21
FINDINGS AND IMPACT FINDINGS AND IMPACT

44 %
Together, these add up to a used generative AI tools did so at or encouraged by managers or
fundamental gap in understanding: least once a day. This suggests that organisations (28%). Six months later,
participants simply haven’t seen what while many workers are this pattern remained largely
effective AI use could look like in their experimenting with AI, they are not unchanged, with 70% continuing to
specific professional contexts. yet fully integrating it into their daily adopt AI independently, while the
However, our findings show that workflows. Our findings indicate that proportion of those encouraged by
targeted training helps bridge this a key barrier is the lack of clear their employers dropped slightly to
gap by demonstrating AI’s relevance guidance on how AI can be effectively 22%. This organic pattern of
to everyday tasks, increasing applied to their specific roles. engagement suggests that AI training of UK workers who said
confidence and ultimately driving — especially when provided or
greater adoption. Workers are taking the lead endorsed by employers — could
they used generative AI
without employer support dramatically boost usage by offering tools did so at least once
Frequency of usage is also lagging AI adoption is largely happening both the skills and official endorsement a day (pre-training)
Even among those who had used without official workplace guidance. that many people need before fully
generative AI tools at work, reported A significant majority of workers embracing a new way of working.
frequency of usage was low. Just surveyed (71%) chose to use AI tools
44% of UK workers who said they on their own rather than being asked “It might be because we’re quite
young and quite digital but there One of our expert interviewees,
are people asking for [AI] faster the chief executive of a national
than we can give it to them right membership organisation, observed:
now. I actually think that’s the “If you look at the distribution of the
case in lots of businesses.” use of technologies and therefore the
— Head of digital at a large levels of productivity within a sector,
employer it is vast. There’s no sign of that

Women over 55
shrinking over time as technology
AI adoption gaps risk trickles or diffuses.”
increasing workplace inequality
AI has the potential to radically In this light, targeted employer-led

are four times less change workplaces, but AI adoption

(and the productivity gains it brings)
is far from evenly distributed. Women
training represents not just a
productivity opportunity, but also an
essential intervention to prevent skills

likely to use AI than

over 55 are four times less likely to gaps within the workforce becoming
use AI than men under 35.8 And even further entrenched. And it
without targeted training and needn’t be expensive. The

men under 35
support, the gap between those who technologies showcased in our
can harness AI effectively and those training were mainly low cost or
left behind will only widen. free to use, and training platforms
such as Google Digital Garage are
free to access for both employers
8. Public First, National Worker Poll, March 2025. and employees.

22
FINDINGS AND IMPACT

Immediately post-training, estimates of previous

usage and anticipated future usage
1. The results were Feedback from participants
highly promising confirmed these findings, with a
Multiple times a day Once a day A few times a week
strong majority across all sectors
There were large increases in reporting they were more likely to A few times a month Once a month A few times a year
adoption after just a few hours continue using AI tools. When Once a year Less often Never
of training asked what drove their increased
engagement, participants
We found that a few hours of training consistently mentioned: better
100%
was sufficient to triple estimated daily understanding of what AI can do,
AI usage with unions, and double greater confidence in trying new
utilisation among SMBs and teachers. applications, and tangible
90%
time-saving benefits.
“I’ve been playing around with it
and trying to use it every day so I
80%
think it’s just becoming more natural
for me to use.”
— Education, post-training group 70%
participant

Public First, Aggregate across cohorts post-training surveys, Nov. 7, 2024–Jan. 15, 2025.
This increased engagement appears 60%
to stem from three key factors:

•A
 I tools are surprisingly easy to 50%
use — more like learning a new app
than mastering coding or similar
technical skills. 40%
• T he wide range of practical
applications helps users quickly
build regular habits. 30%
• Users are rewarded with immediate
time-saving benefits that reinforce
continued engagement. 20%

10%

Before taking the training, Anticipated future usage,

self-reported usage self-reported estimate

23
 FINDINGS AND IMPACT

You said you will use generative AI tools

more often going forward. Why is that?
“I think it’ll make me more time- out there to make your job easier
Select all that apply efficient. Because, although I won’t and your job enjoyable, and you
be able to use it necessarily for don’t feel defeated because they
Education (n=87) SMB (n=63) Union (n=100) everything I do, there’ll be some give you ideas. The AI gives you a
areas of my job that it will make narrative that you can work on and
much quicker and easier to do, start off with.”
80%
particularly if the council allows me — Education, three-month
to use it on my work machine.” post-training group participant
— Union member, post-training
72%

71%

69% group participant “What I loved the most about the

67%

training was the aspects of keeping

%

63%
63

60%

61%
61%
59%

Participants in our groups often the human in the loop and AI being

56%
reported enjoying using AI. What an augmentation tool rather than a

54%
54%

53%
began as curiosity quickly turned to replacement. So that was the biggest
appreciation as they discovered how shift; to go from this, almost like,

45%
44%

43%
43%
40% AI can make their work easier, more fear to this confidence of it won’t
%
41

do my work for me, but it will help

38%
creative, and even more enjoyable.
me do my work more effectively.”
“It’s given me a lot of hope. It’s given — SMB, three-month post-training
me a confidence boost. There’s tools group participant
%

24%
24
20%

Public First, All cohorts post-training surveys, Nov. 7, 2024–Jan. 15, 2025.
It will help save me time

I am more comfortable experimenting

with it now I know the basics

I have a greater understanding of where

Al tools can be helpful

I understand how to use it more effectively

It will help me be more creative

I understand how it works

I trust it more having done the training

“It’s given me a lot of hope.

It’s given me a confidence
boost. There’s tools out there
to make your job easier and
your job enjoyable.”
— Education, three-month post-training group participant

24
 FINDINGS AND IMPACT

Proportion predicting daily usage in the next three

months after training, compared to the proportion Most notable was the transformation we
reporting daily usage three months later saw among demographic groups often
underrepresented in technology adoption,
Pre-training daily usage specifically older participants (those over 55)
Predicted daily usage and women. We found that:
Reported daily usage
•W
 omen’s daily AI usage increased from 18%
to 45%, and weekly usage from 36% to 75%.

Public First, All cohorts post-training surveys, Dec. 20, 2024–Feb. 19, 2025.
• Those over the age of 55 increased their daily

and Public First, All cohorts impact surveys, Nov. 7, 2024–Jan. 15, 2025.
60%
usage from 13% to 35%, and their weekly
usage from 23% to 62%.10

There was also a near-tripling of daily usage for

40% women aged over 55 (from 9% to 29% daily,
and from 17% to 56% weekly), though the
sample was small.

We saw a consistent relationship between

20%
predicted and actual usage patterns in all our
cohorts. Immediately after training, 57% of
education and 59% of SMB participants
predicted they would use AI tools daily. Three-
month follow-up data closely matched these
Union Education SMB projections, with 47% of education participants
and 60% of SMB participants reporting daily
usage. Improved understanding of role-specific
applications of AI was commonly cited as both a “I think my confidence has
driver for initial adoption and for continued use. really grown. It has become
2. High predicted usage Adoption patterns proved resilient It is worth noting that the education impact a kind of personal assistant.
matched actual usage, after training. Both predicted usage survey took place immediately after the To be very candid, I use it for
with minimal drop-off (measured immediately after training) Christmas holidays, which may have contributed almost everything.”
and actual usage (assessed three to a slight drop-off in reported daily usage as — SMB, three-month
To understand whether training months later) maintained consistent participants returned to their regular routines. post-training group participant
created lasting change, we surveyed levels with very little drop-off. The
workers both immediately after their percentage of participants using AI
training and again three months later. tools at least a few times a week
This approach let us compare what increased dramatically — from 37%
people thought they would do with pre-training to 71% three months
9, 10. As the initial sample included many more SMB responses with higher usage rates, Public First weighted
what they actually did. post-training — with union members to the composition of the impact surveys on cohort. This minimises the risk of the change being a result of
in particular making significant gains.9 fewer SMBs featuring in the impact sample.

25
 FINDINGS AND IMPACT

How much time do you estimate you’ve saved in an

average week using Al tools? (Three months post-training)
3. Workers reported time became more comfortable
No time saved Less than 1 hour 1–3 hours savings of over 122 hours integrating AI into their workflows.
per year — exceeding These findings underscore the
4-6 hours 7-10 hours Over 10 hours
modelled estimates by 22% potential of AI training to drive
Don’t know long-term efficiency gains, even
Across all cohorts, participants for those who are still adapting to
reported significant reductions in the technology.
100% the time spent on routine and
administrative work, with estimated “I have noticed it is gradually
annual savings surpassing our initial reducing workload. I've been
projections.11 planning English [lessons] quite
a lot recently, so I have noticed
While some workers experienced it has helped a lot.”
immediate benefits, others — Education, three-month
75%
highlighted a gradual shift as they post-training group participant

50%

These findings
Public First, All cohorts impact surveys, Dec. 20, 2024–Feb. 19, 2025.
underscore the
potential of AI
25%

training to drive
long-term
Education (n=101)

SMBs (n=94)

Unions (n=103)

efficiency gains
11. Public First, Google’s Impact in the UK 2023, April 2024.

