1、Unlocking the UKs AI Potential:Harnessing AI for Economic GrowthT:+44(0)2036 872 761E:infopublicfirst.co.ukwww.publicfirst.co.ukPublic First is an independent consultancy that works to help companies and organizations develop new policy proposals,better understand public opinion,and model their econ

2、omic and social impact.Public First is a member of the Market Research Society.Unlocking the UKs AI Potential:Harnessing AI for Economic GrowthContents5 Executive Summary9 Introduction13 The UKs Digital Advantage17 Infrastructure&Compute21 Digital Skills25 The Returns from Digital Investment27 Reduc

3、ing Barriers to Digital Adoption31 Unlocking Value in the Public Sector33 MethodologyThe UK has the most advanced tech sector in Europe and is second behind only the US in its ability to take advantage of AI-however,this lead is relatively slight over other leading economies such as Germany,Canada o

4、r South Korea.Executive Summary5The UK has the most advanced tech sector in Europe and is second behind only the US in its ability to take advantage of AI-however,this lead is relatively slight over other leading economies such as Germany,Canada or South Korea.Over the next decade,digital technologi

5、es such as AI and cloud have the potential to create a half a trillion opportunity for the UK economy,increasing GDP by over 550 billion by 2035.AI could help make individual workers more productive,level up skills across the workforce,reduce waste,speed up scientific innovation and enable entirely

6、new types of products and services.6550 BnU.S.A.U.K.GERMANYCANADAS.KOREAThe UK could be held back from achieving its full potential by three core bottlenecks:The UK does less well at providing the basic infrastructure needed to enable the build of datacentres.This is only likely to become more impor

7、tant as demand for additional compute and data from larger AI models continues to grow.A shortage of digital skills continues to be a bottleneck for companies looking to adopt and apply AI and other digital tools into their own workflows.In new business polling,we found that digital skills were seen

8、 as the type of skills businesses thought would be helpful to have more of,with this being particularly true for larger enterprises.Many companies are still unaware of the potential benefits to their business from greater use of more digital technology.In our business polling,we saw that around half

9、 of businesses were unaware of compelling use cases.7The next five years could be particularly crucial ones for putting in place the right enablers to ensure the UK can take full advantage of AI.Adding an additional 5 years to the time it takes for AI to fully roll out would reduce the size of the e

10、conomic impact in 2035 by over 150 billion.Investing more in infrastructure,cloud,AI and digital skills could help overcome these obstacles.In total,we estimate that investment in digital technologies and skills could have an average societal Return on Investment(ROI)of over 5:1 in the next decade.M

11、icrosoft,who commissioned this report,has committed to investing 2.5 billion over the next three years in the UK,more than doubling its datacentre footprint and training more than one million people for the AI economy.Alongside its impact in the private economy,digital technology could play an equal

12、ly powerful role in the public sector.In our modelling,we found that the average augmentation potential for public sector occupations was only 15%less than that in the private sector.8Introduction9The promise of technologies like AI are coming into sharper focus for businesses,citizens,and public se

13、rvices.The UK Governments National AI strategy sets out a 10-year plan to ensure all parts of the country can reap the benefits of this technological opportunity.In this short report,Microsoft commissioned independent consultancy Public First to explore:The opportunities created by digital technolog

14、y for the UKThe supporting investments that will be needed in infrastructure,skills and business supportThe current barriers to adoption,leading to unequal take upAs part of our research,we ran a new extensive survey of over 1,000 senior business decision makers in the UK,and created new modelling o

15、f the potential returns from AI,cloud,data centres and digital skills.Digital technology is already one of the most important drivers of growth in the UK.In our polling,a majority of businesses told us that they were using digital technology in most of their core processes,from communicating with cu

16、stomers to monitoring key business metrics.In total,we estimate that in the era of the PC,digital technology has created over 400 billion for the UK economy.1My business uses digital technology for this1 Public First modelling based on the Conference Boards Total Economy Database,looking at the aggr

17、egated impact of ICT capital since 1990.1011Looking forward,AI and the cloud could have an even larger impact on growth.In our polling,59%of senior decision makers expected digital technology to have a greater impact on their business in the next five years than it had in the last five-compared to j

18、ust 2%who thought it would be lower.In total,we estimate that cloud and AI could create a half a trillion-pound opportunity for the UK,increasing GDP by over 550 billion by 2035.Thats the equivalent of raising annual growth rates by 2%a year.However,as this report reveals,the UK risks foregoing this

19、 opportunity.The next five years are likely to be the most important for putting in place the right enablers to ensure the UK can take full advantage of AI.Already,the UK suffers from relatively low levels of compute capacity,significant delays in building new infrastructure,a long tail of SMBs who

