Reflections from FutureScot Public Sector AI 2026

Introduction

On 21 May 2026, the University of Strathclyde’s Technology and Innovation Centre hosted FutureScot’s Public Sector AI conference. More than 400 senior public sector officials, digital leaders, academics and vendors spent the day discussing how artificial intelligence will reshape public services across Scotland and, by extension, across the rest of the United Kingdom.

The headline number from the day came from a recent paper by Storm ID, Automate Tasks, Not Jobs, presented by Paul McGinness. Their analysis put the prize at 62.1 million staff hours of recovered capacity per year by 2030 in their high adoption scenario for Scotland alone (Storm ID, February 2026). That is the equivalent of 30,000 full-time teachers, doctors, nurses, social care workers and police officers. The prize is capacity, not cuts.

The interesting question, posed quietly but persistently by speakers throughout the day, is why the UK public sector is not closer to that prize already. The conference offered a clear answer, repeated from very different vantage points by government, by clinicians, by academics, and by the vendors themselves. The constraint is not the technology. It is the operating model, processes and capability that sit around the technology. Until those are addressed, AI investment continues to deliver pilots rather than scaled capability.

This blog sets out the case for that argument, draws on the day’s evidence and on published UK research, and explains how Reinvigoration’s Simplify4Scale® methodology is built to close the gap.

The constraint is not the technology. It is the operating model, processes and capability that sit around the technology. Until those are addressed, AI investment continues to deliver pilots rather than scaled capability.

The size of the prize

Storm ID’s analysis of fifty high-priority workflows across Scotland’s public services identified an annual baseline of 178 million staff hours spent on routine administration, coordination and records maintenance, with the bulk of the automation opportunity concentrated in a small number of repeatable workflow patterns (Storm ID, February 2026). The concentration matters. It means the prize is not scattered evenly across hundreds of unrelated problems; it sits in a handful of operational patterns that recur across every part of the public sector.

Scaled to the United Kingdom, the prize is materially larger. The Tony Blair Institute for Global Change has published a series of estimates on AI productivity potential across UK government, ranging from £10 billion annually by the end of the current parliament to £40 billion annually with the right data infrastructure (Tony Blair Institute and Faculty, Governing in the Age of AI, 2024 and 2025). The Institute’s local government analysis projects £8 billion in annual productivity gains for English and Welsh councils, equivalent to one million work hours per council per year (TBI, Reimagining Local Government, May 2025). The UK Government’s own AI Opportunities Action Plan, published in January 2025, frames the potential savings even more ambitiously.

The National Audit Office’s March 2024 study of AI in government found that 70 per cent of UK central government bodies surveyed were piloting or planning AI use cases, but only a minority had deployed any AI in production (NAO, Use of artificial intelligence in government, 2024). Two years on, the gap between exploration and operational deployment has narrowed, but it has not closed. The implementation gap is the difference between what public sector leaders intend to do with AI and what is actually delivering value in operational reality.

None of this is theoretical. It is sitting on the desks of UK public sector chief executives, finance directors and operations leaders right now. The question is what to do about it.

70 per cent of UK central government bodies surveyed were piloting or planning AI use cases, but only a minority had deployed any AI in production.

Why the gap exists

The day at Strathclyde surfaced four root causes of the implementation gap. Each was raised independently by speakers from different organisations and different disciplines. Each is, fundamentally, an operating problem rather than a technology problem. Each is also exactly the kind of problem Reinvigoration has been solving for the last twenty years.

Strategy that does not cascade into operational priorities

Scotland’s AI Strategy 2026 to 2031, published on 20 March 2026, is ambitious, well-constructed and credibly resourced. The same is true of the UK Government’s AI Opportunities Action Plan. The challenge is not in the strategy documents themselves. It is in the discipline that takes a published national strategy and converts it into operational priorities at agency, council and trust level.

