Generative and agentic AI technologies hold huge promise for governments. From proactively serving citizens to improving policies, they can help deliver next-generation citizen services at scale and create a smarter, more efficient government that benefits everyone.

Yet despite a strong appetite for adopting these digital technologies, various factors are preventing governments from unleashing their transformative power.

Our new point of view explores how governments are experimenting with Gen AI and agentic AI and what could be holding them back. It also covers how they can overcome these barriers and start realizing big gains. All while being transparent and ethical about where they deploy AI and how they use citizens’ data.

64% of public sector organizations are exploring or actively working on Gen AI initiatives, and 90% plan to implement agentic AI in the next 2-3 years

“From ambition to execution: data and AI mastery in government”, Capgemini Research Institute, May 2025

How agentic AI differs from Gen AI

While Gen AI excels at creating new text, audio or visuals, agentic AI can make decisions and complete tasks independently, without the need for human input.

Take your email inbox. Gen AI may already be helping you to draft responses and summarize long threads. An agentic AI assistant could take this up a level – auto-scheduling meetings, flagging conflicting priorities, nudging you to reply to important emails and suggesting what to say.

Agentic AI can also deploy multiple AI agents to achieve a common goal. When these specialized agents combine their expertise, they can make more nuanced decisions, and address complex challenges more effectively, than a single system attempting to manage everything.

Three ways Gen AI and agentic AI could transform government

Here are just three ways governments could deploy these game-changing technologies:

Agentic AI could spot potential issues early and step in to offer support before problems arise – marking a shift from reactive to proactive services. It could also help remove duplication of effort and budget behind the scenes while leading to faster responses and better decisions.

Agentic AI could act as a citizen’s personal assistant (with their consent), automating the process of claiming welfare benefits and proactively suggesting support. Gen AI and agentic AI could also help people with accessibility needs to access digital services.

Policymakers rely on vast amounts of written and numerical content to draft regulations, analyze research, and assess policy impact. Both Gen AI and agentic AI can streamline this process, so decision-making is faster, more informed and more inclusive.

How to implement Gen AI and agentic AI safely and at scale

Our point of view recommends that governments start by thinking big about the widespread transformation these tools could bring. It then suggests how they can start small and scale iteratively, putting the basics in place and using a tried-and-tested approach from pilot through to widespread adoption.

Find out more