20、have not yet fully adopted key background technologies such as cloud and growing demand for both basic and advanced digital skills.In the rest of this report,we look at how the UK can ensure it doesnt miss out on this growth opportunity and explores some of the investment that will be needed to brin

21、g it about.How AI will support the economyAs part of the research for this report,we produced new modelling of the potential from AI,using GPT-4 to classify to what extent AI can support workers in different tasks across the economy.While significant uncertainty remains over the pace of potential fu

22、ture AI adoption-and to what extent the technology will continue to develop-we think a plausible central estimate is that with the right enablers in place,AI has the potential to increase the size of the UKs GDP by 550 billion by 2035.AI will impact the economy in multiple ways,including:Saving time

23、 for individual workers.For many tasks,AI will help workers get more done more quickly,with early studies showing it can often produce productivity improvements of around 20-30%.Levelling up the skills of individual workers.AI tools seem to be particularly impactful for workers at the bottom or midd

24、le of the wage distribution:helping them catch up their skills and expertise with the leaders in their field.Reducing waste and increasing efficiency.AI can help organisations in both the private and public sector triage their resources:identifying emerging problems earlier,and using both economic a

25、nd environmental resources more efficiently.Enabling entirely new types of products and services.Just as the PC enabled the spreadsheet and the Internet the search engine,the rise of AI is likely to create entirely new types of tools and entertainment that would not have been possible before.Speedin

26、g up the development of new science and technology.AI based simulations are already speeding up the time it takes to develop or understand new biological and chemical materials.For example,recent work by Microsoft and scientists at PNNL used a combination of AI and high-powered computing to help dis

27、cover a new type of battery material in just weeks,rather than years.While many people have expressed concern about the potential impact of AI on unemployment,our modelling finds that this is overstated.The vast majority of occupations are made up of a combination of tasks:some which AI can help wit

28、h,and others where it cant.In our modelling,less than 4%of the occupations we looked at saw a majority of their tasks able to be automated.The potential gains from AI will not happen automatically,however-and a slower adoption curve would significantly reduce the potential.Adding an additional 5 yea

29、rs to the time it takes for AI to fully roll out would reduce the size of the economic impact by over 150 billion.12The UKs Digital Advantage13The UK has the most advanced tech sector in Europe,with the highest level of private investment,the most mature consumer digital market and globally leading

30、companies such as Wise,Sage and Builder.AI.Since 2016,the UK has already attracted over 20 billion of private investment in AI,including 2.5 billion of investment announced by Microsoft in November 2023.2 In total,the UKs digital economy and online retail sector directly contributes 227bn in GVA and

31、 supports over 2.6 million jobs in the UK.3 Beyond its tech sector,the UK has other structural advantages in research,culture and skills that make it well placed to take advantage of AI and data.In our business poll,senior decision makers from companies that operated internationally pointed to the U

32、Ks scientific research,regulation system and entrepreneurial culture as particular advantages.Compared to other advanced economies(eg the US,Canada,EU,Japan),how would you say that the UK compares?(internationally operating companies only)2 https:/assets.publishing.service.gov.uk/media/641d71e732a8e

33、0000cfa9389/artifical_intelligence_sector_study.pdf3 Public First modelling for the CCIA,https:/ccianet.org/research/reports/uk-digital-economy/14More systematically,we compiled a new index looking at the UKs relative strengths and weaknesses in underlying economic competitiveness;research;skills ba

34、se;infrastructure;commercial ecosystem and regulation.4 We find the UK is second behind only the US in its ability to take advantage of AI-however,this lead is relatively slight over other leading economies such as Germany,Canada or South Korea.Public First AI IndexWith countries such as France,Isra

35、el and Germany amongst others all making significant investments in AI,the UK will need to keep its foot on the pedal to not fall behind.4 Public First AI index drawing on data from the World Banks Global Competitiveness;Heritage Foundations Index of Economic Freedom;Tony Blair Institute;Statista;St

36、anfords AI Index Report;Tortoises Global AI Index 1516The UK has the most advanced tech sector in Europe,with the highest level of private investment,the most mature consumer digital market and globally leading companiesInfrastructure&Compute17Over the past 13 years,the amount of computing power use

37、d to train AI models has increased by a factor of 350 million.5 Moores law has been beaten with the amount of computing power devoted to AI doubling somewhere between every 3-6 months.The more we scale up our AI models,the more powerful they become-and this trend is expected to continue to drive inc

