Artificial Intelligence is reshaping Digital Transformation across the UK public sector, but rising ambition is exposing gaps in governance, quality assurance and operational resilience. Rob Jones, UKI Public Sector Lead, Tricentis, tells us that while AI can accelerate delivery and improve efficiency, only strong testing foundations, clear accountability and embedded quality controls will ensure trust, stability and safe adoption in mission-critical public services.
Across the UK public sector, Artificial Intelligence (AI) is rapidly shifting from concept to reality. Public bodies are under sustained pressure to deliver more with fewer resources, modernise ageing systems and improve citizen services in an environment of heightened scrutiny. AI promises help on all fronts. It can automate manual tasks, spot errors faster than humans and accelerate software delivery.
But as the pressure to deploy AI grows, an uncomfortable truth is emerging. Confidence in AI is rising faster than readiness to use it safely. And in government, that imbalance carries real risk.
Speed matters, of course: citizens expect responsive services and policymakers want quick wins. But in essential public services, trust matters much more. If AI is going to shape the next phase of digital government, it will not be decided by who adopts it first, but by who deploys it responsibly, transparently and with quality at its core. In some cases, getting this wrong is not just inconvenient; it can have life-or-death consequences.
Confidence is outpacing control
Recent research paints a clear picture: UK public sector technology leaders are enthusiastic about AI. The vast majority (92%) plan to increase AI usage in software delivery and quality assurance over the next year, while almost as many (88%) are confident in autonomous AI making decisions about when software is ready to go live.
At the same time, only one-third (34%) believe they have adequate controls in place to validate those AI-driven decisions. This gap between ambition and assurance should concern everyone, from senior leaders to citizens who rely on these systems every day.
When critical systems fail, the impact can be widespread, whether those systems sit in the public or private sector. What matters in government is that these platforms underpin essential public services, including mission-critical healthcare delivery, social care and public safety. Even short-lived outages can erode public trust, which is far harder to restore than any piece of infrastructure.
The hidden cost of poor quality
This is not a hypothetical risk. Poor software quality is already costing the UK public sector dearly. Public sector organisations lose an average of £1.34 million every year due to software quality issues, according to the public sector results from the Tricentis Quality Transformation Report.
More worrying still is the operational risk. Nearly two-thirds of public sector organisations (62%) believe they are at risk of a system outage. Many also report increased security and compliance risks, along with falling staff morale and higher turnover linked to constant firefighting.
These challenges did not start with AI, but AI can amplify them. When organisations automate decisions on top of fragile systems, they risk scaling failure rather than fixing it. There is also a cultural issue at play. Speed is often rewarded more visibly than stability. Teams are under pressure to release software quickly, even when testing is incomplete or the understanding of end-to-end processes is unclear. The Quality Transformation Report findings show that more than three-quarters of UK public sector organisations routinely release untested code. In that context, introducing AI without strong governance simply speeds up existing risks rather than delivering genuine innovation.
Trust is built on basics, not buzzwords
If trust is the goal, then “getting the basics right” must come first. That sounds simple, but in complex public sector environments it requires deliberate effort.
The first basic is clear ownership. AI systems rely on data and software that often span departments, agencies and suppliers. Without clear accountability for data quality, system changes and decision outcomes, it becomes impossible to explain or defend an AI system’s actions. Systems must be explainable if they are to be used responsibly.
The second is governance that keeps pace with technology. This does not mean slowing everything down with bureaucracy. It means defining where AI can make decisions autonomously, where human oversight is required and how decisions are validated before they affect live services. Quality assurance should be embedded throughout delivery, not treated as a final hurdle before go-live.
The third is a shift from reacting to failures to preventing them. Too many teams spend their time fixing issues in production because they lack the tools or time to test properly upstream. AI can play a valuable role here, provided it reinforces testing and assurance rather than cutting around them. Using AI to make decisions faster is only valuable if those decisions are correct.
Shared services, shared responsibility
These principles are especially important in the context of government shared services. Large-scale platforms such as Software-as-a-Service ERP systems promise consistency and efficiency across departments and arm’s-length bodies. But they also introduce complexity.
Regular updates are pushed centrally, often on tight schedules. While shared service teams are responsible for delivering the core platform, individual organisations still rely on local systems and processes that connect to it. If those end-to-end journeys are not tested thoroughly, small changes can have outsized impacts.
Quality assurance in this model is a shared responsibility. It requires collaboration, clear communication and an understanding that trust in the service depends on how well it works for everyone, not just whether the central system was updated on time.
Quality as a measure of success
For years, Digital Transformation has been measured by speed, cost savings and innovation. Those metrics still matter, but as AI becomes more deeply embedded in government operations, quality and trust must be treated as core measures of success in their own right.
Quality is not a technical detail to be left to IT teams. It is a strategic issue that affects budgets, reputations and public confidence. When citizens interact with digital services, they do not see algorithms or release pipelines; they see whether the service works, is fair and can be relied on.
The next phase of AI in government will be shaped by these perceptions. Departments that rush ahead without sufficient safeguards may move quickly, but they will also stumble more often. Those who invest in strong foundations will move with confidence, even if progress feels slower at first.
In the long run, trust compounds. It allows organisations to innovate without fear, because they know their systems are resilient and their decisions defensible. In public service, that trust is everything.
AI will undoubtedly play a major role in the future of government. But it is trust, not speed, that will decide whether that future is a success.