26
 FINDINGS AND IMPACT

Overall, how confident are you using generative Al tools?

4. The training increased
Very confident Somewhat confident Not very confident fluency and drove
Not confident at all Don’t know experimentation

Our training participants started with

lower confidence than the average
worker polled in the initial survey.

100%
But the training increased their
confidence and fluency. This
90% allowed participants to experiment
with the technology and discover
80% new use cases.

“I am not very good with computers,

70%
but I now feel confident using AI. It
used to frighten the life out of me.
60% So, yeah, if I can use it, anyone can.”

Public First, All cohorts pre-training surveys, Sept. 4, 2024–Jan. 15, 2025 and Public First,
— Union member, post-training

All cohorts impact surveys, three-months post-training, Dec. 20, 2024–Feb. 19, 2025.
50% group participant

After three months, communication AI to brainstorm new ideas and for

40%
and writing tasks remained the creative thinking, 41% to assist with
primary AI applications across our learning about a new topic, and 54%
30% cohorts (69%), but a majority of to assist with summarising or
participants also began to innovate simplifying longer documents.
20% with new use cases. For example, Seventy percent said they had
48% said they had used generative discovered a new use or application
for AI tools in the three months since
10%
the training.

“I think it’s made me more

All workers

Trade union workers

Education sector workers

SMB workers

Union cohort pre-training

Education cohort pre-training

SMB cohort pre-training

confident. Although I think the

confidence comes more from using
the tools rather than from the
training per se. What the training
did was inspire me to think about
using it in different ways.”
— SMB, three-month post-training
group participant

27
 FINDINGS AND IMPACT

Which of the following will you use generative Al to help

with in the future? 5. Cultivating responsible use of AI
Union Education SMBs
After training, participants developed a more
nuanced understanding of when and how to use
AI outputs. Many reported implementing their
own checks and balances, cross-referencing
multiple tools, considering the values of
different AI providers, and being mindful of
80%
data confidentiality. This shift reflects a deeper
level of critical engagement with AI.

60%

Advancing AI safely and

Public First, All cohorts impact surveys, three-months post-training, Dec. 20, 2024–Feb. 19, 2025.
40%
responsibly at Google
20%
Combining the best of Protecting your privacy We strictly uphold
AI and human insight with AI that is secure responsible data
by default practices, put users
We rigorously test our in control of their
models and infrastructure As we advance the information, and are
at every layer of the future of generative AI, actively implementing
Help you draft a document or report

Research or learn about a new topic

Help you write an email or text

Help brainstorm ideas

Help to be more productive

Help you analyse data

Reduce the manual load of admin tasks

Help you edit or generate a new image

Help you program or code

stack, combining the we leverage the same privacy safeguards

best of AI with our world industry-leading security tailored to the unique
class teams of safety infrastructure that needs of our AI products.
experts. This end-to-end protects billions of users
approach enables across all of our products.
advanced AI experiences
that put safety first.

To learn more visit ai.google/responsibility/safety/

28
FINDINGS AND IMPACT FINDINGS AND IMPACT

6. As usage increases, so The influence of AI training extends jumped by 22pts, from two-thirds Some participants noted that seeing
does optimism about the beyond just building technical skills: it (66%) to well over three-quarters AI’s practical applications firsthand
technology’s impact fundamentally reshapes participants’ (87%). Similar patterns emerged gave them a sense of agency over
perspectives on the role of AI in across other cohorts, with optimism the technology, and made them more
society. After completing the training, rising by 13pts among SMBs and 9pts confident about integrating it into
the share of education workers who among trade union members. This their daily routine.
felt optimistic that technology would finding is particularly significant for
have a positive impact on society policy development, suggesting that “I work in marketing, so I think it [AI]
improved workplace adoption will completely change my role, and
is likely to have a positive impact on that’s one of the reasons that I was
overall public sentiment about AI so keen to learn about it. I think you
and its evolving role in society. can stick your head in the sand and
you’ll just get overtaken by those
Perception of optimism towards
who are embracing the technology.”
technology pre- and post-training
— SMB, three-month post-training
group participant

100%

80%

and Public First, All cohorts post-training surveys, Nov. 7, 2024–Jan. 15, 2025.
Public First, All cohorts pre-training surveys, Sept. 4, 2024–Jan. 15, 2025.
60%

40%

20%
Education
pre-training

Education
post-training

SMB
pre-training

SMB
post-training

Union
pre-training

Union
post-training

National poll
pre-training

National poll
post-training

29
FINDINGS AND IMPACT FINDINGS AND IMPACT

TRAINING IMPACT
7. AI as an equaliser: enhancing “I have a few people [staff] who
accessibility for everyone are neurodivergent and have used
it in some incredible ways. There’s
Beyond improving efficiency, many participants someone who has ADHD … Since
found that AI played a crucial role in making the training she stepped up the AI
work and education more accessible and usage to help translate things in a
inclusive. Some teachers reported that AI tools way that she understands.”
helped them adapt lessons more effectively for — SMB, three-month post-
pupils with special educational needs. In training group participant
workplaces, AI was seen as a powerful support
system for neurodivergent individuals, assisting “I’ve got into HR because I have
with information processing and communication a passion for fairness, for equal
in ways that made tasks more manageable. This opportunity. English is not my first
newfound accessibility not only boosted language. I’m autistic. I have had
individual confidence but also reinforced AI’s so many people help me in my
potential to create more equitable learning and career, make little adjustments for
working environments. me here and there, so I could
deliver incredible work, and I feel
“Being a special needs teacher, we find it quite like AI can do what these
difficult to adapt lessons, because there’s not incredible people have done for
very much resource out there ... And me with a lot less effort. So it’s
sometimes trying to put that into a lesson plan that equaliser. You know, it helps
can be quite difficult, so being introduced to me be that equaliser.”
Gemini has cut my workload quite a lot.” — SMB impact group participant
— Education, three-month post-training
group participant

“I think you can stick your head

in the sand and you’ll just get
overtaken by those who are
embracing the technology.”
— SMB, three-month post-training group participant

30
FINDINGS AND IMPACT FINDINGS AND IMPACT

8. Scaling AI skills training “The area of prompt engineering

for national adoption was something that I found
extremely useful. Because, before
Feedback across all cohorts then, I’d been using it, but when I
highlighted a few key elements that got to see how I could further use
made the training particularly it to enhance my work, I mean, it
effective: just blew my mind.”
— SMB, three-month post-training
• Interactive sessions that group participant
encouraged hands-on
participation. Participants also praised the 1:1
• Practical demonstrations sessions, particularly those that
showcasing real applications allowed for deeper discussion on AI use
relevant to participants’ work. cases relevant to their work. This level
• T
 ailored content designed for of personal engagement was especially
each sector’s specific needs. valuable for those unsure about how AI
applied to their daily tasks.
Sessions on AI prompting and
refinement received especially “The session I had, it was wonderful,
positive feedback. Participants because he just demonstrated how I
emphasised how seeing real-world could personally use it. I was doing a
examples helped them better research project, so he just showed
understand AI’s practical value, me how I could use it to carry out
particularly for writing documents my project. Getting that insight was
and handling administrative tasks. something I never expected. It was a
Many appreciated the structured good learning experience for me.”
approach where trainers guided — SMB, three-month post-training mode of instruction. This is
them through different applications
and provided immediate feedback.
group participant Participants particularly encouraging as it suggests
that optimising for scalability should
Post-training analysis revealed no emphasised how be possible without sacrificing impact.

seeing real-world
significant difference in AI attitudes or
usage based on delivery format, It is also noteworthy that the training

examples helped
whether in-person or through had relatively low dropout rates.
webinars. The consistency of response We speculate that this could indicate

them better
across training methods suggests that that organisational buy-in played a
the content and structure of the significant role in sustaining

understand AI’s
training mattered more than the engagement, with participants being
more likely to complete the training

practical value
if they believed that it aligned with
their organisation’s priorities.

