《微软：2025代理型AI变革潜力：英国组织的机遇、挑战与实施路径研究报告（英文版）（37页）.pdf》由会员分享，可在线阅读，更多相关《微软：2025代理型AI变革潜力：英国组织的机遇、挑战与实施路径研究报告（英文版）（37页）.pdf（37页珍藏版）》请在三个皮匠报告上搜索。

1、How agentic AI can supercharge business performance,transform public services and unlock the UKs economic growthAgentsof ChangeContentsForeword36The Age of Agentic AI13Opportunities and Obstacles26Performance Redefined31A Blueprint for Agentic Success37Appendix2If the UKs AI journey so far has been

2、transformative,the coming wave of agentic AI is set to be truly revolutionary.Darren HardmanCEO,Microsoft UKForeword2.5bnMicrosoft is on this path too both for our own business and as a trusted AI partner for our public and private sector customers.In the UK,we are investing 2.5 billion in major AI

3、infrastructure,skilling and education programmes and have trained more than 1 million people so far.We are also committed to continuing our work with the government by exploring new opportunities to support its vision of AI-enabled economic growth and public service reform.As you will read,this repo

4、rt provides empirical evidence of the vital role agentic AI,the modern change agents of our time,will play in achieving that vision.The UKs eyes are open to the opportunity,now we all have to act to seize it.The UK has bold ambitions for a thriving AI economy.The scope and scale of the governments A

5、rtificial Intelligence(AI)Opportunities Action Plan is just what is needed to drive growth,create jobs and improve living standards for all.And we are seeing more and more UK organisations invest in AI solutions to optimise workflows,supercharge innovation and improve experiences for customers and e

6、mployees.But this is also just the beginning.If the UKs AI journey so far has been transformative,the coming wave of agentic AI agents that can make decisions and take actions on our behalf is set to be truly revolutionary.Routine and repetitive tasks will be taken off workers bulging to-do lists.Fr

7、eeing them up to focus on more creative,high-value elements of their jobs and careers.Business-critical functions like cybersecurity,customer relationships and risk management will be streamlined,increasing agility and resilience.Collaboration,between different AI systems and between humans and AI,w

8、ill be redefined,boosting performance and competitiveness.However,as this report underlines,many UK organisations are not currently unlocking the prosperous future agentic AI promises.While the majority of leaders and employees acknowledge the benefits on offer,far fewer believe their organisation i

9、s in a position to seize them.Barriers of workforce readiness,strategy and regulatory uncertainty persist.These barriers can and must be overcome.In the pages that follow,we assess exactly where UK organisations are when it comes to the maturity and implementation of their AI strategies today.They m

10、ake clear the need to balance innovation and progress with robust frameworks for governance,transparency and ethics.And they lay out a clear and practical blueprint to help organisations accelerate into this new agentic era,including empowering a new generation of AI-augmented workers.The Age of Age

11、ntic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixExecutive SummaryThe Age of Agentic AI:The Promise,Potential and Responsibility3Agentic AI represents a significant step forward in the evolution of artificial intel

12、ligence,promising transformative benefits for organisations along with renewed economic growth and improved public services for the UK as a whole.Whether agents retrieve,generate or process information,operate autonomously or even perform all of these capabilities together,the potential of this next

13、 wave of technological progress is near limitless.Led by Dr Chris Brauer,Director of Innovation at Goldsmiths,University of London and commissioned by Microsoft,this report explores the opportunity AI agents offer,discusses the barriers to widespread adoption and shares the experiences and perspecti

14、ves of industry leaders,subject matter experts,academics and employees.Executive SummaryThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixExecutive SummaryThe Age of Agentic AI:The Promise,Potential an

15、d Responsibility4 4Key FindingsThere is clear and urgent demand for agentic AI72%of leaders expect AI agents to be fully integrated across their operations,delivering significant value.21%of leaders expect this to happen within the next 12 months.0304The UK risks becoming stuck in neutral on AI Lead

16、ers recognise agentic AIs transformative potential,but barriers of skills,strategy and trust mean just 20%of organisations have scaled AI adoption to date and 50%of leaders report a gap between AI ambition and action.02Agentic AI presents an opportunity to grow the UKs economy,transform its public s

17、ervices and relieve pressure on stretched workersThe research proves a direct and positive link between AI strategy and agentic readiness with the financial performance of businesses and the productivity of public sector organisations.UK workers are under severe pressure More than half of UK workers

18、(52%)feel they are having to do more work than is possible for one person,with 36%of leaders and 25%of employees claiming to do the work of at least two members of staff.01An AI divide is emerging within the UK economy and workforceData shows a lack of AI strategy,preparedness and impact in around h

19、alf of UK organisations while 57%of leaders cite a difference in the efficiency and productivity of workers who use AI and those who do not.05A bright agentic future awaits if UK organisations act nowWe finish by setting out a clear and practical blueprint for agentic success.By laying the right AI

20、foundations today and by following Microsofts three-phase approach to agent adoption,UK organisations can equip themselves to thrive in the agentic AI era and,ultimately,support the governments bold ambitions for AI-enabled prosperity.06The Age of Agentic AI:The Promise,Potential and ResponsibilityO

21、pportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixExecutive SummaryThe Age of Agentic AI:The Promise,Potential and Responsibility5The Age of Agentic AI The Promise,Potential and Responsibility01Executive SummaryOpportunities and ObstaclesPerformance RedefinedA Blu

22、eprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and Responsibility6 6The world stands on the brink of a seismic shift in the evolution of AI.Up to this point,large language models(LLMs)and AI assistants have been busy transforming the way we search for information,creat

23、e content and organise our life and work.But now,the emergence of agentic AI is set to rewrite the rules again,moving the technology from simply responding to our instructions to acting on our behalf.It is undeniably exciting.For organisations across sectors,AI agents open the door to significant im

24、provements in everything from efficiency and productivity to innovation,cybersecurity and more.Meanwhile for employees,they have the power to enhance personal performance,unlock game-changing levels of capacity,reduce digital fatigue and create a better work-life balance.The Age of Agentic AI Satya

25、Nadella,CEO,MicrosoftYou can build a very rich agentic world,defined by a tapestry of AI agents,which can act on our behalf across our work and life,across teams,business processes,as well as organisations.Executive SummaryOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Succe

26、ssAppendixThe Age of Agentic AI:The Promise,Potential and Responsibility7 7The first step for UK organisations hoping to seize this opportunity is to understand exactly what agentic AI is,how it differs from the tools they already use and,most importantly,the transformative impact it could have.Micr

27、osoft defines agentic AI as an AI system with varying capabilities,including making decisions and taking actions to achieve specific goals with reduced or no direct human intervention.This can mean agents with the ability to retrieve and generate content,which have surged in accessibility and scalab

28、ility since the advent of generative AI.But it can also involve more advanced autonomous actions,with agents operating independently and orchestrating other systems to transform business operations.See Three core agentic actions.For many leaders,the potential value of these agents is near limitless.

