Public sector organisations are exploring how trusted, governed data foundations can help AI initiatives move beyond trials and deliver measurable public value, says Errol Rodericks, Product and Solutions Director, Denodo.
If you attend any public sector technology event today, it won’t be long before the conversation turns to AI. From central government and local authorities to NHS trusts and emergency services, organisations are exploring how AI can help improve services, reduce administrative burdens and make better use of limited resources.
While there is no shortage of AI projects, far fewer are operating at scale or delivering measurable public value. Many initiatives begin with genuine promise but struggle to move beyond small trials or tightly controlled environments.
When that happens, the discussion often centres on the AI itself when, in reality, many organisations discover the same obstacle waiting for them: the data simply is not ready.
Why good ideas struggle to scale
The public sector has spent decades collecting information. Citizen records, operational data, case files, financial information, healthcare records, planning systems, sensor data and countless other sources generate huge volumes of information every day.
The challenge is rarely a lack of data but knowing where it is, who owns it, whether it can be trusted and how quickly it can be accessed.
Anyone working in government will recognise the problem. Information often sits across multiple systems built at different times for different purposes. Teams may hold separate versions of the same data. Access requests can take weeks. Valuable insights remain locked away because joining information together is harder than it should be.
Whilst none of this is new, AI now exposes these weaknesses very quickly. In fact, recent research suggests this challenge is far from isolated, with 61% of public sector organisations saying they struggle to identify trustworthy data or prepare and integrate it for AI initiatives.
The average public sector AI initiative relies on 479 separate data sources, while more than one in five organisations are working across more than 1,000.
An analyst may be able to work around fragmented information. A motivated project team can often compensate for gaps through manual effort. AI systems, however, are far less forgiving.
If the underlying information is incomplete, inconsistent or difficult to access, the results quickly become unreliable.
Trust matters more than speed
Public sector leaders also face a different challenge from many commercial organisations. Rather than simply looking for efficiency gains, they are making decisions that affect people’s lives.
Whether the issue is healthcare, social care, benefits administration, safeguarding or public safety, there is an expectation that decisions can be understood and justified. That is why trust sits at the centre of every serious conversation about AI in government.
People need confidence in where information comes from, how decisions have been reached and that governance has not been sacrificed in pursuit of speed.
This becomes much harder when information is fragmented across multiple systems and organisational boundaries. Without trust in the underlying data, confidence in the technology itself quickly begins to erode.
Data needs context
There is another dimension to this challenge. AI doesn’t simply consume data, it needs to understand how government and public sector organisations work.
To produce reliable answers, AI must understand what information represents, how departments, organisations and services relate to one another, which policies apply and the responsibilities that shape public sector decision-making.
High-quality data remains essential, but without this operational context AI risks drawing conclusions that are technically accurate yet disconnected from how the government actually functions.
Creating this richer layer of meaning and context is becoming increasingly important as organisations look to scale AI. By connecting data with governance, business definitions and operational knowledge, organisations can give AI the context it needs to generate responses that are not only accurate but also relevant, appropriate and explainable for the public sector.
A change in mindset
For years, the answer to fragmented information was often to bring everything together. Large data programmes were launched with the goal of centralising information into a single repository before innovation could begin. Sometimes those programmes delivered value. But often they became long-term transformation projects that consumed significant resources before users saw any meaningful benefit.
Increasingly, organisations are beginning to ask a different question. Instead of focusing on where data should live, they are focusing on how it can be accessed, governed and shared more effectively.
One local government organisation responsible for delivering hundreds of citizen services recently faced this exact challenge. Information was spread across departments, agencies and operational systems, making it difficult to build a complete picture of citizen needs or support new AI initiatives.
Rather than embarking on another large-scale data migration programme, the organisation focused on creating a trusted data foundation that enabled governed access to information wherever it resided. This improved visibility across services, reduced duplication and created the conditions needed to support AI-powered citizen services, while maintaining governance and security.
The result was faster decision-making, better collaboration between teams and more responsive public services, demonstrating that progress came not from moving the data but from making trusted data easier to access.
What happens next
The public sector’s relationship with AI is still evolving. There will undoubtedly be new tools, new regulations and new opportunities over the coming years. Some will live up to expectations, whilst others will not.
What seems increasingly clear, however, is that the organisations making the greatest progress are paying close attention to the foundations beneath the technology.
They understand that AI is only as useful as the information that supports it. While it is unlikely to generate the same attention as a new model or a new application, it is often the factor that determines whether a promising idea becomes a practical service.
The next phase of public sector AI is unlikely to be defined by bigger ambitions. It will be shaped by organisations that have quietly done the hard work of making their information accessible, with the right context and understanding surrounding it to make it trusted and ready to use.
Ultimately, before AI can transform public services, confidence is needed in the data that powers it. Those that get this right will be on track to achieving better public outcomes.