31
Education Sector Pilot
A deep dive on findings and
observations from our training pilot

In partnership with
05
 In
EDUCATION SECTOR PILOT EDUCATION SECTOR PILOT

“We’re always looking for new ways to improve the

experience of our staff and pupils and we see a number
of ways that AI can help us to do just that. AI can help us
to better cater to the various needs of the children in our
classrooms, capture the imaginations of pupils struggling
summary
to grasp new concepts, and even assist our support staff 1. AI could be particularly 3. Teachers frequently 6. Following the training,
important for teachers, expressed concern that teachers began using AI
with the day to day running of our schools, and this pilot as they manage current using AI constituted for a much broader range
is key to achieving these goals.” workloads and the “cheating” — suggesting of tasks than before, from
responsibility of helping explicit institutional document writing and
the next generation permission is critical for administration, lesson,
­— Phillip Hedger, CEO, LEO Academy Trust
understand this new widespread adoption and curriculum planning,
technology. Yet adoption within education. to student assessment
remains surprisingly low; and feedback.
“AI has tremendous potential to support our teams with our landscape polling 4. After bespoke training,
indicates that AI training adoption increased 7. Finally, one of the most
administrative tasks and transform the learning experience is a low priority for dramatically, delivering unexpected outcomes
for our pupils. It’s incredibly important for the nation’s education workers, with time savings exceeding of the training was how
educators to be equipped with the knowledge needed to only 22% choosing to our projections by almost quickly participants
teach and inspire the next generation about a technology focus on AI training if 10%. Success stemmed became advocates for
given a choice of from the tools’ intuitive AI, actively sharing
which is set to transform the way we live and work.” different training options. usability, integration into techniques and best
daily routines, and practices with colleagues.
­— Martin Simpson, Director of Technology, Lift Schools 2. Our initial hypothesis immediate benefits. This organic peer-to-peer
assumed trust was the learning suggests that
primary barrier to AI 5. How will adoption take even small scale training
adoption; instead, hold? Administrative tasks interventions can spark a
perceived relevance to will serve as the primary much larger cultural shift,
specific job tasks proved pathway for AI in schools, embedding AI adoption
more significant. addressing a pain point in far beyond the initial
teacher retention. For group of trainees.
“high-stakes” applications
— classroom instruction
and direct pupil
engagement — additional
leadership endorsement
appears necessary.

33
 EDUCATION SECTOR PILOT

AI could relieve The problem is not the core

teacher workload teaching role — 78% of teachers
say they enjoy their job “most or
AI has the potential to transform all of the time”. The problem is
education. Public First modelling administrative burden. 74% of
found that AI could enhance teachers report spending too
74% of roles in education and much time on administrative tasks,
potentially save teachers up to the and nearly half (49%) say the
equivalent of 109 hours a year, administrative burden has
which exceeded our previous increased in recent years.14 This
estimates of 100 hours a year level of workload is actively
by almost 10%.12 This time saving harming retention and
would be equivalent to a 16% recruitment. One-third (34%)
increase in the teacher to pupil of teachers and leaders are
ratio, or over 31,000 new teachers considering leaving the profession
(a 7% staff increase). within the next year, with the
overwhelming majority of these
We know that time matters in (90%) citing excessive workload as
education. Nearly half (49%) of the primary reason. And the inflow
teachers and school leaders believe of new teachers can’t keep pace
their workload is unmanageable — recruitment for 2023–24
and beyond their control, and reached just 50% of its target,
just 22% say their workload is continuing a long-term trend of
acceptable.13 In the initial landscape under-recruitment.15 Improving job
poll, 77% of those working in satisfaction could play a crucial role
education reported working in supporting the government’s

Education
outside their contracted hours. pledge to recruit 6,500 new
teachers during this Parliament,
helping to attract and retain talent
in the education sector.

Results from our AI training pilot

34
EDUCATION SECTOR PILOT

51
Would you say you use generative AI tools voluntarily

%
AI adoption is low, and AI or because you are required to by your employer?
training is a low priority All workers Education workers

“I’ll be honest, I was a bit dismissive.

That’s why when I saw the
opportunity, I was like, ‘How much 80%

75%
longer can I go on dismissing this?’.” reported no use of generative AI
tools at work in the past 12 months

71%
— Teacher, three-month
post-training education
group participant 70%

The initial landscape survey revealed

significantly low AI adoption within
60%
the education sector. While teachers
receive regular professional Perhaps unsurprisingly, generative
development — more than four-fifths AI remains largely unused within

Public First | Education workers | Landscape poll | Fieldwork Aug. 6, 2024–Aug. 14, 2024.
reported being offered training within educational settings. The majority
50%
the past year — less than one in four of education professionals in the
identified AI training as a priority. This landscape survey (51%) reported no
lack of prioritisation was mirrored in use of generative AI tools at work in
institutional provision: only 23% of the past 12 months. Where such tools
40%
respondents indicated that “AI tools were used, adoption was
[have been] required or provided as an predominantly teacher-initiated
option for use by employees”. For the rather than institutionally directed.
majority of teachers who had taken 30%
part in training, the focus of their From these baseline findings, we

23%
most recent training was compliance tailored our training to answer two
(36%) or soft skills (20%), and only fundamental questions:

18%
4% reported a focus on AI/machine • How can we effectively increase 20%
learning and 8% on advanced AI adoption among teachers and
technological skills (e.g. coding other education professionals?
skills, big data analysis). • Does AI adoption yield the

7%
time-saving benefits predicted 10%

5%
by our models?

1%

1%
I do this My employer My employer Don’t know
voluntarily encourages makes me
12. Public First, Google’s Impact in the UK 2023, April 2024. me to do this do this
13, 14, 15. Department for Education, Working lives of teachers and leaders: wave 3, Nov. 2024.

35
EDUCATION SECTOR PILOT EDUCATION SECTOR PILOT

How we ran the training

Google partnered with two two • Practical application and Public First measured usage and Our initial hypothesis
Multi-Academy Trust (MATs) — skill development — providing attitudes to AI using pre-training assumed trust was the
LEO Academy Trust and Lift Schools bespoke training on specific surveys and focus groups; major barrier to adoption
— to offer 300 education staff an AI tools, including those post-training surveys and focus
initial 2.5-hour in-person training on available through Google groups; and then a final survey We developed a hypothesis for
AI. This was followed by a series of for Education alongside and focus group three months teacher groups that was designed
voluntary online workshops over platform-agnostic solutions. after the training concluded. to be applicable across sectors:
three months. Overall we trained Teachers were equipped with Demonstrating how AI tools work
475 education staff through the frameworks for effective and behave is key to building trust
programme. prompting (from basic to How our in the product and giving users the
advanced applications); cohort compared confidence it will deliver effectively.
The training programme was methodologies for matching Our trainees skewed more female
structured around two core tools to specific tasks; and and were less confident with We chose this hypothesis because
components: mechanisms for experimentation, AI than the average education research on the adoption of new
including developing worker. Overall the group was less technologies often shows a
• Explaining AI — introducing personalised workbooks likely to have proactively engaged correlation between trust,
participants to AI as a concept; containing effective prompts. with AI than their sector peers. confidence, and product efficacy
how generative AI works in and adoption. Concerns about i) the
general terms, it’s inherent privacy of AI; ii) the reliability of AI;
strengths and limitations and iii) the quality of AI were the top
(including hallucinations, biases, cited reasons not to expand it in the
and privacy implications). workplace.16 In many of our expert
interviews, trust was also cited as a
core barrier for employees.

FIRST LANDSCAPE SURVEY PRE-SURVEY COHORT In fact, relevance mattered

more than trust
Female 61% 84% Testing our hypothesis against the
initial landscape poll revealed that
Male 39% 16% “trust was not the primary barrier to
adoption among education workers”.
Mean age 42 years 41 years

Confidence 34% of AI users 13% of users

using AI tools very confident very confident Survey responses:
(among users) 84% of AI users confident 85% of users confident Pre-training, 506; post-training, 471;
impact (three months on), 104 16. BCG survey, Jan. 2023.