29、As international retailer Pets at Home has recently found,an agent able to scan millions of customer transactions every day to detect fraud could have the potential to drive a seven-figure annual savings.Or in the public sector,imagine an AI agent in a GPs surgery that automatically retrieves and su

30、mmarises your medical history,so you dont need to repeat yourself.It could interact with other agents to arrange tests at the hospital in real time and automatically send you a letter or notification to confirm.Results and follow-ups are seamlessly organised and no one gets lost in the system.GPs ca

31、n then spend more time with their patients,knowing theyre receiving higher quality care.Agentic AI is an AI system with varying capabilities,including making decisions and taking actions to achieve specific goals with reduced or no direct human intervention.Agentic AI can play a key role in removing

32、 digital drudgery,giving workers the opportunity to spend more time on creative and value adding tasks.At Microsoft,were helping to build an AI economy,investing in digital skills and tackling the AI divide,all pre-requisites to driving AI fuelled economic growth for the UK.Darren Hardman,CEO,Micros

33、oft UKNot so secret agentsExecutive SummaryOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and Responsibility8 8This report considers agentic AI through the lens of three core capabilities,each of which can benefit UK

34、 organisations in different ways.In ActionIn ActionIn ActionThree core agentic actions Retrieval actions Fetch information from grounding data,reasoning and summarising insights and responding to user questions.In action:A customer-facing chatbot for a bank.After a customer logs in and authenticates

35、 their details,they ask,“Whats my current loan balance?”The agent then consults secure internal databases,fetches relevant data,reasons over it and returns a concise,direct answer.01Task actionsWhen triggered by a person or event,enabling organisations to automate workflows and replace repetitive ta

36、sks for users.In action:An HR departments workflow manager.When triggered to create a new employee onboarding workflow,a task agent automates multiple tasks,such as sending welcome emails,setting up accounts and scheduling orientation sessions.Prompted by email interactions with the new member of th

37、e team,the agent can populate forms,initiate processes in HR,learning&development and finance systems,while notifying team members of progress with an estimated start date.02Autonomous actions Operate independently to dynamically plan,act,learn and escalate while orchestrating other AI tools and sys

38、tems to deliver ongoing outcomes.In action:An autonomous supply chain agent.When the agent detects an upcoming storm likely to cause shipping delays,it adjusts reorder thresholds based on sales data and weather forecasts to secure extra supplies of high-demand items.If needed,it contacts suppliers f

39、or revised delivery timelines and consults a human supervisor for authorisation in unusual cases.Finally,it reallocates resources-adjusting production,inventory,and transport schedules-and updates connected systems and the logistics team with the new schedule.03SimpleAdvancedExecutive SummaryOpportu

40、nities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and Responsibility9 9As with previous AI tools,it is important to note that the adoption of AI agents is not about replacing human workers;it is about augmenting them.And this

41、research clearly shows the need for tools that can help reduce the workloads of workers across the UK,whether in the private or public sector.More than half of UK leaders(60%)and employees(52%)feel their current workload is more than that of one person.A third of leaders(36%)and a quarter of employe

42、es(25%)go even further,believing they currently do the job of at least two members of staff.Agentic AI can be a powerful and positive source of change here,evolving job roles,creating new career paths and lightening the load on over-stretched workers.For staff,this will boost productivity and perfor

43、mance,helping increase their own job satisfaction and wellbeing while enabling them to deliver better experiences for customers and other external stakeholders too.Meanwhile,armed with a round-the-clock team of AI agents to support,connect and drive collaboration among their workforce,businesses wil

44、l be better placed to gain a competitive edge at home and abroad.The benefits dont stop there.By freeing leaders and employees from the constraints of over-work while supercharging innovation and performance among public and private sector organisations,agentic AI will play a vital role in achieving

45、 the UK governments growth ambitions.It will also be central to the countrys plans for AI leadership recently set out in the AI Opportunities Action Plan.Supporting actors36%of UK leaders and 25%of employees feel they currently do the job of at least two people.10Executive SummaryOpportunities and O

46、bstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and Responsibility10Read more about how were using agents internally at Microsoft to improve customer and employee experiences.Given the transformative benefits,it is easy to see why the v

47、ast majority of UK leaders see the integration of agentic AI into their organisation as a question of when,not if.Indeed,72%expect it to be fully scaled across all operations,with 39%predicting this will happen within the next two years and more than a fifth(21%)expecting this transformation to take

48、 place within the next 12 months.See Figure 1.The edge of transformation.Meanwhile,two-thirds of leaders(69%)say AI agents will have a positive impact on innovation in their industry during the next five years and more than half(53%)believe they will benefit the nations economy overall.Encouragingly

49、,the findings of this report back up that confidence.It provides compelling evidence of the extent to which agentic AI can drive economic growth and reform public services.We discuss this in more detail during the following sections.Imminent integrationFigure 1.The edge of transformation.21%Within a

50、 year8%Within 6 months39%Expect this to happen within two years72%Expect agentic AI to be fully integrated within their organisations,delivering significant value 11Executive SummaryOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promi

51、se,Potential and Responsibility11Of course,there are challenges to overcome.Despite widespread enthusiasm for AI agents among leaders and employees,the cutting-edge nature of the technology means most UK organisations remain some way off realising their full abilities.Worryingly,this report finds a

52、gap emerging between organisations that are taking concrete actions to prepare for agentic AI and those already at risk of falling behind.In many cases,barriers of skills,budget and regulation continue to hold back deployment and/or impact.This underlines the need for a strong blueprint for how orga

53、nisations accelerate and scale the adoption of AI agents in future.We provide this in Section 4 of this report.Leaders must also accept a renewed(and ever evolving)duty of responsibility to their customers,employees and society at large.The more that AI systems are able to operate,orchestrate and co

54、llaborate without human intervention,the more important it becomes to build robust governance frameworks that ensure they are working to the highest standards of security,transparency and ethics.What is clear is that with the right strategies and risk management systems in place,the arrival of agent

55、ic AI can bring a new wave of social and economic prosperity to the UK.Enhanced operational performance and competitiveness.A happier,more productive workforce.An unwavering commitment to protecting the safety of users.This is the promise,potential and responsibility for leaders who lean into the ne

56、xt chapter of AIs remarkable story.Those that choose not to may find themselves written out altogether.We need to establish a 360-degree analysis of approaches for the new era of AI and emerging tech.This means intertwining trust with economic value while collaboratively shaping the right mindset an

57、d culture to tackle resilience challenges in cybersecurity,technology,and the economy.The story continuesProfessor Hoda Al KhzaimiFounder,EMARATSEC Center for Emerging Tech and Advanced Research 12Executive SummaryOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic SuccessAppendi

58、xThe Age of Agentic AI:The Promise,Potential and Responsibility12Opportunities and Obstacles02Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles1313As we saw in the previous chapter,UK

59、 workers are under pressure,with many feeling as if they are working far beyond what is possible for a single employee.Understandably,this is taking a toll on their own effectiveness and morale as well as the overall performance and growth of the organisations for which they work.There is no single

60、silver bullet solution-yet undoubtedly AI can help.This report uncovers a direct correlation between organisations agentic readiness(i.e.,the maturity of their AI strategy and how this can support the integration of AI agents into their operations)and their overall performance.For businesses,this pe

61、rformance uptick is measured according to their financial success and productivity,while for public organisations,it concerns productivity alone.It can also have a telling impact for employees,with two-thirds(67%)of leaders and nearly half(46%)of employees believing they would be more productive if

62、AI could autonomously handle many of their time-consuming or repetitive tasks.In other words,AI agents may hold the key to alleviating the pressure on stretched UK workers while at the same time improving public services and unlocking the nations broader economic growth.Opportunities and ObstaclesEx

63、ecutive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles1414.1487%believe that further transforming the workplace by speeding up their organisations adoption of AI is important over the next y

64、ear.31%say AI-augmented workers are happier and better satisfied in their jobs.38%LinkedIns 2025 Work Change report reveals that more than a third of UK C-suite leaders(38%)feel employees using AI at work respond faster to customer needs and are more productive and creative.15It is very exciting tha

