高瞻远瞩，从小事做起-在整个政府中释放Gen AI和代理AI的变革力量

What’sinsideFrom future vision to daily realityHow does agentic AI differ from Gen AI?The current use of AI in citizen digital servicesThink big to transform how government worksWith big potential come big risksHow to implement Gen AI and agentic AI safelyand at scaleFinal word 45681112193 AI is everywhere right now; you probably cannot get throughyour morning coffee without hearing about it. When it comes togovernment services, though, AI is not just hype. It has the potentialto transform how millions of people access vital public services everyday – and we are just at the start of the journey.Around the world, there are examples of governments usingautomation, personalized healthcare and intelligent chatbots toserve their citizens. As these technologies advance, and adoptionincreases, public services as we know them may change completely.Rather than being discrete offerings from individual departments,they may merge into integrated, intuitive experiences that respondproactively to citizens’ needs.Examples could include renewing driving citizens’ licensesautomatically, flagging support they are eligible for or guiding themthrough complex applications. And who knows – in a few years’time, a single government web page with a prompt box could be theaccess point for all citizen-facing services.We have generative and agentic AI to thank for these current andpotential ways of delivering public services. Generative AI respondsto a user’s prompts to create new content (text, audio or visual).Agentic AI is a broad term, but in essence, it means AI that can makedecisions and complete tasks independently, without the need forhuman input.1Agentic AI clearly marks a big leap in the evolution of thesetechnologies. But it also begs the question: are governments readyto put their faith in the machine? And if so, how can they do soeffectively and at scale – in a way that protects citizens from harm?1.https://cohere.com/blog/what-is-agentic-aiThe answer lies in thinking big but startingsmall.In this point of view, we will explore how generative and agentic AIcan help governments deliver better services while maintaining theircommitment to human-centered design and digital service standards.We will cover how digital teams globally are experimenting withthese transformative technologies, including our service designteams at Capgemini Invent. (Go to page 13 to see our results.)We will explore what could be holding governments back beforesuggesting how they can start to unlock the value of AI. And, forthose who have already started on the journey, we will suggestways to scale AI iteratively – including by embedding it into existingtransformation programs. All while being transparent and ethicalabout where they deploy AI and how they use citizens’ data.From future vision todaily reality4 | Unleashing the transformative power of Gen AI and agentic AI across government How does agentic AI differfrom Gen AI?Gen AI, as we know it today, excels at creation. Butit operates within defined boundaries, rather thantaking pre-emptive action. Agentic AI, on the otherhand, focuses on doing, and as such, represents amore sophisticated, proactive approach.Think of it this way:Gen AI is like your creativeteammate who comes up with content, code andimagery. Agentic AI is more like your get-things-doneaccomplice who turns plans into action without youneeding to ask.Consider your email inbox for a moment. Instead ofdrowning in emails, a Gen AI copilot helps you draftresponses and summarize long threads.An agentic AI assistant could take this support a stepfurther by auto-scheduling meetings and flaggingconflicting priorities. It could even nudge you to replyto an important email while proposing what it shouldsay – so you stay on top of what truly matters.Another aspect that distinguishes agentic AI is itscollaborative architectural configuration. This meansit can deploy multiple AI agents – for example, onefocused on analyzing data, another on optimizingprocesses, and another on assessing risk – to achievea common goal. Each agent can access differentdata, connect to different tools and follow differentinstructions, creating distinct experts within the sameworkflow.When these specialized agents combine theirexpertise, they can make more nuanced decisions,and address complex challenges more effectively,than a single system attempting to manageeverything. It marks a shift from the “single expert”model to an orchestrated team of AI specialists, eachfocusing on what they do best.² ³2. https://www.cambridgemc.com/accelerating-local-government-with-agentic-ai3. https://www.simpson-associates.co.uk/agentic-ai-making-ai-work-for-you/ The current use of AI in citizendigital servicesWith public finances more constrained than ever,governments globally are under growing pressureto transform and become more efficient. At thesame time, their stakeholders – including citizens,businesses and international partners – expectthem to provide modern, intuitive and