36
 EDUCATION SECTOR PILOT

You mentioned you do not use

generative Al tools at work. Why is that? When asked why they didn’t use Training impact:
generative AI at work, the most boosting AI adoption
common answer was lack of and reducing workload
relevance to my job (38%), followed
38%

40% by lack of training (27%), and “I use AI to log and summarise

availability (24%). In contrast, only things. It just saves time. For
18% were concerned with data example, I was using AI today to
35%
privacy, 16% with trust, and just help with my workload, which meant
12% with ethics. I had an extra 20 minutes. I was able
to spend those 20 minutes with a
30%
27%

pupil to turn their diary around and

get them back into lessons.”
24%

25% — Education post-training

group participant
21%

18%

18%

20%

Public First | Education workers | Landscape poll | Fieldwork Aug. 6, 2024–Aug. 14, 2024.
16%

13%

15%
12%

9%

10%
7%

7%

7%
6%

5%
2%
Not relevant to my job tasks

Insufficient training on how to use Al

No available Al tools in my workplace

Prefer traditional methods

No company policy on using these tools

Data privacy concerns

Lack of trust in Al technology

Lack of confidence

Ethical concerns

Costs

Legal risks

Integration issues with current systems

Management does not support Al use

Other (please specify)

Don’t know

37
EDUCATION SECTOR PILOT

Change in Al use from before and after the training

AI usage increased dramatically AI was also seen as aiding human
Comparing usage pre-training with judgement rather than replacing it
Multiple times a day Once a day A few times a week
three months after training, the — a development viewed as positive
proportion of education workers in our post-training focus groups. A few times a month Once a month Less often
using AI weekly increased from 46% Never Unsure/don’t know
to 78%, while daily usage more than “It’s going to be transformational for
doubled, increasing from 19% to 47%. teaching because teachers spend
so much time doing administrative 100%
Three months post-training, tasks that actually draw them away
education workers reported saving from the teaching and learning
2.9 hours per week using AI. This and curriculum development. The
translates to 109 hours a year, things they actually like.”
greater than our initially modelled — Teacher, three-month post-

2.9
estimate of 100 hours. training education group participant

Public First | Education workers | 3 Months Post-Training Survey | 20th December - 5th February
75%

Public First | Education workers | Pre-Training Survey | Fieldwork 4th Sept - November 8th
Education workers reported saving

50%

25%

Pre-training Three months

hours per week using AI after training

38
 EDUCATION SECTOR PILOT

How would you rate the application of Al tools in your

work using the following scales? Teachers also felt AI had a tangible “We’ve had time to play around with
impact on their working hours, with the different apps and different
one noting, “AI has given me the applications … to do something during
100% chance to focus more on other areas the week and then come back to it.”

Irrelevant to my work

Pointless

Difficult to use
and projects, reducing my workload in — Vice Principal, post-training
the evenings”. This suggests that AI in group participant
education will deliver benefits that go
beyond productivity gains to enhance Why adoption increased: Building
80% teacher wellbeing and ultimately confidence and trust
increase staff retention. Education workers left the training
significantly more likely to describe AI
Why adoption increased: as relevant, useful, and easy to use.
Inculcating habits and encouraging
60%
experimentation The relevance and understanding also
appeared to directly impact trust, with
“I think when we went into the training teachers explaining that their
in September, it gave everybody that confidence in AI increased as initial
security blanket that this is OK for us, fears dissipated.
that it’s not cheating using AI. It’s just
40% finding the time to play.” “I don’t think I really knew what I was
— Female, SEN Teacher doing, and I didn’t really trust what I
was getting. So sometimes it was just
Relevant to my work

Our immediate post-training focus a case of learning how to better use it

Public First | Education workers | Pre- and post-training surveys | 2024

groups revealed that AI usage habits to help, and I have used it now in the
20% are quickly and easily formed and classroom in front of children, which I
Easy to use

these habits lead to more wouldn’t have done before.”​

experimentation, with participants — Three months post-training
Useful

finding new ways to use the education group participant

technology. In our post-training
survey, more than eight in 10 (86%) How teachers used AI after training
said that the training opened their Before training, education workers
Pre-training

Post-training

Pre-training

Post-training

Pre-training

Post-training

eyes to how AI could be applied to primarily saw generative AI as a writing

their work, and 90% said they were and communication tool. After training,
surprised by what AI could do. the use cases broadened significantly.
Writing and communications remained
the most common uses, but all other
categories of task saw large increases.
Respondents were asked to rate each statement on a sliding scale from 1 to 5,
where 1 represents one end of the spectrum (e.g. “Irrelevant to my work”)
and 5 represents the opposite (e.g. “Relevant to my work”).

39
EDUCATION SECTOR PILOT

77 %
And how have you used a generative Al tool in the
past [three months/12 months] at work?
Select all that apply.

Three months post-training Pre-training (prior 12 months)

used a generative AI tool to assist with 90%

communications or writing post-training

77%
80%

74%
70%
Below we have outlined some Creative and visual content design

60%
examples of how AI training has given “I use Canva to help advertise and
workers the confidence to explore create the posters for the school 60%

51%
and adapt the technology in ways musical.”

47%
they had not initially considered.
50%
“To create images for PowerPoints

41%
Public First | Education workers | 3 Months Post-Training Survey | 20th December - 5th February
Personal productivity for my children at school in terms
“I’ve discovered some Chrome app of lesson planning. Because

Public First | Education workers | Pre-Training Survey | Fieldwork 4th Sept - November 8th
40%
called Text Blaze. It’s an add-on on obviously you can use Google

27%
Chrome, and it basically summarises Images, but the images aren’t
30%

23%
it. You put in anything, any sort of precisely what you’re looking for. I
writing you want. You give it a found it really handy that you could

18%
command and it will copy it all out type into ChatGPT and just say,

15%
20%
for you on our document. So what ‘I need an image of something
I’ve been doing is, I’ve been getting completely ridiculous’, and it will
10%

5%
my templates for emails or whatever just give you this thing that you’ve

3%
I need, creating these templates in created, put on your slides, and
AI, moving it with text layers, giving that could be a monkey roller
its own command.” skating or whatever.”

To assist with communications

or writing

To assist with summarising or

simplifying longer documents

To assist with lesson

planning

To assist with
problem solving

To assist with analysing complex

information

Other (please specify)

Planning and organisation We also saw signs of peer-to-peer
“When I’ve been on Google Meet, it learning emerging organically after
comes up and it’s like, ‘Oh, do you the training, creating a ripple effect.
want me to make notes on the This is extremely important for
meeting?’ And that’s quite helpful, driving scale. Training doesn’t
because then I have a summary simply increase AI usage for direct
afterwards about what we’ve said recipients, it also creates organic
without having to write it down, momentum and offers a sustainable
which is quite handy.” route to widespread adoption.

40
EDUCATION SECTOR PILOT XX

“A lot of my family are also teachers. case for new technologies, where trust
Speaking to them, they were exactly typically develops over time. Even after
where I was before the training. So as training, 38% of teachers reported that
soon as I said to them, ‘I’ve done this’, worries about AI’s reliability would stop
and I showed them how I was them from using it more. Trust in AI did
planning, they were like, ‘I wish I’d improve post-training, but 20% of
known that. I’ve literally planned all of education workers still do not trust AI
this, and it could have saved me half for high-stakes tasks. Privacy concerns
the time.’ So showing them the little were the second-biggest issue, cited
tricks that I’ve learned, they’ve now by 31% of education workers.
tried it, and they’ve noticed the
difference as well.” Uncertainty about whether using
— Three-month post-training AI is allowed adds another layer of
education group participant hesitation. Teachers in focus groups
frequently expressed concern about
“In the geography network meeting, whether AI use was considered
we talked about different applications acceptable. One participant describes
that people were using and what they the mindset among teachers three
were using it for, and that was quite months after training: “It’s not cheating
useful to get a brainstorm of different using AI. It’s OK. I think there was a lot of
people’s ideas about how they were misconception from people that if they
using it as well.” used it, they were seen as cheating.”
— Three-month post-training
education group participant Another referenced problems with a
lack of guidance from government:
“I think it (education policy) is so
Remaining challenges: loose that people are hesitant.
Trust, reliability, and They’re a bit like, ‘We’re not really
institutional uncertainty sure if we’re meant to be doing
this, so let’s not do it just in case
Questions about reliability and it’s bad’ … That’s part of the issue.”
repeatability were the biggest concerns — Three-month post-training
among education workers post- education group participant
training, with doubts lingering
for some teachers about whether This suggests that clearer guidance is
the technology could be trusted in needed, reinforcing that AI is not just
“high-stakes” classroom situations, permitted but also beneficial. Without
such as presenting in front of a class. explicit institutional support, teachers
These concerns remain the largest will remain hesitant to integrate AI into
barriers to adoption, as is often the their day-to-day work.