65、t we are starting to conceptualise AI applications through the lens of autonomy,of acting and interacting.We now have to ensure these models are well-constructed and that we have a good dialogue around what they mean for individuals,for business and for society.Professor Nick JenningsVice Chancellor

66、 of Loughborough University and renowned AI thought leaderExecutive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles71%are looking for ways to create cost reductions through automation.64%seek

67、 efficiency and productivity gains through AI-led workflows and processes.16Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesAI agents can deliver on these goals,helping streamline op

68、erations,reduce bottlenecks and increase speed-to-execution without the need for significant human involvement.By providing real-time insights on operating conditions,customer sentiment and workforce utilisation,they can also increase agility and resilience in the face of market changes,unlock new r

69、evenue streams and enable organisations to deliver better experiences for the people they serve.The value of using AI to augment or automate repetitive tasks should not be underestimated either.In the public sector alone,highly skilled frontline workers such as doctors,nurses,police officers and tea

70、chers could save as much as 23 million hours through the use of generative AI tools.This would,in turn,leave them free to concentrate on higher-value activities,boosting frontline services across education,healthcare,local government and more.When you consider UK organisations strategic priorities,t

71、he case for embracing agentic AI strengthens even further.Almost three-quarters(71%)say they are looking for ways to create cost reductions through automation while two-thirds(64%)seek efficiency and productivity gains through AI-led workflows and processes.Many organisations are also aiming to stre

72、ngthen business resilience by using AI to better detect fraud,forecast risks and enable real-time responses to market changes.Meanwhile,more than half see AI-driven personalisation as a way to enhance service delivery and would welcome agentic assistance for leadership teams in both strategic planni

73、ng and decision-making(58%and 56%respectively).Learn more about how Pets at Home built a fraud detection agent in Copilot Studio.Figure 2.Top priorities.Agents in action:Taylor Wimpey17Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Pote

74、ntial and ResponsibilityOpportunities and Obstacles“We recently made the investment to increase our Copilot licenses from 300 to 1,000,reflecting both a 92%adoption rate and meaningful efficiency gains.Users report an average daily time saving of about 28 minutes-rising to two hours for some-which a

75、llows them to focus on more value-add tasks that may otherwise get sidelined.Employees are also using that extra capacity for upskilling and personal development.Even saving 10 minutes between meetings can reduce stress and boost energy levels.Our pilot has also extended into Microsoft Dynamics AI t

76、ools,providing high-level sentiment analysis of customer emails,classifying them into positive,neutral,or negative.Early results show it helps our customer service teams spot emerging issues faster,before they escalate.Additionally,we are using AI to further analyse the results of our customer surve

77、ys,taken at the eight-week and nine-month mark after homeowners move in,to further understand which topics most affect net promoter scores(NPS)across our business.Because Copilot is our first widely deployed AI tool,adoption and change management have been paramount.Were building a community of user

78、s who are growing comfortable and confident with these tools at a pace that maintains engagementpeople are less likely to drop off if theyre not rushed.AI is part of a broader change effort called Innovate TW,aligning with our ambition to become the UKs first digital house builder.Its about shifting

79、 mindsets and processes to harness technology for greater efficiency,innovation,and competitiveness.While the construction industry has traditionally adopted technology more slowly,Taylor Wimpey is moving at pace to develop its IT capabilities.We encourage our colleagues to experiment with AI and to

80、 submit their own ideas for innovation so they can reap time savings and ensure best practice is shared across the business.It really helps that our CEO Jennie Daly and board champion these tools,offering sponsorship and leading by example.We also communicate that Copilot is meant to enhance,not rep

81、lace,employees.Its about freeing up capacity for higher-value work,such as supporting our customers and improving build efficiency.One of the biggest advantages is that Copilot sits within the Microsoft 365 tools we already use daily,like Word and Excel.Adding a digital assistant on top feels intuit

82、ive.Though the pilot initially focused on office-based roles,were now planning workshops with specific functions and roles,for example production directors and managers to develop role-specific training.By mapping their on-site tasks,well see how AI can streamline processesfor instance,using OneNote

83、 to capture notes and then Copilot to structure and share them.Internal communication remains vital for drumming up enthusiasm.Newsletters spotlighting success stories and a network of change champions help people see how AI can improve their daily work.Of course,responsible AI and governance are es

84、sential.We ensure data security in SharePoint,review access levels,and label sensitive information appropriately.Everyone signs a Copilot acceptable use policy before getting a license,and we are also developing a wider AI policy with our data protection team.Looking ahead,were taking small steps in

85、to agentic AI,exploring Copilot agents for tasks like onboarding or knowledge retrieval.The goal is for teams to create their own agents in Copilot Studio without needing a full-scale IT project each time.Were also planning a hackathon to crowdsource fresh ideas and identify challenges agentic AI mi

86、ght address.With a wide range of operational functions across our business -from technical planning to sales and marketing-theres huge potential to bridge process gaps and make our cross-functional work even more seamless.”Emilia TurnerIT Business Partner,Taylor WimpeyTaylor Wimpey is one of the UKs

87、 largest residential developers committed to delivering new homes responsibly for customers and communities.The goal is for teams to create their own agents in Copilot Studio without needing a full-scale IT project each time.18of UK leaders say their organisation have a formal AI strategy in place-a

88、 notable increase since 2024 but not high enough if the UK is to unlock the full potential of agentic AI.46%Ready or waiting?So,are UK organisations ready to capitalise on the agentic AI opportunity?The simple answer is that today,not all are.In fact,research reveals around half have a plan for AI w

89、ith ambitious ideas for implementation and a serious approach to ethics but around half do not.Roughly half have their data and people prepared for AI but around half do not.And approximately half are feeling the impact of AI through optimised processes,competitive advantage and bolstered cyber defe

90、nces but around half are yet to fully explore this.For example,only a fifth(20%)claim to have advanced and scaled AI adoption more than double the 9%figure we saw in previous Microsoft research in 2023,but still far too low.Meanwhile,half of leaders(50%)say there is a gap between AI ambition and act

91、ion in their company and 41%admit they dont know where to start with AI.Only 2 in 5(40%)say their business is using it to strengthen cybersecurity.Concerningly,just 46%of leaders say their organisation has a formal AI strategy in place another notable increase since 2024 when that figure was 29%acco

92、rding to previous Microsoft research but,again,still not high enough if the UK is to unlock the full potential of agentic AI.Meanwhile,according to a report from market intelligence firm,IDC,just over a fifth(22%)are struggling to identify clear use cases that align with their business needs.Executi

93、ve SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesThe research also reveals a clear distinction between what we term assertive AI organisations and tentative ones.For the former,AI is seen as

94、 a tool for efficiency and customer engagement and they are willing to take a process-first approach to transformation.Their key challenges include training employees and measuring ROI.In the case of tentative AI organisations,leaders recognise the technologys potential benefits but require more dev

95、eloped AI strategies if they are to implement at scale.We consider in more detail how UK organisations are delivering on their AI strategies and preparing for the arrival of agentic AI during Section 3 of this report.Watch how Amey are keeping their front line employees safe with a Health&Safety age

96、nt.19Time to change gearAcross the board,there remains an overwhelming sense that many UK organisations are stuck in neutral gear at the very time they should be accelerating.As well as limiting the value generated by their AI tools,this threatens their ability to future-proof for new innovations wh

97、ile jeopardising their aspirations for growth and competitive edge.After all,if they are waiting to act on agentic AI,their competitors,whether domestically or internationally,may already be surging ahead.As for whats stopping them,the reasons cited are myriad but by no means insurmountable.Nearly t

98、wo-thirds of UK leaders(64%)point to the stymying effect of regulatory uncertainty,a significant increase from 44%in 2024.However,the growing shift towards more innovation-led AI policies and legislation,evidenced by the AI Action Opportunities Plan,may see these concerns reduce in the months ahead.