Dr Tom Wilkinson, Scotland’s Chief Data Officer, was disarmingly direct on this point. Do not rush to apply AI. Focus first on good process for maximum effect. That is the voice of a Chief Data Officer asking his audience to slow down and do the strategy deployment work properly, before the AI investment is committed. Eilidh McLaughlin, Deputy Director for Digital Ethics, Inclusion and Assurance at Scottish Government, reinforced the same theme from the inclusion and trust angle. Understand the personas, understand the use cases, really understand the problem, before designing the solution.

Without that discipline of cascade, national AI strategies become headlines rather than operating plans. The NAO’s 2024 review found that the draft strategy for AI adoption did not even set out which department had overall ownership and accountability for delivery (NAO, 2024). Ambition without disciplined deployment delivers slides, not services. This is exactly the work Reinvigoration does. Our strategy deployment practice uses disciplined hoshin and X-matrix techniques to convert published ambition into operational priorities that frontline teams can actually execute against.

Operating models that are not designed for AI-augmented work

The most operationally vivid example of this came from Emily Hill, who leads Agentforce and Agentic AI for the public sector at Salesforce in EMEA. Hill argued that the real constraint in agentic AI deployment is rarely the technology itself. It is the system around the agent. In a crisis, what matters is how fast you can act, and the systems, processes and operating model around the AI are what slow you down.

Her example was Thames Valley Police. The force handles approximately 1.3 million calls a year. Roughly 400,000 of those, just under one in three, are failure demand: calls generated because something else in the system has gone wrong. The deployment of Bobbi, a virtual AI assistant for the public, now handles more than 76 per cent of relevant interactions, with only the remaining quarter requiring human attention. The cost differential is stark: £7.40 for a contact centre call against 24 pence for a Bobbi interaction. The value is not in the agent’s clever language model. The value sits in the operating model redesign that surrounds it: the routing rules, the escalation logic, the human supervision tier, the integration with the systems of record.

Sopra Steria’s Gary Craven made the same point more bluntly. Agency equals redesign. You cannot drop an autonomous agent into an unchanged operating model and expect different outcomes. Roles change. Accountabilities change. Supervision changes. Exception handling changes. Most public sector bodies have not done this redesign work, which is why their agents stall in pilot. Operating model design for AI-augmented work is one of the sharpest needs in the UK public sector today, and it is one of the areas where Reinvigoration is most regularly called in to support clients.

End-to-end processes that AI inherits, broken or otherwise

Colin Birchenall, Chief Technology Officer for the Digital Office for Scottish Local Government, described the operational reality of most councils with admirable honesty. Lots of inefficiency. Multiple systems. Manual double entry. Customer contact captured in one system and actioned in another. AI dropped on top of this landscape accelerates whatever process it inherits. If the underlying process is broken, the AI delivers broken outcomes faster.

 If the underlying process is broken, the AI delivers broken outcomes faster.

The structural insight from the day was that most public sector inefficiency is not unique to each organisation. It clusters into a small number of repeatable workflow patterns that recur across health, education, justice, policing and local government. That is good news. It means the process redesign work, done once well, can be reused across many services. End-to-end process improvement, applied with discipline and rigour, is where the bulk of the AI implementation gap is actually closed. It is also Reinvigoration’s core craft.

Absent continuous improvement discipline when AI stumbles

Perhaps the most illuminating session of the day came from Dr Dervla Carroll, Clinical Innovation Fellow at the Digital Health Validation Lab and a resident doctor in NHS Greater Glasgow and Clyde. Carroll described the deployment of AI to support radiology in the Queen Elizabeth University Hospital Emergency Department, against a backdrop of 143,000 people in Scotland waiting for testing, 61 per cent of them waiting more than six weeks, and a 29 per cent UK-wide shortfall in radiology workforce.

Her observation was simple and important. When an AI tool appears to fail, the temptation is to switch it off. Most clinical teams reach for the off switch when they experience something they lack confidence with. Carroll’s experience, however, is that the failure is almost always in the system around the AI rather than in the AI itself. Sticking with the technology, applying continuous improvement principles, running structured root cause analysis on what actually went wrong: this is what gets you from a failed pilot to a scaled clinical capability. She described an A3 problem-solving approach to optimising the trial of new use cases that any practitioner of operational excellence would recognise immediately.