41
EDUCATION SECTOR PILOT EDUCATION SECTOR PILOT

CASE STUDY CASE STUDY

Streamlining Simplifying
lesson planning and teaching tasks and
engaging pupils personalising
Cheryl Narayanan, a
dedicated special needs
classroom, using Gemini to
streamline lesson planning and
support
teacher at Pioneer School, create interactive activities,
joined Google’s AI Works such as generating custom Dave Sweet, a humanities clear, actionable guidance for
training with the hope of sentences to bring a literacy teacher and department a struggling pupil, transforming
easing her workload and lesson to life. AI has not only head with 15 years of experience what would have been hours of
better supporting her saved her time but also made at Ryde Academy, joined work into minutes. AI also
pupils, many of whom rely learning more accessible and Google’s AI Works training to streamlined his leadership
on different styles of enjoyable for her pupils Inspired explore how technology could duties, from creating bespoke
communication due to their by her own success, Narayanan ease his workload and enhance textbook content for out-of-
special educational needs. now encourages colleagues to pupil learning. date curriculum topics to
explore AI’s potential and plans generating differentiated
Before the training, her to lead a workshop during an While he had some prior reading materials tailored to
experience with AI was limited upcoming inset day. Describing experience with AI tools, Sweet individual pupil needs.
to occasional use for personal the training as a “confidence hoped the training would deepen
tasks, but she was eager to see booster”, she emphasised how his understanding. The training Sweet described the training
how technology could AI has transformed her equipped him with practical skills as a “game-changer”, boosting
transform her teaching. The approach to teaching, giving and new insights. Sweet found his confidence in using AI
training opened her eyes to AI’s her the tools to focus on what particular value in tools such as effectively and responsibly.
potential, equipping her with matters most — helping her Gemini, Canva, and NotebookLM, Looking ahead, he sees AI
practical tools to create more pupils grow. Looking ahead, she which revolutionised his approach becoming an integral part of
engaging, tailored lessons such believes AI will become an to lesson planning, exam education, not just for saving
as creating new images through indispensable part of education, preparation, and resource creation. time but also for creating
AI that will appeal to her pupils. while stressing the importance richer, more personalised
of ethical safeguards and One standout example was using learning experiences while
Since completing the training, continued training to ensure its NotebookLM to simplify four ensuring ethical boundaries
she has embraced AI in her responsible use. years of examiners’ reports into remain firmly in place.

42
EDUCATION SECTOR PILOT EDUCATION SECTOR PILOT

What we covered 1. Understand

Introduction to AI
2. Explore

Introduction to prompting

in the training
• Understand existing use cases for AI • Use a framework to write
• Define AI and machine learning better prompts
• Explain how generative AI works • Practice prompting for different
scenarios
Creating with AI • Evaluate outputs for accuracy
This training series consisted of up to • Explore different types of models and effectiveness
six 45-minute sessions broken up over • Interact with generative AI tools
Bring AI into your practice
the following three topic areas, with Responsible use of AI • Identify tasks which are a good fit
live demos and Q&As. Prompts, case • Describe strengths and limitations • Explore and compare different tools
studies, and examples were all tailored • Explore privacy and security
to education use cases. A condensed Advanced prompting
• Use AI as a source of personal
2.5-hour session was delivered in person feedback
to an inset day cohort, focusing on • Recognise the importance of iteration
sections 1 and 2.
3. Develop
Enhance your practice
• Reflect on practice to date
• Experiment with new and existing
tools

Plan for the future

• Utilise your AI Works playbook
• Stay up to date with future
development

43
Union Members Pilot
A deep dive on findings and
observations from our training pilot

In partnership with
05
 In
TRADE UNION SECTOR PILOT TRADE UNION SECTOR PILOT

“Community is committed to empowering our

members with the skills and knowledge to thrive in a
world of rapid technological change. Our partnership
with Google is a testament to that commitment.
summary
AI has the potential to positively transform the working 1. AI could help trade union 3. Our cohort of union 6. This supports our
lives of many of our members. By providing them with members contribute an members had lower distinction between
additional £89 billion of confidence in AI than relevance and
the tools and training to harness this technology, we’re economic value, but slow their peers, citing understanding. Most
not only ensuring they remain at the forefront of their adoption means this insufficient training on trade union members
industries, but also unlocking significant economic potential is not being how to use AI as their already recognised AI’s
fully realised.17 main reason for not relevance and potential
benefits for the UK as a whole. This is about ensuring
using the technology. to their role; what they
that the future of work is one where technology 2. Understanding and lacked was knowledge
empowers workers, not replaces them.” relevance proved to be 4. Weekly AI use more than of how to use it in
closely linked — the more tripled three months after their roles. Greater
people understand AI, training once the cohort understanding, not
­ Roy Rickhuss CBE,
—
the more relevant they gained baseline perceived relevance,
General Secretary, Community believe the technology knowledge. Perception drove the usage
to be to their work. In that AI was applicable to increase.
our landscape survey, of job roles only increased
union workers who say modestly, from 56% 7. While training increased
AI is self-explanatory, before training to 64% understanding of
79% think AI is relevant immediately post-training. AI, concern about
to them and only 14% permission remains a
irrelevant. In our union 5. A substantial increase in barrier. Only 46% of
member cohort, the understanding of AI is trade union members
starting level of responsible. When asked were confident in their
knowledge and why usage increased, understanding of their
understanding was trade union members employer’s AI policy,
lower, leading to a were over twice as likely and 58% said security
lower perception of AI’s to select “I understand concerns were likely
relevance to their work. more about how AI tools to stop them using
work” and “The training AI going forward.
helped me understand
how I could use it at work” 17. Public First, Google’s Impact
than any other option. in the UK 2023, April 2024.

45
XX TRADE UNION SECTOR PILOT

AI could massively During our research phase, both

increase the economic our expert interviewees and
value generated by some union members expressed
concern about AI’s impact on
trade union members, jobs — a recurring issue during
and workers are previous waves of industrial
optimistic about the and technological change.
technology’s impact
“I suppose the essay question is,
Trade unions have long supported what is the digital transition for
their members as they navigate blue-collar workers and how do
industrial change and adapt to you get people to a place where
new technology. And unions are they feel more comfortable; that
well-placed to continue this vital technology isn’t coming for their
role as AI rolls out across the UK jobs and is not going to be bad
workforce. Economic modelling for them? And, actually, they can

Trade
from Public First suggests that see how it helps them?”
AI could help union members — Trade union senior leader
contribute an additional £89 billion
of economic value to the UK — and
in the polling, union members were
enthusiastic about AI’s potential to
enhance their work.

union members
Results from our AI training pilot

46
 TRADE UNION SECTOR PILOT

37
How much of an impact, if any, do you think AI

%
is currently having on the way we work?

Trade union members Non-union members

40%

of trade union members think better

training programmes would encourage
them to use generative Al tools more
30%

However, contrary to this expectation, The optimism wasn’t unqualified, with

union members actually proved more union members more sensitive to
likely than non-members to say “AI some potential AI downsides than
20%

Public First | Union workers | Pre-Training Cohort Survey | Fieldwork 25th Sep - 15th Jan 2024
tools could make my job easier” other groups. In the poll, they were
(versus adding complexity) and “AI slightly more likely to say that AI was
tools will let workers focus on the already reducing the need for human

Public First | Union workers | Impact Survey | Fieldwork 27th Jan - 17th Feb 2025
more creative, strategic, or enjoyable workers (17% versus 11% for the
parts of the work” (versus increasing non-union members), and in our
demands on workers to do more qualitative research fear from media
10% with less). reports on AI came up regularly.

This may be the result of shifts in the “All I’d heard was things in the media,
composition of union membership. and I wanted to understand the
Unions traditionally drew upon positives and any negatives more
majority blue-collar jobs, but now than just hearing rumours. I don’t
around two-thirds (62%) of union want to be a dinosaur, I want to be
Major positive impact

Minor positive impact

No impact

Minor negative impact

Major negative impact

Don’t know

members in the UK have a degree able to use this to understand how

or equivalent higher-education it can be used in a positive way.”
qualification18 — a rate 12 pts higher — Union member, post-training
than for non-union workers. focus group participant

18. Department for Business and Trade, Trade Union Membership, UK 1995-2023:
Statistical Bulletin, May 2024.

47
 TRADE UNION SECTOR PILOT

What would encourage you to use generative Al tools

Understanding and relevance proved
more frequently than you currently are? Driving understanding to be closely linked — the more people
was key to accelerating understand AI, the more relevant they
Workers Unionised workers
adoption with this cohort believe the technology to be to their
We developed a hypothesis for trade work. In our landscape survey, of
union members based on expert union workers who say AI is self
interviews with key stakeholders: explanatory, 79% think AI is relevant
37%

40% showing unionised workers generative to them and only 14% irrelevant.
AI applications that are relevant to In our union member cohort, the
their roles will accelerate AI upskilling starting level of knowledge and
34%
33%

and adoption. But in practice, driving understanding was lower, leading to

32%

31%
30%
understanding was the main factor a lower perception of AI’s relevance
28%

27%
27%

26% that unlocked adoption for our cohort. to their work.