99、Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles20More than half(52%)of leaders also worry that agentic AI increases cybersecurity risks while 49%of organisations report a lack of AI

100、 governance talent,making it harder to monitor autonomous decision-making systems effectively.However,AI including agentic must be a core pillar of a modern cybersecurity strategy.In fact,recent research has found that businesses that incorporate AI into their security strategy might suffer 20%lower

101、 financial losses after a successful attack.Historic concerns about AI hallucinations and misinformation have also made some firms hesitant to invest yet,in truth,many of these challenges are already being overcome.Thats because enterprise LLM tools,such as Microsoft 365 Copilot,are grounded in an o

102、rganisations data like files,emails and meetings that are based on users real-life work,making errors and hallucinations much less common.Similarly,while workers still harbour some fears about job displacement as a result of AIs expanding capabilities,especially in mid-level roles,these too are dimi

103、nishing.As discussed in Section 2,there is instead a growing a desire to boost productivity and work/life balance through AI tools that autonomously take on their repetitive,time-consuming tasks.This carries the added bonus of making processes and systems more collaborative and efficient at an organ

104、isation-wide level too,boosting performance across teams,functions and departments.Dr Dimitrios Tsivrikos,Consumer and Business Psychologist,University College London“Employees are consumers of their own business,so leaders have to understand and explore with them how they want to actually use AI an

105、d how it can serve them rather than them serve it.”Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesMeasuring upExecutive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAp

106、pendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesAn undoubted pain point for organisations across the public and private sector and one where rapid improvement is needed is measuring the impact of their AI tools.Indeed,while there has been a steady rise

107、in their ability to do so(from 33%in 2023 to 49%in 2025),fewer than half of organisations feel confident in how they calculate the difference their AI investments are making for their business,their customers and their people.However,one in five(19%)of leaders report they have seen a measurable retu

108、rn on their AI investments(ROI)so far.Given many organisations report still being in an experimentation phase,this figure represents strong progress and should be useful for leaders working to secure boardroom support for further spending.Though,it is important to underscore the need for organisatio

109、ns to support new AI solutions with the necessary infrastructure and culture if they are to move from experimentation to impact and experience ROI.As Dr Nigel Guenole,an expert in management and psychology at Goldsmiths,University of London,explains,“You need to know the in-points and the out-points

110、 where your AI will interface with other technology and of course your people.If you dont have that mapping done properly before you start to implement AI,then youre not going to realise its full value.”51%of UK leaders surveyed say their organisation wants to harness AIs competitive advantage but i

111、s not yet ready to do so.21 Read more about how Virgin Money is using agentic transformation to redefine customer experiences.“We recently complemented our GenAI offering with Microsoft 365 Copilot,which we rolled out firmwide.Copilot makes sense for supporting people in the flow of work,for emails

112、in Outlook,notes in OneNote,meetings summaries in Teams.This is where its head and shoulders above other applications.As for most organisations,but especially law firms,data can be a big hurdle.We have vast amounts of it,but its not necessarily arranged in such a way that makes deploying AI tools ac

113、ross it easy,to get meaningful insights.Were now working with Microsoft and investigating how the Fabric analytics platform could help our data strategy.The change management and training weve offered employees around AI has been hard,intense work,but its paid off massively in terms of engagement.Ev

114、ery single member of staff has signed an acceptable use policy and completed compulsory AI training-which includes getting hands on with various use cases straight away.Teaching people advanced,sophisticated prompting techniques,such as chain of thought-with sufficient context and detail to get the

115、best results from the tools has also made an enormous difference-as has having internal legal and technical AI experts to guide our decision-making.We have great relationships between our technical teams and our lawyers-we talk to each other openly,share ideas,and generally try to have fun and speak

116、 the same language.Thanks to everyones hard work we are in a great place,and plan to keep increasing our adoption of AI tools.“Our users are favouring Laila for general chat with file capabilities,as theyre comfortable with the app and have built the habit of using it.It has compare and contrast fun

117、ctionality between some different AI models,and people can upload multiple files of different media types,using AI vision.We have set it up,so these files are remembered during the chat session but are not stored beyond that session once it is closed.We are building out more agentic self-service fun

118、ctions for Laila,such as Chat with HR,where we have trained it on our HR policy documents using the Azure AI foundry.So,if someone from the UK wants maternity leave advice,they can ask and get the right context specific information.Chat with HR is currently in user acceptance testing and after that

119、were going to be testing our new Chat with IT capability,trained on our IT knowledge base.As agentic evolves,its important people get clear guidance around how agents are differentiated and where they can be used.For us,having strict control over how agents are created and shared is essential.Were e

120、xcited about the possibilities but also keen to tread carefully in this respect.“Agents in action:Linklaters“Nothing weve ever developed internally has had the usage quite like our GenAI chatbot Laila has.More than two million prompts submitted across 7,400 unique users represent incredible numbers.

121、Broadly speaking,pretty much everyone in the firm,globally,has used Laila at least once.Weve never seen a response to a technology or applications weve built like this before.Senior leadership is genuinely excited and backing our AI strategy.”Laila was launched in 2023 as an internal experiment in g

122、enerative AI,built entirely in-house using Microsofts Azure OpenAI infrastructure.We stayed in experimentation mode for a few months,but people loved the tool so much we decided to turn it into a formal product after holding a naming competition!And created a formal product team comprising software

123、engineers,QA,Ops and Product Owner.It has continually evolved since then-its now using several models such as ChatGPT-4o,Gemini,Embeddings&Vision,and user numbers have continued to grow.22Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,P

124、otential and ResponsibilityOpportunities and ObstaclesLaura HodgsonGen AI Lead,LinklatersSu ClarkeHead of Software Development&Testing,LinklatersLinklaters is one of the worlds leading law firms,handling some of the biggest and most transformational mandates for the worlds leading corporates,banks a

125、nd financial sponsors.Weve never seen a response to a technology or applications weve built like this before.Senior leadership is genuinely excited and backing our AI strategy.57%of UK leaders report a productivity gap between employees who use AI and those who do not,but only 18%of workers say they

126、 have received training on how to use AI in their job.Haves and have-nots23Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesThe dangers of a more cautious approach extend beyond the o

127、verall performance of the organisation.Over half(57%)of leaders report a gap emerging between the efficiency and productivity of workers who use AI and those who do not.More than a third of leaders(36%)even say frequent AI users are more likely to be promoted in their organisation.As AI agents creat

128、e new opportunities to work smarter not harder,the gap between these high-and low-performing AI-enabled workers is likely to widen unless organisations invest in literacy and reskilling programmes for all employees.Yet right now,only 18%of UK workers say they have received training on how to use AI

129、in their job.This is a pressing issue that if left unaddressed,could lead to a new structural inequality in the workforce:the haves and have-nots of AI augmentation.As the old saying goes:AI will not take your job but a human with AI skills may well do.If UK organisations need any motivation to over

130、come these barriers,they need only look to the scale of the opportunity agentic AI presents.From enhanced productivity and profitability to transformed public services,they have a chance to unlock a prosperous future for themselves and for the UK as a whole if they act now to create the right condit

