AI in government: new research reveals DDaT priorities

Artificial intelligence has moved from theory to practice in the public sector, but the real question now is not whether government will use AI, but where and how fast.

At the 2025 Government Transformation Summit, over 225 senior public servants were asked to make real choices about where they would invest their time in peer discussions. With competing sessions and only an hour to spare, each delegate’s decision represented a clear signal: what they were willing to prioritise, not just what sounded interesting on paper.

This “revealed preference” approach offers a unique lens on AI adoption in government. It cuts through the noise of downloads and survey rhetoric, exposing where senior leaders are truly preparing to act, and which conversations, however lively, are unlikely to translate into investment any time soon.

What leaders really prioritise

Across more than 100 potential discussion topics, four consistently rose to the top:

• AI strategy – defining departmental vision, aligning investments with goals, sequencing adoption
  
  
• AI use cases – identifying where AI can deliver measurable improvements in services and efficiency
  
  
• Operationalising AI – moving beyond pilots into production, embedding AI into workflows
  
  
• AI governance – creating structures to manage risk, trust and accountability at scale

These four pillars represent what could be called the active demand of AI in government. They are where leaders are ready to commit resources, not just attention.

Other areas sparked curiosity but fell lower in priority. Generative AI is widely discussed but rarely sits at the top of the agenda on its own. Similarly, AI human teaming and intelligent automation attract interest, but leaders remain cautious, waiting for stronger use cases and clearer ROI.

At the opposite end are issues like ethical AI and citizen trust and AI. Important, certainly, but they currently sit in the “speculative noise” category: widely acknowledged, seldom prioritised as standalone initiatives.

Why the divide matters

The gap between what is talked about and what is acted upon is not just academic. For vendors and policymakers, it is a practical roadmap for engagement.

AI strategy, governance and use cases dominate because they are the building blocks of sustainable adoption. Leaders are less interested in flashy pilots than in the frameworks that will allow AI to scale safely across government. As one departmental AI lead put it: “The question isn’t whether we’ll use AI, it’s whether we’ll use it well. My priorities are about making AI a trusted, everyday tool, not a fragile experiment.”

This distinction also reflects maturity. Government has moved past shiny object syndrome. What matters now is integration with legacy systems, risk management and delivering public value.

Different priorities at different levels

The data also reveals a clear split between senior leaders and delivery managers.

• Senior leaders such as directors, DGs and CxOs focus on governance, long-term strategy and proving ROI. They want AI framed as an enabler of resilience, policy alignment and organisational agility.
  
  
• Delivery leaders including IT heads, service managers and transformation leads care about AI that solves immediate operational pain points. Their mantra is clear: plug into workflows, strip out manual steps, show results quickly.

This divergence reinforces a key go-to-market lesson: the AI narrative must be tailored to altitude. A message about resilience and capability growth resonates in senior lesdership, but frontline managers need to hear about efficiency and ease of integration.

Different organisations face different pressures

Not all parts of the public sector face the same AI challenges. The research highlights distinct patterns across central departments, arms-length bodies (ALBs) and local authorities:

• Central departments prioritise governance and trustworthy AI under heavy political and compliance scrutiny. Solutions must be enterprise-grade and interoperable.
  
  
• ALBs sit between policy and service delivery, juggling dual accountabilities. Their top concern is interoperability with parent departments.
  
  
• Local authorities, under relentless fiscal pressure, are most interested in AI that improves frontline services and staff productivity without costly disruption.
  
  

As one local government digital leader put it: “We’re not starting from greenfield. We’re trying to deliver better outcomes using the people and platforms we’ve already got.”

The hidden priorities

Beyond the obvious demand, the research also uncovered hidden priorities. These are topics that didn’t attract wide interest but, when selected, ranked as urgent.

Take building trustworthy AI. It rarely makes the popularity lists, but when leaders choose it, it signals high-stakes conversations about explainability, compliance and ethics in deployment. Similarly, agentic AI is still niche, but matters deeply to those exploring autonomous systems in complex operational environments.

These hidden priorities suggest that beneath the surface, pockets of government are preparing to tackle advanced AI challenges ahead of the mainstream.

Where the conversation is heading

For all the focus on strategy and governance, some of today’s lower priority topics are likely to resurface as tomorrow’s battlegrounds. Generative AI, human-machine teaming and AI for service delivery all fall into this category.

Right now these areas lack ownership, clarity or immediate mandates. But their consistent appearance on “topics of interest” lists signals a growing curiosity. Vendors and policymakers who engage early, helping to define problems, shape governance and align stakeholders, stand to influence how these fields mature.

Cutting through the noise

So what does this mean in practice?

• Active demand areas such as strategy, governance, use cases and operationalising AI are where government is ready to buy, build and scale.
  
  
• Latent demand such as generative AI, automation and human teaming is where conversations are exploratory and thought leadership can still shape direction.
  
  
• Hidden priorities such as trustworthy AI, ROI and agentic AI may offer early-adopter opportunities for those willing to go deep.
  
  
• Speculative noise should not be ignored, but is unlikely to unlock budgets without being tied to higher-demand agendas.
  

The bottom line

AI adoption in government is no longer about pilots and experiments. Leaders are focused on frameworks, governance and embedding AI into the machinery of public service. What they want is practical, trustworthy and scalable solutions, not hype.

The signal from this research is clear: if you want to engage government on AI, focus less on what’s fashionable, and more on what’s actionable.

You can download the full research and behavioural deep dives below.