The shift from AI experimentation to AI readiness in UK public services

Bacon: The biggest constraint we’re seeing is everything around the model. Many public sector organizations are working with estates built over decades, with technical debt, dispersed data and dependencies that make even small changes difficult.

AI increases the pressure, making integration, security, assurance, skills and service management even more important. The challenge is now to create the right conditions for governing, supporting and scaling AI.

Cheal: Large, estate-wide transformations can still be necessary and deliver important outcomes. However, scale isn’t the same as progress. A major program can look decisive, but it can also create long timelines, complex dependencies and pressure to show value before the underlying work is completed. These issues are particularly challenging in public services because continuity isn’t optional.

A better approach is to think in terms of what has to change to improve a specific service or outcome. Some systems need substantial modernization. Others may be stable, well-controlled and fit for purpose. The goal isn’t to replace everything but to make continuous changes that remove barriers without introducing unnecessary risk.

Cheal: It starts with a service problem, not a technology target.

Here’s a good example from the United States. When the US federal government began to wind down pandemic-era programs, one state government had to ask more than 1.5 million existing recipients to reapply for benefits. Older technologies and working practices in the state’s contact centers forced citizens to wait up to two hours on calls, and post-call work could take as long as 25 minutes.

As part of a broader modernization program, Kyndryl helped replace interactive voice response and call center solutions, redesign workflows, and improve routing and authentication. After launch, nearly 3,900 agents were able to automatically receive information about past conversations and transfer calls more easily to relevant community partners. The platform also supports generative AI, and improved reporting and controls give both the public and the legislature better visibility into operations and performance.

But the technology work isn’t the core lesson. What matters in this context is that modernization was anchored in service capacity and continuity.

Cheal: Selectively rather than as a single, universal requirement. Since the UK government doesn’t have an overarching digital sovereignty policy and there’s no universal definition or agreed outcome, it’s more useful to treat sovereignty as a workload question than as a blanket objective for the entire estate.²

Leaders should instead focus on understanding where sensitive data is processed, who can access it, how services can be recovered, how risk is managed and whether workloads can move or work together as requirements change. The right answer may differ by service.

Bacon: The work becomes more deliberate. Agencies first need to align on the service outcome, priority services, acceptable levels of risk and decision-making criteria. From there, they can assess readiness gaps across the estate and operating model, then prioritize opportunities based on value, feasibility and risk.

The next step is to prove the approach through a bounded delivery or use case that tests the technology and the controls and operating model needed to support it. Once that foundation is proven, agencies can scale by reusing assurance evidence, platforms, skills and delivery patterns spanning other services.

Overall, the question shifts from “Can we deploy this?” to “Can we operate it safely, recover it when necessary, and improve it over time?”

AI is giving public sector leaders a reason to rethink modernization, not as a one-time replacement program but as a series of targeted decisions. Within this framework, the goal isn’t to modernize everything at once. The strategy is to make deliberate changes that strengthen resilience, improve services and create the foundation to scale AI so it can consistently deliver meaningful value.

A Kyndryl AI Readiness Assessment and Executive Workshop can help you identify priority opportunities, uncover readiness gaps, define a practical 90-day modernization roadmap, and establish a controlled proof of value.