131、ions for success.In the next section of this report,we therefore look more closely at the effectiveness with which private and public sector organisations are implementing their AI strategies and the impact this is having on their readiness to embrace AI agents going forward.The adoption of AI is a

132、continuum,with each wave preparing the ground for the next.Here,the UK has a comparatively strong platform upon which to build.It is the third largest AI market in the world1 and there is widespread acknowledgement of the positive impact agents could have for organisations,their customers and their

133、employees.Yet currently,just two in five(40%)leaders say their organisation is currently harnessing AI to drive a competitive edge let alone future tools.Knowing the opportunity exists is one thing.But to unlock agentic AIs benefits tomorrow,they must first overcome the barriers that are holding the

134、m back today.Acknowledgement to action1 https:/www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-planExecutive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obst

135、acles242424“Theres a need to to lay strong architectural foundations-supporting interoperability,compliance and zero-trust security,so organisations can evolve their agentic AI as the technology develops,rather than having to reset or rebuild.”Drew Winlaw,Partner and Global LLM lead,Simmons&SimmonsT

136、he number one problem in construction is resource availability,theres not enough high-quality experts to do all the work we could do.There is a real opportunity to improve productivity in our industry.The promise of agentic is that we can start to create process experts trained by our human experts

137、in core,critical and complex areas that,for example,keep onsite people safer,reduce errors and redoing things that erode margins,and eliminate waste to improve sustainability.I couldnt simply point a simple or public large language model at our working at height process because the questions and con

138、siderations are too complex and important.We need an agent thats more knowledgeable about the context and right questions to ask,so they can help guide people through the correct process 100%of the time.Weve identified quality assurance as a key area where AI agents can add significant value,particu

139、larly in how we test what we build and install.Our inspection and test plans rely on multiple data sourcesthe build plan,manufacturer specifications,and site conditionswhich need to be considered together.AI agents can help by analysing and cross-referencing these inputs,debating between themselves

140、to determine the most appropriate inspection and test plan.A final agent then reviews and validates the recommendations,providing a human-reasoned output that guides decision-making.In construction,where multiple interdependent processes create complexity,AIs ability to decipher,reason,and streamlin

141、e decision-making is particularly valuable.Were seeing this come to the forefront in both quality control and safety management.This means health and safety inspectors can spend more time on site rather than filling out or reviewing forms on a PC for a risk and method statement.Just like doctors the

142、 biggest gain with AI is getting them more time with patients.We want our experts at the point of work,rather than sitting behind a laptop in a site office.”“One exciting use case is an in-house solution weve built for categorising job defects-which can be in the thousands for large infrastructure p

143、rojects.It organises issues according to what you should work on,and in what order,to maximise the chance of a first-time commercial handover with the customer.Previously,a team of eight to ten people would log them in an Excel,sift through manually and task them out to onsite staff.Now,the tool can

144、 successfully identify which are critical and which could be deprioritised-like distinguishing between dust in a light fixture and an actual structural crack-with around 80%of the accuracy of a construction director with 25 years experience.Ive also learned that its impossible for a technology centr

145、e to sit in the middle of an organisation and anticipate every opportunity that might exist for AI at the front end.Its better to create a listening post and mechanisms to capture ideas from all the experts in the business,like our My Contribution employee led innovation programme.This helps spot wh

146、ere AI can create the most value out there in the business,and we recently mined these suggestions for a fun and engaging hackathon we ran with Microsoft.By the end of it I couldnt spot who was an IT professional and who wasnt!This created a lot of excitement and ideas that solved a problem 50%to 60

147、%of the way but getting the rest of the way and into production is the challenge.Agents in action:Balfour Beatty25Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesJon OzanneCIO,Balfou

148、r BeattyBalfour Beatty is a leading international infrastructure group with 26,000 employees across the UK,US and Hong Kong.It is leading the transformation of its industry to meet the challenges of the future.We want our experts at the point of work,rather than sitting behind a laptop in a site off

149、ice.PerformanceRedefined03Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles26In the previous section,we considered the opportunity that agentic AI offers UK organisations to thrive al

150、ong with the obstacles that must be overcome if they are to seize it.The good news is that there is a growing recognition among both leaders and employees about what is at stake.Already,three-quarters of businesses(74%)and two-thirds of public organisations(64%)are looking to AI agents to automate r

151、epetitive workflows and reduce operating costs.Likewise,60%and 56%respectively see them as a way to better manage business critical processes,such as detecting and escalating attempted fraud.“I can see agentic AI carrying out an array of currently undifferentiating work,allowing our domain experts t

152、o focus on the work that brings most value to clients while still providing validation and assurance of the work we deliver,”says Tom Heath,Global Director of Data Science and AI at professional services firm,Arup.“Agentic applications could also optimise our workflows by resolving process issues at

153、 different stages of a design project or across disciplines.We have got very smart people,and this is about making the best possible use of their time:focusing on the creative,complex problem-solving that really moves projects forward.”Innovation,speed of delivery and pricing models are also likely

154、to improve,while 60%of private companies and 56%of public organisations feel AI agents can successfully personalise customer interactions and,in doing so,boost satisfaction and loyalty.See Figure 3.As for employees,many are hoping the emergence of agentic AI will boost their career prospects by catc

155、hing errors in their work(63%),helping them better prioritise tasks(50%)and uncovering ways for them to seek promotion(40%)or contribute beyond their existing role(39%).Meanwhile,48%believe AI agents can fill skills gaps and 46%like the idea of being proactively informed about their teams needs.Figu

156、re 3.Leaders priorities for agentic AI“We have got very smart people,and agentic AI is about making the best possible use of their time:focusing on the creative,complex problem-solving that really moves projects forward.”Tom Heath,Global Director of Data Science and AI,Arup27Automation of repetitive

157、 workflows to improve operating costs71%Autonomus detection systems that escalate cases of potential fraud59%AI that personalises interactions to improve customer satisfaction58%AI chatbots that reason to provide reccomendations on how to improve customer experience54%AI that identifies the most pla

158、usible sales targets52%AI that automatically identifies opportunities to grow market share51%AI to proactively predict customer churn50%AI pricing models that dynamically set prices to improve margins46%PerformanceRedefinedExecutive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendix

159、The Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance cultureBut to understand where they have to go next,organisations across the private and public sectors first require a clear picture of where they are now.Using a combination of quantitative and qua

160、litative data analysis,the researchers have created a new benchmark of performance.They then used these insights to classify the nations businesses and public organisations into three categories:Leaders;Chasers;and Laggards.What they found is that the best performing organisations(measured by financ

161、ial success and productivity in the private sector and by productivity in the public sector)are those that have established a clear strategy for integrating AI tools and solutions into their operations and workforce.This,in turn,is allowing them to move more quickly from experimentation into full de

162、ployment at scale.What is more,the core foundations of a business that has successfully integrated AI into their processes data,infrastructure,governance,and culture will be critical to successful agentic adoption.These high performers are best positioned to evolve and adapt their strategies to unlo

163、ck the power of AI agents in the future.See Figure 4 and 5.Figure 4.Private sector performance Are proactive and experimental with agentic AI,seeking to use autonomous systems for lead generation,market expansion,fraud detection,and customer retention.Integrate domain-specific AI models that are tai

164、lored to their specific industry or field.This means they can achieve results faster with a lower data requirement,enhancing the speed and accuracy of decision-making.Use dynamic workforce agentic AI for team optimisation and AI-enabled employee engagement for productivity.Frequently use autonomous

165、energy management and dynamic pricing AI agents to drive efficiency and profitability.Are very confident in the ability of agentic AI to make decisions without human intervention.Are adopting AI cautiously,primarily using it to automate routine workflows,sales and customer service.Experiment with AI

