The public sector is bullish on agentic AI – but data readiness will be its undoing
Public sector agentic AI efforts are pointless without the necessary data infrastructure and governance frameworks


Nine-in-ten public sector organizations worldwide are planning to explore, pilot, or implement agentic AI in the next two to three years, according to research from Capgemini.
But while two-thirds are already either exploring or actively working on generative AI initiatives, only 21% of public sector organizations told researchers they have the right data for training and fine-tuning models.
The Capgemini study found that, within the next two to three years, 39% of public sector organizations plan to evaluate the feasibility of agentic AI, 45% intend to explore pilot programs, and 6% plan to scale their existing agentic AI initiatives.
Nearly two-thirds have progressed to pilots and scaled deployments or are exploring generative AI tools, with this number rising to 82% in defense agencies, 75% in healthcare, and 70% in security.
"With rising citizen demands and stretched resources, public sector organizations recognize the ways in which AI can help them do more with less,” said Marc Reinhardt, public sector global industry leader at Capgemini.
"However, the ability to deploy Gen AI and agentic AI depends on having rock-solid data foundations."
It’s this aspect of agentic AI adoption where organizations are facing acute problems, the study noted. Eight-in-ten public sector executives cited data security issues as a main barrier to adoption, while three-quarters cited limited trust in AI-generated outputs.
Get the ITPro daily newsletter
Sign up today and you will receive a free copy of our Future Focus 2025 report - the leading guidance on AI, cybersecurity and other IT challenges as per 700+ senior executives
In the EU in particular, there was a distinct lack of confidence about complying with the EU AI Act, with fewer than four-in-ten saying they were up to the task.
In terms of data management and utilization, there's been little progress in the last five years, Capgemini found. Only 12% of organizations considered themselves very mature in activating data, while 7% report being very mature in nurturing data and AI-related skills.
Only a fifth of the public sector organizations surveyed said they had the required data to train and fine-tune AI models, for example.
AI success requires visionary leadership
As the focus on data quality increases, chief data officers (CDOs) and chief AI officers (CAIOs) are coming to the fore, according to Capgemini. Two-thirds of public sector organizations now have a CDO, while a quarter plan to appoint one.
Meanwhile, more than a quarter of public sector organizations have a chief AI officer, with 41% planning to appoint one.
"Looking ahead, governments can be more agile and effective as AI augments the work of government employees to source information, conduct policy analysis, make decisions, and answer citizen queries," said Reinhardt.
"However, to reach this future, governments need to focus on building the right data infrastructure and governance frameworks."
AI could be the key to public sector efficiency
Public sector efficiency has been in the spotlight on both sides of the Atlantic for some time now, but with the advent of generative AI in late 2022, tech industry and government leaders alike have highlighted the potential gains of the technology.
A report from Microsoft last year, for example, found that UK public sector employees could save up to four hours a week per person by using AI tools, which would represent a significant boost to productivity if actualized.
UK prime minister Kier Starmer has been highly vocal about public sector AI adoption, with the Labour government charging ahead with a host of initiatives aimed at driving efficiency through the technology.
But once again, infrastructure challenges have been cited as a key barrier. The report from Microsoft noted that any widespread adoption of the technology would require significant investment to modernize data infrastructure.
A separate report published in January this year warned that “archaic” legacy infrastructure is holding back the public sector from modernization efforts, and in turn limiting the potential for generative AI to deliver benefits.
MORE FROM ITPRO
Emma Woollacott is a freelance journalist writing for publications including the BBC, Private Eye, Forbes, Raconteur and specialist technology titles.
-
CrowdStrike prepares for battle as Delta given go-ahead for outage lawsuit
News Delta Air Lines has been told it can take CrowdStrike to court to seek damages after last summer's outage forced the airline to cancel 7,000 flights.
-
Michael Dell talks up the power of human and AI collaboration – but not everyone’s singing the same tune
JPMorgan Chase head of technologies outlines plans for greater generative AI use at Dell Technologies World 2025
-
ReliaQuest targets international growth, agentic AI gains with $500 million investment
News Cybersecurity firm ReliaQuest has raised $500 million as part of a funding round aimed at accelerating international growth and product development.
-
Generative AI adoption is 'creating deep rifts' at enterprises: Execs are battling each other over poor ROI, IT teams are worn out, and workers are sabotaging AI strategies
News Execs are battling each other over poor ROI, underperforming tools, and inter-departmental clashes
-
‘If you want to look like a flesh-bound chatbot, then by all means use an AI teleprompter’: Amazon banned candidates from using AI tools during interviews – here’s why you should never use them to secure a job
News Amazon has banned the use of AI tools during the interview process – and it’s not the only major firm cracking down on the trend.
-
Former Scale AI leaders raise $4.2 million to simplify agent deployments
News Applied Labs wants to take the hassle out of AI agent deployments
-
A big enforcement deadline for the EU AI Act just passed – here's what you need to know
News The first set of compliance deadlines for the EU AI Act passed on the 2nd of February, and enterprises are urged to ramp up preparations for future deadlines.
-
This startup wants to simplify AI adoption through a single “orchestration” platform – and it includes access to popular models from Meta, OpenAI, and Anthropic
News Nexos.ai wants to speed things up for enterprises struggling with AI adoption
-
Execs are happy to let AI make decisions for them, and it’s got IT workers worried
News IT decision makers are more cautious than the C-suite with AI adoption
-
Accelerated, gen AI powered mainframe app modernization with IBM watsonx code assistant for Z
whitepaper Many top enterprises run workloads on IBM Z