30%

26%

25%
24%
22%

22%
20%
“At first, I thought when AI
19%
20%

takes over, it’s going to be

14%
10%
like robots taking over the
world. But it really changed
my perspective. AI can
actually enhance my work.
Better training
programmes

Demonstrated benefits
to job performance

Clearer guidelines
and regulations

Assurance of data
privacy and security

Simplified Al tools

Continuous
technical support

Improved user
interface of Al tools

Increased
management support

N/A — nothing would encourage

me to use Al tools more often

The way we are growing and

Landscape Survey - Union Workers

shifting in a lot of workplaces,

it’s a quite useful tool.”
––– Trade union member, three-month post-training group participant

48
TRADE UNION SECTOR PILOT TRADE UNION SECTOR PILOT

How we ran the training This suggests that lack of knowledge

about how to use AI, rather than a
In our training, we wanted to The training included: Public First measured usage and failure to see its potential in their
increase understanding of how • Introduction to AI — providing attitudes to AI in pre-training roles, was the obstacle.
union members can integrate AI union members with a basic surveys and focus groups;
into their daily workflow. Google understanding of key principles post-training surveys and focus Our trade union cohort reported
partnered with Community union and the strengths/weaknesses groups; and then a final survey and different adoption barriers compared
to offer 404 union members three of AI tools. focus group three months after to the broader union population from
one-hour live webinars, as well as the training completed. the landscape poll; relevancy was not
three optional in-person training • Tangible use cases — a key barrier to adoption for our union
sessions specifically curated to demonstrating specific tasks cohort unlike in our landscape survey.
assist union representatives in and tools relevant to trade union How our Our cohort, started with significantly
their role supporting members, members’ work contexts. cohort compared lower engagement levels, with only
as well as 30-minute bookable 1:1 Our union cohort was older 60% of those who had used AI
mentoring sessions on AI available • Prompt engineering and less confident using AI tools expressing confidence using AI
to the whole cohort. guidance — teaching than the average union member. versus 86% in the general trade union
participants how to write While they volunteered for the member sample; and low adoption
effective AI prompts. programme, they represented a within our cohort stemmed primarily
lower-engagement demographic from confidence issues and
within the union membership. knowledge gaps. When asked what
prevented them from using AI, almost
half of our union participants cited
insufficient training.

Unsurprisingly, union members

identified training as the key factor
FIRST LANDSCAPE SURVEY PRE-SURVEY COHORT that would encourage increased
AI adoption — placing even higher
Female 50% 52% importance on this than the average
worker. Interestingly, they showed no
Male 50% 47% strong preference for who delivered
this training, whether peers, external
Mean age 42 years 53 years consultants, managers, or union
representatives. Our data also
Confidence 38% of AI users 11% of users
revealed that union members placed
using AI tools very confident. very confident.
greater emphasis on concrete
86% of AI users confident. 60% of users Survey responses:
pre-training, 145; post-training, 119; job-performance benefits compared
confident.
impact (three months on), 122 to the average worker.

49
 75%

25%
50%
100%
Never

Pre-training
Multiple times a day
A few times a month
Once a day
Once a month
Change in AI use pre- and post-training

Unsure/don’t know

after training
Three months
Less often
A few times a week

Public First | Unions workers- Pre and Post-Training Survey. Public First | Union workers | Pre-Training Cohort Survey | Fieldwork 25th Sep - 15th Jan 2024
Public First | Union workers | Impact Survey | Fieldwork 27th Jan - 17th Feb 2025
TRADE UNION SECTOR PILOT

50
TRADE UNION SECTOR PILOT

12.5
And how have you used a generative Al tool in the past
Union members also [three months/12 months] at work?
used AI for a broader Select all that apply.
range of tasks after
the training Impact Pre-training

Summarising information
working days annually “It is amazing at summarising 80%

74%
information. Other government
was the reported

69%
organisations are constantly
equivalent time saved pumping out documents for us 70%
through AI application to try and read and make sense of,
and AI just bangs them out. They’re

55%
publicly available so you don’t even 60%
need to copy and paste them in.

49%
You can just ask for a summary of
Training impact: big the relevant changes and things.”
50%

increases in adoption

40%
37%
37%
and time savings Concept development and 40%

34%
Public First | Union workers | Pre-Training Cohort Survey | Fieldwork 25th Sep - 15th Jan 2024
brainstorming

29%
29%
A three-month follow-up survey “Quite recently, I asked one of the

28%
revealed significant growth in usage [AI] platforms to give me [some 30%

Public First | Union workers | Impact Survey | Fieldwork 27th Jan - 17th Feb 2025
among our union cohort. Daily AI use ideas]. I thought I’d really test
increased from 9% before training to the limits. I said, ‘[Give me] 50

15%
29% afterward, while weekly usage campaigning ideas around health 20%

12%
tripled from 17% to 61%. Importantly, inequality that would be of interest
union members reported saving to the Government’.”

8%
an average of 2.1 hours per week 10%

through AI application — equivalent Drafting and editing

to around 12.5 working days annually. “So I put it into AI. I said, ‘Create
an evaluation and feedback form’.

To assist with communications

or writing

To assist with summarising or

simplifying
longer documents

To assist with learning

about a new topic

To assist with brainstorming

new ideas and creative
thinking

To assist with analysing complex

information

To assist with problem

solving

Other
(please specify)
And, you know, whoosh, I had 20
questions, which were in different
sections. So I think we’ve tweaked it
very slightly, but I would say probably
98% of what was created using AI
was brilliant.”

51
 TRADE UNION SECTOR PILOT

65 %
You said you have used generative Al tools in the
past three months at work. Which of the following, Addressing key
if any, best explains why? misconceptions about
Select up to three AI accessibility and
impact on jobs

The training effectively dispelled

66%

said the training helped them

65%

70%
a common perception that AI was
understand how they could too technically demanding to use.
60% use AI at their work In our focus groups, participants —
particularly older workers — frequently
expressed surprise at how accessible
the technology proved to be:
50%

Increased usage was “I was really surprised that it’s

40%
driven by increased not difficult to use, because I am
understanding of AI rubbish on computers. It’s just not
part of my job. I never did it in school
29%

29%

The training transformed both or anything.”

30%
understanding and intended usage — Union member,
of AI among our union cohort. In our post-training group participant
post-training survey, 58% of union
16%

20%
participants reported they were likely Workers emphasised that the training

Public First | Union workers | Impact Survey Fieldwork 27th Jan - 17th Feb 2025
to increase their AI usage due to gave them a solid foundation and the
newly gained understanding of helpful confidence to start using AI, making it
10%
applications. The same percentage feel more approachable and less
4%

3%

3%

acknowledged that the training had intimidating:

introduced them to AI use cases they
wouldn’t have previously considered. “It just gave us a really good
I understand more about
how Al tools work

The training helped me understand

how I could use it at work

I got into the habit after

taking the training

I trust the tools more

than I did before

Al tools have been

introduced in my workplace

I was encouraged to use it

by my manager

Don’t know

Other (please specify)

grounding and made me feel that,

“I thought it was something really yeah, I could give it a go. It wasn’t
complicated, and something you anything to be worried about. And
needed a coding language to use. I have used it and I hadn’t before,
But I learned that it’s actually quite because I really didn’t know
simple, and we all are now using it.” anything about it.”
— Union member, — Union member, three-month
post-training group participant post-training group participant

52
TRADE UNION SECTOR PILOT

Privacy, workplace
policies, and other
organisational barriers
Despite strong improvements in AI
confidence and adoption, there are
still some major impediments to
adoption. The biggest uncertainty
for union members is whether their
organisation supports AI use. Only
46% of union members feel confident
that they understand how their
employers would like them to use AI
tools. After training, concerns shifted
notably — from lack of training and
questions about relevance to more
practical issues around privacy
protocols and organisational policies.

As in other cohorts, the lack of clear

organisational policies also manifest
as concern that using AI could be
perceived as cheating:

“I’d almost feel like I was cheating

in some way to use it. There’s a
negative commentary around it
that’s something to be fearful of.”
— Union member, post-training
group participant

53
TRADE UNION SECTOR PILOT TRADE UNION SECTOR PILOT

CASE STUDY CASE STUDY

Cutting admin Enhancing

time to focus learning and
on care knowledge
Tania da Silva, a team manager in a supported living
home, joined Google’s AI Works training to explore
sharing
how AI could enhance her work.
Carl Ravenhill, a steelworker at Tata Steel, joined
Initially unfamiliar with AI’s potential, da Silva was Google’s AI Works training to stay ahead of technological
surprised by how tools such as Canva, which changes he saw shaping the future of work.
she had associated with creative jobs, could streamline
her work. She now uses it to design invitations for While he had dabbled with AI for personal use, he lacked
residents’ birthday parties and postcards for holidays, the skills to use AI effectively. The training introduced him
while Gemini assists with report writing, staff training to a range of tools, revealing AI’s versatility beyond simple
materials, and meeting preparation. queries.