166、-driven executive support but struggle to scale pilot programs and realise ROI from limited rollouts.Are often just tentatively dabbling in agentic AI opportunities and risks.Are less confident than Leaders in the ability of agentic AI to make decisions without human intervention.Are off the pace in

167、 AI adoption and at risk of falling further behind,often relying on scripted chatbot workflows.Remain sceptical of AIs impact and are slow-moving,with senior leaders rarely using AI tools.Are twice as likely as Chasers and three times as likely as Leaders to have no AI strategy in place for growth.H

168、ave low or no confidence in the ability of agentic AI to make decisions without human intervention.Prefer to maintain direct control and involvement in all workflows rather than allowing AI agents to make decisions or take actions on their behalf.LeadersTop 25%by performance ChasersMiddle 50%by perf

169、ormanceLaggardsBottom 25%by performance Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and Obstacles28Going for growthThese classifications underline again how much work is to be done to turn th

170、e promise of agentic AI transformation into reality.But they are also by no means inescapable or permanent.With the right strategies and mindset,the 75%of UK organisations that make up the Chasers and Laggards today could soon progress to become the Leaders of tomorrow.What is more,by preparing for

171、and,ultimately,deploying AI agents,organisations can go beyond simply future-proofing their own operating models and competitiveness.They can also play a crucial role in unlocking the UKs vision of AI-enabled economic growth while making the country a more equal,enjoyable and prosperous place to liv

172、e and work.To redefine performance for the era of agentic AI that is the challenge and the opportunity facing UK leaders and employees alike.In the final section of this report,we lay out a two-part blueprint for how they do it.Figure 5.Public sector performance Have high confidence in AI to drive i

173、nnovation,diversity and inclusion.Feel positive that AI leadership will enhance the UKs position on the world stage.Are keen to embrace agentic AI that promotes national security and combats fraud.Leaders and employees use AI tools in their day-to-day jobs and share a strong belief that AI can impro

174、ve their productivity and ability to perform their roles effectively.Are very confident in the ability of agentic AI to make decisions without human intervention.Are still in the process of learning how to efficiently integrate AI into processes.Tend to prioritise the use of advanced AI to strengthe

175、n cybersecurity defences.Have a good understanding of the resources,talent and upskilling requirements to be successful with AI.Leaders believe their operating model will need to change dramatically in the next five years.Employees feel frequent AI users are more likely to be recognised or promoted.

176、Are less confident than Leaders in the ability of agentic AI to make decisions without human intervention.Yet to complete any training to improve workers understanding and confidence to use AI in their work.Make no or little use of AI to improve decision-making and rarely have a clear or formal AI s

177、trategy.Are not seeing measurable ROI on their early AI investments and do not feel able to efficiently integrate AI into their processes and systems.Have low or no confidence in the ability of agentic AI to make decisions without human intervention.Prefer to maintain direct control and involvement

178、in all workflows rather than allowing AI agents to make decisions or take actions on their behalf.ChasersMiddle 50%by performanceLaggardsBottom 25%by performance LeadersTop 25%by performance Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promis

179、e,Potential and ResponsibilityOpportunities and Obstacles29“At Arup,we have long seen the opportunity for AI to be a tool to augment our employees work,enhancing the design and advisory services we provide to clients.Right now,we are applying the technology in three ways.Firstly,through classic mach

180、ine learning,supporting client facing work and internal operations,such as predicting project health.Secondly,deploying generative AI like Microsoft 365 Copilot to improve productivity and efficiency across the organisation.And thirdly-our most advanced work-through proprietary applications built by

181、 our specialist Analytics and AI team.These include custom chatbot applications to support our technical specialists and AI models that augment our engineering design processes.One example is computational fluid dynamics(CFD)simulations for wind comfort modelling.Traditionally,these simulations take

182、 hours or days to run,limiting the number of scenarios engineers can test.Now,however,by using AI models to approximate CFD simulations,we can run them in seconds,which in turn allows us to optimise across thousands of design variants.This creates the possibility to have live,interactive design disc

183、ussions with clients,testing solutions in real time,while still using the traditional simulation methods to validate the preferred designsI see agentic AI as a bridge between these three AI applications,particularly between off-the-shelf yet customisable tools like Copilot and Arups proprietary mode

184、ls.Accessible AI tools like Copilot are helping us bring the entire organisation along on this journey.Theyre introducing people to AI,showing them how to integrate it into their workflows,and building confidence in the agentic aspects of AI.One of the most valuable applications of this today is dis

185、covery where employees use Copilot as an AI agent to quickly surface internal information about past projects.For example,someone might ask,Which metro station projects have we worked on Asia?Who led them?This is transformative for knowledge management and business development,as it enables employee

186、s to uncover insights they didnt even know existed.Arup is a data rich organisation.Experimenting with some of the graph extensibility mechanisms in the Microsoft 365 environment to bring our proprietary structured databases into Copilot is enabling us to lean on AI to improve the way people are wor

187、king,while we continue to evolve how we manage the underlying information.The next step is showing people how to configure their own agents with Copilot Studio so they have more visibility and control over say,the information sources used to ground that agent and tailor it to deliver responses suite

188、d to their unique role and specific use cases.This will further improve the user experience.After this we have clear plans to look at using agents to carry out more repetitive tasks for employees,while adhering to a process they get to define.This is an incredibly exciting vision,and where real oppo

189、rtunity lies to create competitive advantage,compared to more commoditised applications.Agents in action:ArupI can see agentic AI carrying out an array of currently undifferentiating work,allowing our domain experts to focus on the work that brings most value to clients while still providing validat

190、ion and assurance of the work we deliver.Agentic applications could also optimise our workflows by resolving process issues at different stages of a design project or across disciplines.We have got very smart people,and this is about making the best possible use of their time:Focusing on the creativ

191、e,complex problem-solving that really moves projects forward.”30Executive SummaryPerformance RedefinedA Blueprint for Agentic SuccessAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesThis is an incredibly exciting vision-where real opportunity lies to c

192、reate competitive advantage.Tom HeathGlobal Director of Data Science and AI,ArupArup is a global collective of more than 18,000 designers,engineers and technical experts around the world who are dedicated to sustainable development in the built andnatural environments.A Blueprint for Agentic Success

193、04Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success31Such issues must be addressed if these organisations are to prepare for and succeed in the agentic era.In the following pages,we t

194、herefore set out a practical blueprint for how they transform and thrive with agentic AI.This blueprint has two parts.The first lays out a series of recommended steps organisations can take now to evolve their AI strategy for the agentic era.These steps articulate the groundwork needed for the succe

195、ssful adoption of AI agents in future,based on the actions the highest performing companies are taking today.See Figure 6.Gearing up.To seize the agentic opportunity tomorrow,UK organisations must geton the front foot today.The second part of the blueprint builds from there.Once organisations are co

196、nfident they have the necessary AI foundations in place,the next step is to assess their level of maturity and identify opportunities to evolve and enhance their agentic transformation.These actions centre around the following key areas:Here,Microsoft has developed a simple,three-phase approach that

197、 UK businesses and public sector organisations can follow to progress from initial exploration and experimentation with AI agents,through integration and scaling and,finally,onto continuous innovation and improvement.See Figure 7.A three-phase approach.123Organisational culture and capabilitiesBusin

198、ess strategyAI governance and responsibility123A Blueprint for Agentic SuccessWhile every organisations AI journey is unique,the findings of this report underline the extraordinary promise and potential of AI agents to revolutionise operations,increase resilience and drive growth for everyone.Yet to