This is the most under-discussed dimension of the implementation gap. Organisations without continuous improvement maturity respond to AI failure by retreating. Organisations with continuous improvement maturity respond to AI failure by investigating, learning and adapting. The difference, over time, is the difference between a pilot graveyard and a scaled operational capability. Reinvigoration’s twenty-year history is in building continuous improvement maturity inside client organisations. That same discipline is now the assurance mechanism that AI investment quietly depends on.

A useful nuance on data readiness

One further observation is worth recording, less as a root cause and more as a piece of useful context. The day exposed a quiet but visible split between public sector and private sector speakers on the question of data readiness. Public sector voices, including Tom Wilkinson, Mark Parsons of EPCC and Eilidh McLaughlin, were notably more cautious. Parsons noted that while Britain ranks fourth in the global OECD index for data and digital design, it ranks twenty-second for open, useful and reusable data. The data exists, but it is siloed and inaccessible. Private sector speakers were considerably more bullish, urging the audience not to wait for a data nirvana and to trust the technology they already have. Both positions are rational. Bridging them, with the right operating model and the right governance, is part of the work.

What good looks like: Simplify4Scale® as the operational answer

Simplify4Scale® is Reinvigoration’s signature methodology for organisations facing operational complexity. It is modular by design, allowing clients to begin small with a focused diagnostic and to scale into full transformation as confidence and value are proven. Crucially, it is built around four sequential phases that map directly onto the four root causes of the AI implementation gap exposed at Strathclyde.

This is not a coincidence. The methodology was developed over twenty years to address exactly the kind of operational complexity that the AI conversation has now exposed at national scale. What follows is how each phase of Simplify4Scale® responds to the evidence of the day.

Discover: quantifying where the prize sits

Every public sector body needs to know where its own operational burden lies before it commits AI investment. National-level analysis is useful for sizing the prize, but each organisation’s share of it must be quantified locally before the right interventions can be designed.

Our Discover phase is the diagnostic instrument that does this. A fixed-price engagement, typically four to eight weeks long, that baselines performance, maps core value streams end-to-end, identifies and quantifies AI use cases and improvement opportunities, and identifies the specific barriers to AI adoption. It is a low-risk, low-commitment entry point that gives leadership the data they need to make confident investment decisions. It also provides a stronger foundation for implementation.

Enable: building the internal capability AI requires

The most consistent theme of the day, repeated by speakers across every sector, was that capability must be built inside the organisation rather than rented from outside. Chloe Celani of Dual Track Capability put it bluntly: the best transformations are led from within, not outsourced. Lee Dunn, Head of the Digital Academy at Scottish Government, made the same case from a policy angle, describing the work to build digital and AI capability across the public sector workforce.

Our Enable phase trains the client’s own change team in Simplify4Scale® through a Lean Competency System accredited programme, equipping them to repeat the methodology self-sufficiently. This is the bridge between Discover and delivery. It is also the difference between a transformation that fades when consultants leave and one that compounds over time. Capability is the asset. The methodology is the means of building it.

Simplify: redesigning the processes the AI will inherit

This is where the implementation gap is mostly closed. The Simplify phase co-creates future-state processes with the client’s newly trained change team, implementing the AI use cases and process improvements identified in Discover, and realising the benefits for the organisation and its customers. It is the operational answer to Sopra Steria’s "agency equals redesign" and to Emily Hill’s point about failure demand at Thames Valley Police.

Most importantly, this is the phase that ensures AI sits on a solid operational foundation when it is implemented. The AI inherits whatever process it is dropped on. If the process has been simplified, end-to-end, with handoffs eliminated and failure demand removed, the AI accelerates good work. If it has not, the AI accelerates broken work faster. The Simplify phase is what makes the difference.

