Artificial intelligence has gone from boardroom buzzword to budget line item in UK local government. Freedom of Information requests to some of the country’s largest city councils confirm that AI spending is climbing year-on-year, with projections pointing to significant growth over the next two to three years. Councils are investing in workflow automation, predictive analytics and digital collaboration tools, all in pursuit of the efficiency gains that Westminster is demanding. The Prime Minister has suggested that AI adoption could save the UK economy £45 billion annually, and public sector AI contract spending hit a record £1.17 billion in 2025 according to Tussell.
The direction of travel makes sense. For local authorities under relentless pressure to deliver more with shrinking budgets, AI is a natural place to look for productivity gains. But the findings tell a consistent story about what happens when organisations adopt new technology without addressing the complexity already sitting underneath it. A global survey of over 700 professionals across IT, customer experience, finance and operations, found that organisations lose seven percent of annual revenue to software complexity, and that nearly one pound in every five spent on software is effectively wasted on unused tools, failed implementations and hidden costs. Scaled across the UK, this complexity tax amounts to roughly £32 billion a year.
Councils are operating in exactly the kind of environment where this complexity takes hold. Many harbour legacy systems that fail to integrate cleanly, with data held in silos that no single team can access in full. Layering AI on top of that without a clear plan for how it fits into existing infrastructure risks compounding the very inefficiencies it is supposed to solve. The good news is that the steps required to avoid this are practical and well understood.
The first is auditing before procuring. The instinct when facing a new challenge is to buy a new tool, but the research suggests this frequently backfires. Over half (53%) of organisations surveyed had not seen the expected return on their software investments, and 77% of implementations ran longer than planned. Before committing to new AI platforms, councils should be evaluating what they already have, identifying which tools are actively used, which are sitting idle, and where functionality overlaps. Understanding the true state of the technology estate means any new investment lands on solid ground rather than adding another layer to an already fragmented environment.
The second is ensuring that AI amplifies what teams can already do. The research found that employees lose an average of 6.8 hours per week navigating complicated systems and fragmented tools, equivalent to nearly a full working day spent managing technology rather than doing meaningful work. With this, for council staff already stretched thin, that time is coming directly out of service delivery. Councils that opt to pilot AI in small, well-supported use cases before scaling will get better results than those that rush to deploy across departments. This can also be achieved through investing in training alongside the technology itself. When teams understand how a tool fits into their workflow, adoption sticks, but when they’re left to figure it out alone (which 32% of respondents said happened when vendor support fell short), frustration builds and the investment does not deliver.
The third is making simplification a measurable objective. Specifically, councils should be tracking the number of redundant processes across departments, the time staff spend on manual workarounds, as well as the number of tools that perform overlapping functions. These metrics create a baseline against which AI’s actual impact can be assessed, moving beyond the assumption that more technology automatically means more efficiency.
From FOI data, a mixed picture is painted, some councils published formal principles for responsible AI use while others are still developing their approach. Councils that are able to set clear governance frameworks before going all in on deployment, defining what success looks like and how AI fits within a coordinated programme, will be far better positioned to demonstrate value.
The opportunity that FOI data is pointing to is genuine. Councils are committing real budgets to AI and central government at Westminster is actively encouraging them to do so. AI can transform how local authorities allocate resources, engage with residents and deliver services. But the research is clear that technology alone does not drive productivity gains. The councils that will turn that rising spending into real returns are those willing to simplify before they scale, to define the problem before selecting the product, and to measure the impact of what they adopt. The alternative is more systems, more fragmentation and more of the complexity that is already costing UK organisations billions each year.