199、 seize the agentic opportunity tomorrow,UK organisations must get on the front foot today.And as we have seen,there is a clear and compelling relationship between organisational performance and AI strategy across both the public and private sectors.High performers(those classified as Leaders in Sect

200、ion 3)tend to have more clear AI roadmaps,more value-driven use cases and a culture of purpose-led change.Yet in contrast,mid-performing organisations(Chasers)may experiment with AI in silos but struggle to scale it more widely.Meanwhile,those at the lower end of the scale(Laggards)find themselves g

201、rappling with the stymying effects of scepticism,analysis paralysis and limited leadership buy-in.Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success32Figure 6.Gearing up.This first asp

202、ect of the blueprint applies to those organisations classified as Laggards and Chasers during Section 3.By securing stakeholder buy-in for AI experimentation,upskilling workers and shifting their AI strategy from siloed piloting towards enterprise-wide adoption,they can bolster performance today whi

203、le equipping themselves to become the AI Leaders of tomorrow.Getting your house in orderLaggard Chaser Chaser LeaderLaggard Chaser Chaser LeaderChallenge AI scepticism and establish a clear AI strategy:Directly engage in AI education and strategic planning and establish pilot programs with clear per

204、formance benchmarks.Move from basic chatbots to intelligent generative AI:Deploy AI-driven retrieval and task agents to handle customer interactions,sentiment analysis and real-time query resolution.Increase AI adoption among leadership teams:Ensure executives integrate AI-driven analytics tools,for

205、ecasting and decision-making into daily operations.Shift from passive to active AI integration:Develop a structured AI adoption roadmap,starting with automation projects in customer service,sales and fraud prevention.Seek basic business process automation:Drive efficiency gains with simple AI automa

206、tion in customer service ticketing,sales pipeline management and administration.Optimise lead generation and market expansion:Use agentic AI-driven models to autonomously identify growth opportunities,convert leads and dynamically adjust targeting strategies.Overcome ROI challenges with scalable AI

207、deployment:Transition from isolated AI pilots to organisation-wide scalability,ensuring AI investments drive revenue growth through optimised customer acquisition and service automation.Embed predictive customer retention and service personalisation:Use real-time agentic sentiment analysis to person

208、alise customer interactions,thereby reducing churn and increasing loyalty.Implement AI-driven workforce training and upskilling:Introduce AI-powered reskilling programs that autonomously adjust employee training based on market shifts,customer demand and future workforce needs.Use AI to sharpen deci

209、sion-making:Expand AI beyond basic automation into strategic planning and decision-making,including scenario forecasting and risk assessment.Overcome AI hesitancy and develop a clear AI strategy:Establish transparent AI roadmaps,ensuring structured AI adoption plans and leadership alignment.Increase

210、 workforce AI training and digital readiness:Develop AI literacy programs for workers,ensuring basic AI proficiency in administrative functions and service automation.Transition from manual to AI-assisted decision-making:Introduce AI-powered decision support tools,ensuring leaders leverage AI insigh

211、ts for data-driven policymaking.Build confidence in AI for service automation and productivity:Implement pilot projects for AI-assisted automation,allowing gradual adoption while maintaining human oversight.Strengthen cybersecurity and risk management through AI:Deploy basic AI-driven cybersecurity

212、tools to boost public data protection and fraud detection capabilities.Improve AI workflow integration and service delivery:Create standardised AI adoption roadmaps across departments,ensuring seamless integration into day-to-day operations.Address leadership skills gaps:Develop AI programs to prepa

213、re and upskill leaders for AI-driven governance and service delivery transformations.Strengthen AI investment and measure ROI more effectively:Introduce ROI tracking mechanisms to quantify AI-driven productivity improvements and efficiencies,ensuring sustained AI funding.Incentivise AI career advanc

214、ement:Implement AI adoption incentives,ensuring workers who engage with AI tools receive career development opportunities.Build confidence in AI-driven decision-making:Increase pilot programs for autonomous decision-making,using safe testing environments without full-scale deployment risks.Private S

215、ector Public Sector 12345123451234512345Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success33Figure 7.A three-phase approachOrganisation Culture&CapabilitiesInward-looking,focusing on a

216、n organisations internal environment and employee behavioursBusiness StrategyOutward-looking,focusing on market positioning and competitive advantageAI Governance&ResponsibilityProcesses and standards that help ensure agents are safe,ethical and aligned with valuesWith their AI house in order,UK org

217、anisations can then move onto realising the full possibilities of widespread agentic adoption.To help business and public sector leaders identify their organisations level of AI maturity and then guide them through the process of agentic transformation,Microsoft has designed a clear,three-phase appr

218、oach.It maps the deployment of AI agents against the cultural,strategic and governance actions required to maximise their impact at every stage.A Blueprint for Agentic SuccessExploring&Experimentation Your organisation is just beginning its agentic transformation,embracing a phase of learning and ex

219、perimentation.You are deploying agents for individuals and select areas,laying the groundwork for future development.What you need to do next Integration&Scaling Your organisation is now actively assessing and defining agents as part of its AI strategy.You are transitioning agents from proof of conc

220、ept into production.Delivering transformation across the organisation with a repeatable framework.What you need to do next Continuous Improvement&Innovation Your organisation is scaling agents across the business,continuously integrating them into end-to-end processes.In doing so,you are realising m

221、easurable value empowering people,processes,data,and technology.What you need to do next Communicate a clear AI agentic vision and develop a skills plan to support a value-driven approach.Focus on learning about AI capabilities,experiment with different use cases and understand the technologys poten

222、tial.Run pilot programmes to test AI agents in specific areas.Build workforce awareness of the importance of agents,and the value they offer the organisation.Integrate agents across existing systems and scale up their use across different departments and functions.Ensure diversity of roles and exper

223、iences among those involved in AI projects.Designate agent champions and advocate ongoing learning as the standard.Empower the workforce to build agents and establish co-creation partnerships between business users and IT.Augment organisational capacity by deploying agents continually to support the

224、 delivery of your business strategy.Establish an inter-linked ecosystem of agents that talk to each other,report decisions,execute actions and,summarises performance and lessons learnt all to drive greater efficiencies.Continuously improve the utilisation of agents to maintain effectiveness and adap

225、t to evolving needs.Identify value-led agents by automating processes,enhancing customer experiences,and providing data-driven insights.Establish a cross functional AI council to oversee and guide the development,deployment and evaluation of AI solutions.Utilise agents to enable smarter decision-mak

226、ing and provide data-driven recommendations on strategy.Develop a clear investment plan infusing AI agents across the business,with a clear mapping to business objectives and KPIs.Integrate agents with existing processes and systems to enhance their transformative power.Utilise agents to optimise pr

227、ocesses,reduce costs,and further sharpen decision-making.Enhance business resilience by infusing capacity with agentic solutions to deliver business priorities.Proactively monitor value and evolve AI investment against a clear investment plan.Drive innovation and explore new business opportunities t

228、hrough the use of agents.Build your familiarity with Microsofts Responsible AI principles and raise workforce awareness of the ethical implications of agents and the need for responsible AI practices.Establish ethical guidelines and principles to ensure AI systems are developed and used responsibly.