Scale: embedding continuous improvement as the AI assurance mechanism

The fourth phase aligns executive teams, establishes an AI and continuous improvement centre of excellence, and embeds the capability across the enterprise. It is the operational answer to the consensus across the day that public sector AI needs centres of excellence working in common ways. It is also the answer to Dr Dervla Carroll’s observation about clinical teams reaching for the AI off switch. Organisations with embedded continuous improvement maturity do not switch AI off when it stumbles. They run root cause analysis, they apply A3 thinking, they evolve the solution. This is how a pilot becomes a scaled capability.

The Scale phase is where the prize is actually realised. The headline AI productivity numbers, whether quoted by vendors or by Whitehall, do not arrive in pilots. They arrive at scale, sustained by embedded operational excellence, supported by aligned leadership, and assured by a centre of excellence that owns the long-term capability.

The Reinvigoration position

Most public sector AI pilots fail not because the technology is wrong, but because the operating model around it is broken. Simplify4Scale® is the operational bridge between AI ambition and AI value. We diagnose where the burden sits, redesign the processes the AI will sit on, and build the internal capability your organisation needs to scale safely. The result is AI that delivers in operational reality.

Reinvigoration is not an AI technology firm. We do not build agents, architect data platforms or deliver machine learning operations. The day at Strathclyde was full of organisations that do, and they do it well. What we do is the operational and organisational work that those technologies need in order to land. We are the methodology, process and capability partner that AI delivery partners need their clients to have in place. We work alongside the vendors, not against them.

What to do next

For public sector leaders reading this with their own AI ambitions, strategies or stalled pilots in mind, there are three practical steps we would invite you to consider.

Start with a Discover diagnostic

A fixed-price, four-to-eight-week engagement that baselines the operational burden in your organisation, surfaces the barriers to AI adoption, identifies the use cases, and quantifies the prize on offer. This is the lowest-risk way to convert an AI strategy into an operational plan, and it gives leadership the evidence they need to commit further investment with confidence. To start a conversation about a Discover engagement, contact us via reinvigoration.com.

Talk to us about AI strategy deployment

If your organisation has a published AI strategy or roadmap but is struggling to cascade it into operational priorities, we can help. The strategy deployment work we do, using disciplined hoshin and X-matrix techniques, is built for exactly this challenge. It turns ambition into priorities and priorities into operational outcomes.

Follow the conversation

This blog is one of our numerous pieces of thought leadership and opinion in operational transformation and AI in the Public Sector. To receive future pieces directly, follow Reinvigoration on LinkedIn or subscribe via the website.

Closing

The day at Strathclyde did not really expose a new problem. It exposed the same operational complexity problem that has held back every significant technology investment of the last twenty years, dressed in AI clothing. Strategy is published but does not cascade. Operating models are not designed for AI-augmented work. Processes are broken and the AI sits on top of them. Continuous improvement discipline is absent when things wobble.

Simplify4Scale® was built precisely for this kind of problem. The prize is on the table for the UK public sector. The question is whether public sector leaders are prepared to do the operational work that releases it. We are ready to help when they are.

Sources and further reading

• Storm ID. Automate Tasks, Not Jobs. February 2026. stormid.com/research
• Scottish Government. Scotland’s AI Strategy 2026 to 2031. March 2026. gov.scot/publications/scotlands-ai-strategy-2026-2031
• National Audit Office. Use of artificial intelligence in government. March 2024. nao.org.uk/reports/use-of-artificial-intelligence-in-government
• Tony Blair Institute for Global Change and Faculty AI. Governing in the Age of AI: A New Model to Transform the State. 2024. institute.global
• Tony Blair Institute for Global Change. Governing in the Age of AI: Reimagining Local Government. May 2025. institute.global
• Tony Blair Institute for Global Change. Governing in the Age of AI: Building Britain’s National Data Library. February 2025. institute.global
• UK Government. AI Opportunities Action Plan. January 2025. gov.uk
• FutureScot. Public Sector AI 2026 conference, University of Strathclyde, 21 May 2026. futurescot.com