AI has significantly reduced her administrative workload, Ravenhill was particularly impressed by how AI could
allowing her to focus more on supporting residents with simplify complex topics, breaking down technical jargon
learning disabilities. Da Silva has embraced AI as a into layman’s terms, and by creative applications like
time-saving, accessible tool, while remaining mindful of designing visuals through Canva.
the need to verify outputs.
He now uses AI to streamline training materials and
She actively encourages colleagues to explore AI, organise knowledge into concise, accessible formats.
demonstrating how it refines writing, simplifies
communication, and enhances creativity. Looking ahead, Ravenhill is an enthusiastic advocate, encouraging
da Silva sees AI becoming an integral part of her role, not colleagues to explore AI and sharing tips from his
as a replacement but as a powerful support tool, making training. He sees AI as an essential tool for the future
her work more efficient while freeing up time to focus on — not just enhancing individual productivity
the human connection essential in care settings. but also empowering entire workforces.

54
TRADE UNION SECTOR PILOT TRADE UNION SECTOR PILOT

What we covered 1. An introduction to AI

• Key concepts of AI: what is AI,
2. Getting hands-on
with AI tools

in the training
breaking down jargon and • Using AI tools to support writing:
understanding it in practice web applications, voice typing,
• How AI works: types of generative AI transcribing handwriting
models and how it works • Using AI tools for designing:
• Using AI responsibly: limitations and creating posters & images
This training series consisted of three watch-outs, responsible use • Using AI tools for understanding
one-hour sessions with live demos and and condensing large volumes
of information
Q&As. Prompts, case studies, and examples • Using AI tools to learn new skills
were all tailored to sectors represented • Key considerations when selecting
within Community union. A condensed new AI tools
in-person version was delivered over
2.5 hours for union representatives, 3. Writing effective prompts
with examples tailored to tasks and
responsibilities of union representatives. • An overview of prompt engineering
• Techniques for crafting
effective prompts
• Effectively reviewing prompt outputs

55
Small & Medium
Businesses Pilot
A deep dive on findings and
observations from our training pilot

In partnership with
05
 In
SMB SECTOR PILOT SMB SECTOR PILOT

“Small businesses already recognise AI is key

to increasing the efficiency and the speed with
which they can efficiently fly through a long list
of complicated and time-consuming tasks. There
summary
is an emerging confidence and skills gap that 1. Reported confidence in 3. Training is the biggest 5. There was a clear
must be addressed if these businesses are going AI is high among small barrier to adoption. shift from ad hoc
and medium business Sixty-two per cent of experimentation before
to truly benefit from the transformative power (SMB) workers, but those who participated training to purposeful
of this technology.” adoption is low and in our training said that integration of AI in their
usage is infrequent. they didn’t use AI more workflow three months
Workers need to be at work because of after training. In our
­ Emma Jones CBE, Founder,
—
shown clear use cases insufficient training, focus groups, workers
Enterprise Nation through training to see more than the second often referenced the
the full benefits of AI. and third most-cited positive effects of the 1:1
problems combined. training sessions and the
2. But small businesses practical, demo-based
lack the capacity to do 4. Following tailored training Lunch & Learn sessions.
training themselves. on clear applications of AI
SMB workers receive for SMBs, adoption was
disproportionately transformed. The number
less training than of SMB workers making
their counterparts at daily use of AI doubled
larger businesses. (from 29% to 60%). In our
focus groups, workers
repeatedly said that being
trained on relevant use
cases of AI was key to
increasing their usage.

114
57
INTRODUCTION SMB SECTOR PILOT

Reported confidence in AI among SMBs is high,

but adoption remains low
SMB workers expressed strong SMBs in digital adoption,
confidence in their digital skills scoring an average of 7.6 out
in the initial landscape poll: 92% of 10 in technology utilisation,
said they felt confident using compared to 7.3 for medium-
technology and 84% who use sized businesses and just 6.6
AI specifically reported feeling for small businesses.20 This may
confident in doing so. However, be because larger businesses
high reported confidence did provide more training: 69% of
not equate to adoption. Despite those employed by SMBs in our
71% of SMB workers seeing AI as landscape survey had received
at least a little applicable to their some formal training, compared
work, only 38% of SMB workers to 81% of those in larger
reported actually using AI in the organisations. Asked how
past 12 months. they had learned the most
recent skill they had acquired
This pattern reflects wider at work, nearly a third of SMB
trends in digital adoption. UK employees (29%) had taught
SMBs invest less in technology themselves, compared to 19%
and have lower digital adoption of those at larger organisations.
than their G7 counterparts, but This was even higher among
believe they are at the same employees at the smallest
level as their international peers.19 organisations (under 10
Large organisations in the UK employees), with 42% of those
also consistently outperform having taught themselves.

SMB Results from our

AI training pilot
58
SMB SECTOR PILOT

You mentioned you do not use generative Al tools

at work. Why is that?
Workers agree that lack of From these discussions, we also Select all that apply (N = 89)
training is the main barrier found that many SMB workers were
to adoption aware of AI and occasionally used it,
but lacked understanding of how to 80%
More SMB workers in our training fully leverage its capabilities.
cohort (62%) cited insufficient
training as a barrier to greater usage Without targeted training
of AI than the second and third interventions, the current pattern (high
barriers combined. In our focus confidence but low usage) will persist, 60%

62
groups, workers repeatedly cited limiting the potential economic gains
the need to be shown practical use AI could bring to the SMB sector and
cases of AI. the wider UK economy.

%
40%

Public First | SMB workers | Cohort Survey | Fieldwork 19th Sep - 21st Nov 2024
20%

Insufficient training on how to use Al

Lack of confidence

No available Al tools in my workplace

Data privacy concerns

Lack of trust in Al technology

Ethical concerns

No company policy on using these tools

Prefer traditional methods

Integration issues with current systems

Costs

Legal risks

Not relevant to my job tasks

Management does not support Al use

cited insufficient training as a
barrier to greater usage of AI
19. Be the Business, G7 Productive Business Index, Aug. 2023.
20. The economic and social benefits of digitalisation, A Cebr report for Virgin Media O2 Business, Oct 2024

59
SMB SECTOR PILOT SMB SECTOR PILOT

How we ran the training

In our training, we designed a The training included: Our training incorporated key How our cohort compared
programme to transform SMB adoption • Core webinars introducing habit-formation principles identified Our SMB cohort began with significantly
of AI, encouraging workers to develop foundational AI concepts through through our literature review. These lower AI confidence than the average
regular usage habits and integrate weekly sessions, ensuring regular behavioural concepts significantly SMB worker. Despite volunteering for
these tools into their daily workflows. exposure and skill reinforcement. influence how people adopt and the programme, they showed less initial
Google partnered with Enterprise • Lunch & Learn sessions featuring integrate new skills into their routines. engagement with AI tools than their
Nation to offer 905 SMB workers a panel discussions where relatable Public First measured usage and industry peers.
series of core webinars, practical users demonstrated practical AI attitudes to AI in pre-training surveys
demo-based Lunch & Learn events applications. and focus groups; post-training surveys
(for Marketing, Sales, Operations, • In-person workshops showing how and focus groups; and then a final
and Personal Productivity), and live AI tools could be integrated into daily survey and focus group three months
in-person training day events. tasks and routines. after the training.

PRINCIPLE FROM HABIT RESEARCH HOW WE ADAPTED THE TRAINING FIRST LANDSCAPE SURVEY PRE-SURVEY COHORT

Ebbinghaus forgetting curve: The training programme was Female 48% 57%
Knowledge retention declines spread across weekly webinars and
rapidly after initial learning, with the Lunch & Learns for regular exposure Male 52% 41%
sharpest drop occurring within the and repetition.
first few days. However, spaced Mean age 42 years 47 years
learning or distributive practice can
reduce the rate of forgetting. Confidence 40% of users very confident 11% of users very confident
using AI tools 93% of users confident 77% of users confident
Social support: People are strongly Bringing together groups of SMBs among users
influenced by the behaviour of those for panel discussions where peers
around them. Finding the right social demonstrated real-world AI
support helps habits to form. applications helped participants see
how similar businesses were using
these tools.

