The Public Sector and AI innovation friction

When it comes to tech investment right now, organisations, the public sector included, are being pulled in multiple directions.

At the top of the agenda is a mandate to adopt AI, balanced against strict efficiency targets and shrinking budgets. It is not a dilemma the sector is facing in isolation.

The ambition is not in doubt. What keeps stalling is delivery, and the reasons have stayed remarkably constant while strategies, departments and ministers have come and gone. Start with the money. The National Audit Office has found that government spends around half of its IT budget simply keeping ageing legacy systems running rather than improving services. That is the reality any AI ambition has to work within, and it is not really a technology problem.

The reality check

Before diving into an AI transformation project, it is important to consider the wider governance and budgetary challenges often faced by the public sector.

Joel Hoskins of The Productivity Institute has previously said,

"AI alone will not fix the public sector. While tactical gains can be found in low-effort projects like deploying simple chatbots and off-the-shelf automation tools, these are merely the foothills of the opportunity."

As such, the use of off-the-shelf automation tools, including generative AI tools, must be considered just the beginning of the public sector's AI journey. To take the opportunity to the next level requires the alignment of three fundamentals, Data Culture, Legacy Estate Modernisation, and Public Trust.

Pillar 1: Embedding a true data culture

We have seen AI implementations in the public sector thrive, and we have also seen where they have not reached their potential. Industry research consistently shows that up to 80% of AI projects stall before reaching production due to fragmented and siloed data.

Creating a true data culture demands a change in perspective, one where data hygiene is not seen as an IT tick-box activity but an organisation-wide discipline for which everyone has a responsibility. Before even considering an AI project, clean, accessible datasets must be in place across the organisation, and not just in isolated departments keen to progress on their AI journey.

Pillar 2: Modernising legacy estates

AI also cannot deliver on brittle and disconnected infrastructure. More than nine in ten organisations still rely on legacy systems that consume vast amounts of maintenance budget while creating friction with modern technology.

Transformation starts with considering how modular, cloud-native environments can remove the rigidity of legacy systems, while also helping organisations better manage costs.

With more AI tools and models coming to market, it is important to select the right ones for your organisation. Public sector organisations must consider the right models for their infrastructure and put the correct cost controls in place to prevent runaway consumption.

Doing this demands a shift in how we manage IT tools. Rather than reviewing cost optimisation annually, it must happen in a real-time operating feedback loop, as teams track value against public service outcomes rather than technical consumption.

Pillar 3: Building uncompromising trust

The final pillar, which is unavoidable, centres on establishing trust.

Public sector AI projects cannot compromise on compliance, data sovereignty or public accountability. As we have already addressed, guardrails help the cultural adoption of technology, but they also strengthen governance. Alongside secure data boundaries and verifiable audit trails, these measures provide the essential foundation for public and stakeholder trust.

It is also worth noting that trust is fiscal, particularly when it comes to public funds. Any spend must correlate with measurable service improvements rather than simply being an operational cost.

Ensuring transparency in adoption, and reinforcing the value of human-in-the-loop in AI projects, will not only help the onboarding of internal teams, but also wider stakeholders.

From talk to reality

We have seen this play out firsthand with a North East UK university that moved from its first cloud workload to live AI in the space of a single admissions cycle. The goal was practical rather than experimental. Clearing is the busiest and most unforgiving few days in the higher education calendar, when thousands of prospective students make contact at once and every unanswered enquiry is a place, and a student, potentially lost.

Rather than bolt an AI tool onto the problem, we started with the foundation. We built the university a secure, AI-ready AWS base in a UK-hosted Landing Zone, brought its data into a state where it could actually be used, and mapped the design to the UK Government AI Playbook so governance was built in from the first decision rather than retrofitted. Only then did we put AI to work in student services, with human-in-the-loop oversight and clear guardrails on what it could and could not do.

The result was a Clearing operation that could absorb the surge without adding headcount, answering routine applicant questions instantly and freeing the university's own team for the conversations that actually decide whether a student enrols.

The point is the order of events. The AI worked because the data, the estate and the governance were dealt with first. That is the difference between a pilot that impresses in a demo and a service a public body can stand behind on its busiest day.

Keeping your eyes on the long game

When it comes to sustainable AI adoption, public sector organisations must keep their eyes on the long game. Successful adoption is built on adaptable estates, disciplined spending, and trusted data foundations.

Achieving that works best with an independent partner that knows the public sector, its frameworks and its funding routes, and that can help you access the AWS funding that offsets the upfront cost of your initial proof of concept pilots.

To explore how you can drive innovation in AI without losing control of spend, get in touch with the team or meet us in person at DigiGov Expo at Stand D31.