38、reased demand for compute power for at least the next decade.Having access to ample compute within the UK can be crucial to enabling low latency applications,scientific research,maintaining confidence in data security and allowing new start-ups to experiment.In our business poll,over two-thirds(68%)

39、of senior business decision makers-and four-fifths(79%)of cloud infrastructure users-told us that they thought it was important to have the option to choose data centres located within the UK.The most important reasons they gave for this were legal or compliance(81%),trusting the UKs data privacy la

40、ws(80%)and ensuring low latency or fast access to their data(78%).However,in 2022 the UK had only 1.3%of global compute capacity,6 and on our sub-index for infrastructure the UK came only 11th.London and the Southeast continue to dominate the market with approximately 70%of data centres clustered in

41、 the region.5 https:/www.governance.ai/post/computing-power-and-the-governance-of-ai6 https:/www.gov.uk/government/publications/future-of-compute-review/the-future-of-compute-report-of-the-review-of-independent-panel-of-experts18The Go-Science report and subsequent independent review of the Future o

42、f Compute outlines seven challenges that have held back investment in compute,including a lack of national coordination,skills and concerns over energy usage.In addition to this,many other commentators have pointed to the UKs planning system as a key barrier to delivering faster investment.While pri

43、vate companies such as Microsoft continue to invest heavily and the Government has committed to three new supercomputers by 2025,still more will be needed to meet demand for use cases such as AI,edge computing and other low latency applications.If the UK wants to meet accelerating demand for compute

44、,it will need to do more to unblock current planning barriers and delays,speed up energy connection times and increase zero carbon energy supply to meet electricity demand from new data centres.Infrastructure Index19Microsofts Investment in Datacentre InfrastructureIn November 2023 Microsoft committ

45、ed to investing 2.5 billion over three years,to expand its next generation AI infrastructure in the UK,bringing more than 26,000 GPUs to the UK by 2026.The single largest investment in its 40-year history in the country,Microsoft will grow its datacentre footprint across sites in London and Wales an

46、d potentially expand into northern England.This infrastructure investment will help to meet the exploding demand for efficient,scalable and sustainable AI,cutting-edge compute power,and the needs of the private and public sector waiting to take advantage of the latest cloud and AI breakthroughs.20Di

47、gital Skills21In our polling,digital skills were seen as the type of skills businesses thought would be helpful to have more of,with this particularly true for larger enterprises.Which,if any,of the following general types of skills would you say that your business would find most helpful to have mo

48、re staff with?AI skills were seen as important by around 40%of larger enterprises,but the most pressing current skills shortages were seen to be in more traditional areas such as data analytics or IT.In the next few years,demand for AI specific skills is likely to grow significantly faster as AI wor

49、kflows move increasingly from a prototype to deployment stage.Between 2020 and 2023,demand for AI roles has grown by twice as much as average labour demand.7 7 https:/ specifically,which of the following types of digital skills would you most want to have more of in your business?23Microsofts AI Ski

50、lls Commitment By 2025 Microsoft will support one million people to gain the AI skills they need to start,or move into,a career in technology.The expansion of the Get On programme builds on Microsofts five-year initiative.To date it has trained 1.1 million people in digital skills and helped over 30

51、,000 people build careers in tech.Microsofts expanded AI resources will focus on three key areas:Building AI fluency:Enable workers,job seekers and AI-curious self-learners to build AI knowledge and understanding including responsible deployment.New non-profit partnerships with Generation,Catch22 an

52、d The Princes Trust will support individuals who face greater barriers to employment opportunity.Developing AI technical skills:Enable those with AI knowledge to develop their technical skills and achieve certification through free,online self-learn training on modules including machine learning and

53、 data analysis.Supporting AI business transformation:Support business leaders to manage AI transformation and worker upskilling across new AI products and services in order to take advantage of growth opportunities.Over two-thirds(69%)of senior decision makers told us that digital skills were either

54、 essential or very important to the running of their business-but 40%of businesses reported finding it difficult to find staff with good digital skills,leading to slower deployment of new technology.24Microsoft and Catch22:Digital Edge Digital Edge is a tech employability programme,delivered by Catc

55、h22 in partnership with Microsoft,which supports individuals aged 18+into their first digital job.Paired with an industry expert career coach,participants are offered six months of free coaching,following tech employability workshops,where they work together to learn all about the industry and gain

56、valuable digital and AI skills for their CVs.Since the start,nearly 1,000 participants have started the programme with over 300 progressing into employment or apprenticeships.Naomi Hulston,CEO of Catch22,said:“The application of AI across all sorts of aspects of our lives is hugely exciting and fill