Situational cues: These help We designed practical modules

form habits by associating specific showing exactly when to use AI during
contexts or triggers with particular common daily tasks, linking specific
behaviours, reinforcing the work situations with appropriate
habit loop. AI applications. Survey responses:
pre-training, 594; post-training, 71; impact (three months on), 95

60
SMB SECTOR PILOT

You said you have used generative Al tools in

the past three months at work. Which of the
Our initial hypothesis emphasised the
need to drive adoption by encouraging following, if any, best explains why?
new habits Select up to three. (N=40)

We developed a hypothesis for SMB workers that was

designed to reflect the systemic challenges SMBs face
80%
around digital adoption: Encouraging workers to adopt
new AI-related habits will help to upskill and empower
workers to use AI.

60%

Public First | SMB workers | 3 Months Post-Training Survey | Fieldwork 21st Jan - 19th Feb 2025
40%

20%

I understand more about

how Al tools work

The training helped me understand

how I could use it at work

I trust the tools more

than I did before

I got into the habit

after taking the training

Al tools have been

introduced in my workplace
61
SMB SECTOR PILOT SMB SECTOR PILOT

Change in Al use Multiple times a day

A few times a week
Once a day
A few times a month
pre- and post-training Once a month Less often
Never Unsure/don’t know

25% 50% 75% 100%

Public First | SMB workers | 3 Months Post-Training Survey | Fieldwork 21st Jan - 19th Feb 2025
Public First | SMB workers | Pre-Training Survey | Fieldwork 19th Sep - 21st Nov 2024
Pre-training

Three months
after training

62
INTRODUCTION SMB SECTOR PILOT

AI training did lead to big “I think one of the most eye-opening

daily usage increases as new examples for me was when
habits formed somebody was talking about how
The proportion of SMB workers using they simulated different roles in
AI daily increased from 29% to 60% their businesses so that they could
after training, and the number of brainstorm. The idea of having a
workers who used AI at least weekly virtual team when you work on your
increased to 86%. own was quite mind-blowing.”
— Three-month post-training
In the focus groups conducted three SMB group participant
months after the training finished,
SMB workers consistently commented
positively about the training’s clear
relevance to their work and the
real-world use cases we were able
to share.

86
After training, the number of workers

%
who used AI at least weekly increased to

63
 SMB SECTOR PILOT

How have you used a generative Al tool

in the past three months at work? AI use cases evolved in the “I just wanted a more structured
Select all that apply. months after training way to gain a better understanding
The period after training saw a of its potential and how I could
Pre-training Three-months post-training
significant broadening of AI use cases actually incorporate it, or help my
among SMB workers. But the most clients with it.”
90% common use of AI — assisting with — Three-month post-training
82%

communications and writing — actually SMB group participant

79%

fell slightly. This likely reflects a shift

80%
away from ad hoc experimentation
towards purposeful integration of AI
67%

70% into their daily workflow. In our impact

focus groups, workers highlighted that
60%

59%

while in the past they might have

54%

60%

15
experimented with apps, training
50%

50%

Public First | SMB workers | 3 Months Post-Training Survey | Fieldwork 21st Jan - 19th Feb 2025
allowed them to intentionally use AI
46%

%
50% for a wider range of tasks:

Public First | SMB workers | Pre-Training Survey | Fieldwork 19th Sep - 21st Nov 2024
36%
35%

40%
31%

30%

20%
7%
6%

10%
To assist with
communications or writing

To assist with brainstorming

new ideas and creative thinking

To assist with summarising or

simplifying longer documents

To assist with learning

about a new topic

To assist with
problem-solving

To assist with analysing

complex information

Other

increase, post-training in AI use to assist

with problem-solving

64
SMB SECTOR PILOT

After the training, workers

progressed from using AI for
occasional tasks to integrating it into
their daily workflows — streamlining
content creation and improving
decision-making. This shift occurred
because training helped them
experience the immediate benefits
of AI. Previously, many SMBs
reported abandoning AI tools when
they didn’t see quick results, but our
training helped participants discover
tangible rewards that reinforced
continued usage and helped new
habits to form:

“I’ve been able to create

spreadsheets with it, things that
I wouldn’t have thought. I initially
thought that spreadsheets were
not my thing — doing all these
formulas and stuff, it’s too tedious.
But since learning how to use
Gemini the right way — oh, my god,
Gemini is my best friend. I’m loving
it. It’s been an absolute lifesaver.”
— Three-month post-training
SMB group participant

65
SMB SECTOR PILOT SMB SECTOR
SMB: RESULTS
PILOT

After training, many reported using Creative content generation

AI to streamline administrative “I found the marketing component
tasks they previously found time really helpful. So I never really used
consuming. Others reported using AI for marketing or for anything.
AI as a brainstorming tool to develop We use ChatGPT now and I literally
ideas and refine business strategies. can be like ‘Can you write me a
marketing campaign for 16-to-24-
Increasing productivity year-olds around mental health?’
“What I found really interesting was and it’ll literally come up with one
the various uses that you can use AI in five or 10 seconds. That would
for, whether it’s professional — there’s usually take me an hour.”
the marketing, operational function.
For us now it really has streamlined Drafting and editing
our operations; it’s made things quite “When a client comes in for their
time-efficient for us as well.” first consultation, we’ll then put
together a script. We’ve been using
Concept development and AI to invent these scripts, which
brainstorming is really saving time. And then I’m
“Since the training, I have been able to kind of do a bit of a cut
using AI as a sounding board on and take out what isn’t suited
everything, and just to re-steer from that script, but it has been
my direction when I lose it a little bit quite effective.”
— almost like with a colleague just
bouncing ideas off it or helping “Lately, I changed my website, so
me synthesise different threads of I took some screenshots of each
bits that I have going on. We work of the pages and then submitted
remotely in our team so it’s really them to AI, and it would give me
useful having a tool to just help feedback on everything. It would
me flesh things out, rather than ask me to change this and that. So
bothering colleagues or spending it was really good.”
ages thinking it over until it comes
to me.” “I’ve been using it [AI] in developing
the content for our website and it’s
been very useful … helping me to
develop the content and also giving
me the strategy on how to develop
the website [...] I’ll just tell it ’This is
what I’ve done so far. What do you
think?’ And, surprisingly, it will come
out with some very brilliant ideas,
and I’ll take it from there.”

66
SMB SECTOR PILOT SMB SECTOR PILOT

CASE STUDY
What we covered
in the training
AI approach Webinar series

to inclusion
Using AI Responsibly
This training series consisted of five for Small Businesses
one-hour sessions with live demos and • An introduction to responsible AI

at Grind
Q&As. Prompts, case studies, and examples • Security and privacy risks of AI
were all tailored to small-business use cases. • Checklist for using AI responsibly

Understanding Machine Learning

• What is AI and the sort of problems you Lunch & learn series
Dan Menezes Melo, a people operations manager at could apply machine learning to These 30-minute micro-sessions were
Grind, joined Google’s AI Works training to explore • Understand how some different machine hosted at a lunch-friendly time over
how AI could enhance his work and support his team. learning models work Google Meet with demos presented
• An overview of how to prepare data for by three guest speaker panellists on
With a background in technology and HR, he saw the a machine learning project the following topics:
training as an opportunity to promote responsible, ethical • Personal Productivity
use while streamlining tasks such as policy writing and Boost Your Productivity with AI • Sales
project planning. The emphasis on AI as an augmentation • Understand what generative AI is • Marketing
tool, rather than a replacement, empowered Menezes and how it works • Operations
Melo to use AI for generating document outlines, refining • Tools to apply generative AI across
communications, and ensuring clarity and inclusivity. a range of tasks Each panellist presented specific examples
• An introduction to prompting of AI tools they’ve implemented in their work.
The training also sparked broader adoption across Grind, The format followed:
with colleagues embracing tools such as Gemini for Getting Hands-on with • Task overview
notetaking and brainstorming, and NotebookLM to create AI Tools for SMBs • Identification and benefits of the
audio overviews. One neurodivergent colleague used AI to • Tools to support with writing content AI tool for supporting the task
simplify complex instructions, boosting her independence: • Tools to support with designing content • Bringing others along on the journey
a transformation Menezes Melo deeply related to as an • Tools to support with understanding and overcoming challenges
autistic professional and non-native English speaker. and condensing large volumes
of information
AI now provides him with the same accommodations Live sessions
that he once relied on from supportive colleagues, from Prompt Engineering for Three day-long live sessions consisted
clarifying jargon to ensuring inclusive language. Menezes Small Businesses of a combination of the above
Melo sees AI as an indispensable equaliser, creating a • An overview of prompt engineering modules delivered to groups of SMB
more accessible, empowering workplace for everyone. • Techniques for crafting effective prompts workers in person.
• Effectively reviewing prompt outputs

67
05