229、Ensure AI systems are transparent,with decision-making processes that are explainable to build trust with stakeholders.Use pilot programmes to test AI governance frameworks and identify risks.Implement systems to monitor AI compliance with ethical guidelines and regulatory requirements.Define polici

230、es and guidelines to address AI ethics,data privacy,security,transparency,explainability,and interpretability of result.Integrate AI governance policies into existing business processes and workflows.Identify and mitigate AI-related risks,such as biases and data privacy issues.Continuously update an

231、d improve AI governance frameworks to adapt to changing business needs and regulatory requirements.Utilise AI governance practices to drive innovation and explore new business opportunities.Actively engage with stakeholdersincluding customers,employees and regulatorsto ensure transparency and build

232、trust.010203Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success34Agents in action:Simmons&Simmons35Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and Responsibilit

233、yOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success“This is a watershed moment for AI and automation in the legal sector.Now we have the tools we need to make use of the massive quantities of unstructured data we deal with.There are huge opportunities to streamline thing

234、s,and digital workers completing tasks asynchronously could be very powerful.My role is to help guide the firm through this period of transformation.We began by enhancing AI literacy across our workforce and are now working on developing innovative services to generate new income streams,anticipatin

235、g changes as clients adopt AI.Internally,were using Microsoft 365 Copilot firmwide to achieve incremental productivity and efficiency gains across all the admin work staff must manage,and I think many people underestimate the sheer amount of administration involved in modern work.And the connection

236、to Microsoft Graph data delivers real benefits here.We have also launched PercyAI-an Azure Powered AI Chat for Legal Professionals,designed to support more complex tasks including the forensic analysis of longer,highly detailed documents.Percy uses a specialised system to focus on knowledge that our

237、 people bring to Percy,providing answers that are well-referenced and accurate.When weve shown people how the newest version of Percy can help them its no exaggeration to say weve heard shrieks of excitement!Were expecting it to deliver considerable employee productivity gains and for it to signific

238、antly enhance the way we execute our core legal and business tasks allowing us to go further and to unlock new areas of work.And in the S&S Wavelength team,we have the worlds first regulated firm of legal engineers-a group of professional data scientists who help our lawyers and their clients use AI

239、 to take on tasks that are too big or too complex even for Percy.Examples of these include mass categorisation and tagging of thousands of documents,or say,performing very detailed inference with a set of documents,one of which is 3,000 pages.Lastly,were also helping clients implement AI themselves.

240、This work has included an agent that translates documents while maintaining formatting,another agent for drafting specific types of agreements-which people can email using natural language to ask for amendments.Plus,another agent for comparing contracts against your own legal playbook and risk posit

241、ions,which it will mark up accordingly.In all of this,managing the human factor is most critical for success.The technology is powerful -and to ensure our people can exploit it to its full potential,we need to bring them along on the journey in a step-by-step way.This means working with our AI Champ

242、ions,building employees AI literacy,setting the right expectations of what an LLM can do,helping Theres a need to lay strong architectural foundations-supporting interoperability,compliance,and zero-trust security,so organisations can evolve their agentic AI as the technology develops.them recognise

243、 how to prompt not just for the outcome they want,but also to make it easier to check and verify completions.Its going to be fascinating to see what emerges when we have human and digital workers operating side by side.How do we best manage maintenance for a growing suite of personalised agents,when

244、 people leave a firm or move into a new role,leaving an orphan agent?How do we ensure autonomous digital assistants call for human input or handover tasks to people at the right time,to the right person,and in the right manner?Orchestrating the interplay of human-agent workflows and training profess

245、ional social etiquette will become a key part of the agentic AI culture shift.These questions emphasise the need to lay strong architectural foundations-supporting interoperability,compliance,and zero-trust security,so organisations can evolve their agentic AI as the technology develops,rather than

246、having to reset or rebuild.Its incredibly exciting,because were going to have to think differently about what speed means in the era of agentic AI,and what really great client service looks like.”Drew WinlawPartner and Global LLM Lead,Simmons&SimmonsSimmons&Simmons is an international law firm with

247、2,600 professionals across 21 offices worldwide.36Agents of ambitionExecutive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success“Weve got to challenge that mindset of fear because the far bigger

248、 risk is that if we dont go for it,were left behind by those who do.AI is the way to secure growth to raise living standards,put money in peoples pockets,create exciting new companies and transform our public services.”As UK Prime Minister The Rt Hon Sir Keir Starmer,put it,“New technology can provo

249、ke a reaction a sort of fear,an inhibition,a caution,if you like and because of the fears of a small risk,too often,you miss the massive opportunity.Weve got to challenge that mindset because the far bigger risk is that if we dont go for it,were left behind by those who do.The prize within our grasp

250、 is the path to national renewal.And AI is the way to secure growth to raise living standards,put money in peoples pockets,create exciting new companies,and transform our public services.”Likewise,leaders must take stock of their organisations AI journey to date,using their successes,failures and le

251、ssons learned to identify how they nurture a future-focused culture,empower an agentic-ready workforce,and promote a continuous cycle of measurable improvement.The opportunity is clear:by embedding a new generation of AI agents at the heart of its private and public organisations,the UK stands on th

252、e brink of a future of renewed economic prosperity,enhanced public services and a more productive and engaged workforce.The agentic era is here.We all have to be ready for it.Without doubt,the UKs determination to be a global leader in AI is a cause for optimism.Like LLMs and AI assistants before th

253、em,agents are set to revolutionise how we live,work and connect.By establishing the right balance between of robust governance and trailblazing innovation now,the country can position itself at the leading edge of this change.To support the governments bold AI ambitions,private and public sector org

254、anisations must play their part,moving beyond ad-hoc experimentation to structured,enterprise-wide deployment.Crucially,this is not just about acquiring new AI tools;it requires a pervasive strategic,cultural and operational shift.The Rt Hon Sir Keir Starmer,UK Prime Minister36AppendixResearch metho

255、dologyAgents of Change is a Microsoft commissioned report,published on 5th March 2025 and conducted in partnership with Dr Chris Brauer,Director of Innovation,Goldsmiths,University of London between December 2024 and March 2025.Dr Brauers research team included Research Director Dr Jennifer Barth an

256、d a core group of economists,psychologists,data and social scientists from UK-based research and insights firm Symmetry.They used a mixed method approach to build a model,scorecard and blueprint of agentic AI.Surveys of 1,480 UK leaders and 1,440 UK employees were conducted via YouGov between 21st J

257、anuary and 3rd February 2025.For the quantitative analysis,multiple-item Likert scales were developed,ranging from“strongly disagree”to“strongly agree,”to assess participants responses.These scales were used to test hypotheses about the differentiators of low and high performance using descriptive s

258、tatistics,including means,frequencies and subgroup comparisons.This provided a detailed understanding of the data in relation to their research questions.For the qualitative data analysis,the study used AI-driven sentence transformer embeddings to analyse open-ended text responses.These embeddings s

259、erved as input for k-means clustering,allowing the identification of thematic clusters within the free-text comments.This combination of advanced AI techniques and manual interpretation facilitated a deeper understanding of the qualitative data,complementing the quantitative findings.The research te

260、am also interviewed a range of subject matter experts and contributors,including:Professor Hoda A.Alkhzaimi,EMARATSEC Center for Emerging Tech and Advanced Research Tom Heath,Global Director of Data Science and AI,ArupProfessor Nick Jennings,Vice Chancellor of Loughborough UniversityDr Nigel Guenole

261、,Management&Psychology,Goldsmiths,University of London Dr Dimitrios Tsivrikos,Consumer and Business Psychologist,University College,London 373737Executive SummaryAppendixThe Age of Agentic AI:The Promise,Potential and ResponsibilityOpportunities and ObstaclesPerformance RedefinedA Blueprint for Agentic Success