57、ed with potential.But for some people,this potential is tinged with fear;a fear of the unknown,a fear of being left behind and a fear of not having the right skills to be part of the AI revolution.By empowering people to embrace this new wave of technology,well make sure nobody whatever their age,ba

58、ckground or challenges is left behind.”The Returns from Digital Investment25In our business poll,senior decision makers told us about the many benefits they had already enjoyed from digital technology,specifically utilisation of the cloud:Looking forward,the majority of businesses reported believing

59、 a variety of use cases could be relevant for their business,including AI supported drafting(66%),cybersecurity(63%)and to help answer customer queries(60%.)Which,if any,of the following potential future use cases of AI do you think could be relevant for your business?In total,we estimate that inves

60、tment in digital technologies and skills could have an average societal Return on Investment(ROI)over 5:1 in the next decade-that is,for every extra 1 spent by companies,the economy could grow by 5.8 8 Public First estimate based on a weighted average ROI of investment in AI,cloud,data centres and d

61、igital skills.84%85%81%78%of cloud users told us that it had increased efficiency in their businessof cloud users told us that they had found it easier to keep software up to date of cloud users told us that it had helped support flexible working of cloud users told us that it had helped improve wor

62、ker collaboration26Reducing Barriers to Digital Adoption27Despite the potential of digital technology,we continued to see significant variation in the uptake of key technologies.While over 90%of large businesses in the UK have adopted the cloud,almost a third(31%)of SMEs said that currently they don

63、t use it.The story is even more stark when it comes to using AI tools,with almost half of UK SMEs(47%)saying they use neither dedicated AI tools,AI features built into existing applications nor AI driven robotics or machinery.On average,we found that companies whose revenue had grown by more than 10

64、%in the last year were using over 50%more digital technologies than companies whose revenue had stayed the same or shrunk.When we asked companies what the most important barriers were to them using digital technology,the most significant reason was a lack of awareness-with half of companies saying t

65、hey did not know of any compelling use cases to invest in further digital technology.And which,if any,of the following are important barriers which explain why your company doesnt use more digital technology?28When we asked about the barriers to adoption to AI more specifically however,we found that

66、 more concrete obstacles were seen as important:concerns about the quality of outputs(73%),the potential unreliability of AI systems(72%)and their lack of transparency(71%).How important or unimportant do you think the following are likely to be as barriers to your business making more use of AI too

67、ls?These concerns are likely to fade as AI systems become more mature:becoming more reliable,and with best practice established in how to integrate them into other systems and processes.Like any new technology,it will take time to understand where AI can be most helpfully deployed.However,this nervo

68、usness suggests there is also potentially a role for other partners and guidance to support with and teach best practice in how to deploy AI.29Joos uses Copilot for Microsoft 365 to grow Joos,a supplier of mobile charging stations,is using Copilot for Microsoft 365 as it expands in North America and

69、 Europe.With team members across Europe,North America,Ghana and China working with colleagues across time zones is part and parcel of life at Joos.Using Copilot has been a game-changer for collaboration and sharing across the company.“The fact that Copilot quickly generates meeting recaps with notes

70、 and action items has actually changed the way we structure our meetings”says Jeannette Ikonga,Head of Client Success and Customer experience at Joos.Joos is also using Copilot right across the Microsoft 365 toolset from using Copilot in PowerPoint to create branded sales pitches to summarising long

71、 documents in the onboarding process helping to cut down the time for both onboarding meetings and trainings.30Unlocking Value in the Public Sector31AI could be as significant a boost to productivity in the public sector as it will be in the private sector.In our modelling,we find that the average a

72、ugmentation potential of public sector occupations is only around 15%less than that of those in the private sector.In total,we estimate that greater use of AI to support the completion of routine tasks and administration in the public sector could create over 12 billion in savings for the public sec

73、tor by 2030.By 2035 greater use of AI could save the UKs public sector 17 billion.That would be enough to pay the salary for over 330,000 additional nurses.9 This is broadly welcomed by the British public.Take healthcare-in recent consumer polling by Public First:One area where AI could be particula

74、rly powerful is helping with earlier detection of emerging illnesses.We estimate that the use of AI in the NHS to support earlier detection could save an additional 500 million by 2030.9 Based on https:/www.kingsfund.org.uk/insight-and-analysis/data-and-charts/key-facts-figures-nhsMicrosoft Azure su

75、pports NHS Grampian in trialling AI for breast cancer detectionNHS Grampian in Scotland is trialling Mia,an AI solution developed by Kheiron Medical Technologies and supported by Microsoft Azure,to advance breast cancer screening.Instead of requiring two specialists to read each scan,the trial seeks

