Government departments should rethink how they collect, structure and govern data if they are to realise the full potential of artificial intelligence, according to senior public sector data leaders speaking at Think Data for Government 2026.
During a closing panel on the future of data in government (pictured), speakers from HM Revenue & Customs (HMRC), the Department for Energy Security and Net Zero (DESNZ), the Independent Football Regulator and Snowflake shared how government’s approach to data will need to evolve as AI becomes more widely adopted across the public sector.
Preparing for the next generation of AI
Anna Ibrahim, chief data architect at DESNZ, argued that traditional approaches to data management are no longer sufficient.
“For decades up to now… we have been preparing the data for analytics, for insight, for reporting – structured, rigid,” she said. “I think the frontier is different.”
Instead, organisations should prepare data so that AI systems can interact with it more effectively.
“We want data prepared for interaction with it. We want AI to ask the right questions. We may not know the right questions.”
She added: “How do we prepare the data that takes us further than conventional analytics… towards picking up trends and patterns that are not necessarily clear from the insight from analysis?”
Later in the discussion, Ibrahim suggested government also needs to rethink how data is modelled.
“We’ve learnt to speak by concepts, by storytelling, and maybe we should look at how we model our data the same way,” she said, while stressing she was not suggesting abandoning existing data structures altogether.
AI is changing the economics of technology
Rob Lee, chief data architect at HMRC, said advances in AI are dramatically reducing the cost and time needed to build digital services.
“I think the rise of AI has pushed down the cost of doing technology so much,” he said. “Once upon a time… we’ve got a whole bunch of partners… and now you can sit down in your bedroom over the weekend, and you can write a whole system in AI.”
As technology becomes easier to build, he argued, expectations of government services will inevitably increase.
Lee also said organisations should not admit data into their platforms without understanding exactly what it represents.
“We don’t let any data into our platforms unless we know what it is,” he said.
Public sector AI needs a different approach
Rebecca Norton Price, chief digital and data officer at the Independent Football Regulator, said public sector organisations need to balance growing enthusiasm for AI with realistic expectations.
If you liked this content…
“There is a lot of enthusiasm around AI, particularly from quite senior business leaders,” she said.
She argued that government differs from the private sector because it is not driven by the same commercial pressures to become an early adopter.
“Effectiveness is equally, if not more important.”
Looking ahead, Norton Price said she wanted the regulator to use data to identify early warning signs so it could intervene before football clubs encounter more serious problems.
Data quality remains the biggest challenge
Responding to audience questions, Pollyanna Jones, director of EMEA public sector strategy at Snowflake, warned that poor data quality continues to limit what organisations can achieve with AI.
“I think data debt is the biggest unrecognised problem that we have at the moment,” she said. “Until we improve the quality of the data pipeline and data flows, I’m not sure how useful it really is.”
Jones argued that improving data quality and giving government experts the tools to transform data would be more valuable than relying solely on increasingly powerful AI models.
Telling better data stories
The panel also discussed how public sector organisations can build greater confidence in data initiatives.
Lee said repeated storytelling had helped HMRC build support internally for adopting Unique Property Reference Numbers (UPRNs) as a standard for address data.
“I’ve told stories about it in every forum, so they understand it, and they get it,” he said.
Norton Price said linking data strategies directly to organisational objectives had proved effective in helping colleagues understand both the opportunities and risks associated with technology investment.
Balancing innovation with sustainability
The panel also considered the environmental impact of AI. Drawing on previous work at the Department for Education, Ibrahim described how older or infrequently accessed data had been moved into lower energy “cold storage” to reduce unnecessary resource consumption.
She also argued there is still an education gap around the environmental cost of AI.
“I think that’s an education gap,” she said. “We can address it through architectural designs… but it’s also an educational and human element to it.”
Closing the discussion, speakers agreed that while AI will continue to transform government, its success will depend on organisations investing in better data quality, stronger standards and common approaches to managing information, rather than focusing solely on the latest technology.