76、 to demonstrate that AI working with one highly trained radiologist delivers faster results,addresses the mammogram backlog and enhances detection.Mia analysed over 220,000 mammogram images over a three-year cycle,identifying 271 cancers compared to the 261 that would have been detected by just one

77、radiologist.It also detected 47 additional interval cancers than would not have been identified by two specialists together.AI screening promises scalability through Microsofts cloud.Deployments are underway at 14 additional NHS sites,meaning over 500,000 women a year could be screened by Mia.Mobile

78、 mammogram units can also be used across remote areas,with images transferred via the cloud(with patient data de-identified)and quickly read by Mia before being shared with radiologists.3263%61%65%of the UK public agreed that adopting AI tools in the NHS would improve the quality of health serviceso

79、f the UK public said that they would support the greater use of AI in healthcareof the UK public agreed that adopting AI tools could help GPs to become more efficient in their day-to-day jobsMethodology33Estimated Return on Investment from Digital Technologies and SkillsIn order to estimate the pote

80、ntial societal return on investment from digital technology and skills,we took a weighted estimate of the estimated societal ROI from investing in AI,data centres,cloud services and digital skills:For AI,we first calculated the potential productivity increase from AI tools,drawing our main AI model(

81、see below)and the estimates from Mollick et al(2023).We then applied this to a worker on the average UK wage,assuming they would need to invest in at least 3 AI tools.For data centres,we drew on a literature review of public estimates or financial statements on average return for data centre provide

82、rs.We then added to this additional employee compensation,drawing on ONS Annual Business Survey data on the labour share for data processing,hosting and related activities and telecommunications.For cloud services,we combined estimates from Public Firsts internal dataset on average cost savings,reve

83、nue increases and ROI for cloud users,with a wider review of other published estimates.For digital skills,we drew on a new literature review of the average wage impact of digital skills training,combined with an assumption that around 40%of this impact was additional,based on previous survey work on

84、 digital skills students.34Economic Opportunity from AIIn order to calculate the size of the potential economic opportunity from generative AI for the UK we:Building on the methodology of Eloundou et al(2023),we use GPT-4 to classify the probability that a multimodal LLM,or tools built on top of it,

85、could significantly reduce the time it takes for workers to complete over 17,000 combinations of tasks and occupations drawn from the US O*NET database.For our prompting strategy,we followed the structure of Halawi et al(2024),asking GPT-4 to think step-by-step about reasons about why or why not ind

86、ividual tasks could be affected before making its final judgement.We then aggregated tasks up to occupation level,and used our own developed crosswalk to convert to ONS occupations at a 4 digit level.We aggregated up to look at the impact for the economy as a whole,based on each occupations share of

87、 the total economy wage bill,drawing on ONS Annual Survey of Hours and Earnings data from 2023,and the UK economys labour share of income,again taken from the ONS.Based on the experience of previous General Purpose Technologies,we assume that generative AI will take around 20 years to fully diffuse

88、across the economy,and apply a standard S-curve adoption model,starting from 2022.For our public sector estimate,we look instead at a subset of occupations that are primarily based in the public sector,and then apply the same approach as above.Public First AI IndexOur headline AI index is based on s

89、ix equally weighted pillars:Competitiveness.We draw on data from the World Banks Global Competitiveness Index.Research.We draw on data from Stanford Universitys 2023AIIndexReport and Tortoise Medias GlobalAIIndex.Skills.We draw on data from Stanford Universitys 2023AIIndexReport and Tortoise Medias

90、GlobalAIIndex.Infrastructure.We draw on data from the Tony Blair Institute for Global Changes State of Compute Access report and Statista data on the number of data centres per country.Commercial Ecosystem.We draw on data from Stanford Universitys 2023 AI IndexReport.Regulation.We draw on data from

91、the Heritage Foundations IndexofEconomicFreedom Business Freedom index.For each pillar,we give each country a normalised score between 0 and 1,with the best performing country in each dataset receiving 1 and the worst 0.Where we have more than one dataset to draw on,we report an average.35Potential

92、NHS Savings from Greater Use of AI in Cancer DetectionTo calculate the estimated savings to the NHS from AI we first estimated the cost savings from early treatment as a result of earlier detection,drawing on estimates on cost per patient by Laudicella et al dataonAIcancerdetectionrates NHS data on

93、the number of cancer cases diagnosed by stage each year.We then took into account potential labour force savings given the potential reduction in workload for a radiologist using The Royal College of Radiologists census reports data on outsourcing expenditure.Finally we drew on data on the number of radiology departments and estimated costs to set up the required artificial intelligence technology us to estimate the potential net savings to the NHS from AI early cancer detection.36 Public First-All Rights Reserved 2024