Founded in 2026, AI CLEAR is an independent UK firm helping public-sector boards govern artificial intelligence with confidence, competence and accountability. The work is anchored in CLEAR™, a governance maturity framework developed for boards.
Boards across the UK public sector face the same question: can they direct and account for AI deployment in their institutions? The evidence is that most cannot, yet. The gap between recognising the problem and being equipped to manage it is the defining governance issue of 2026.
The pressure is hardening. The UK Corporate Governance Code Provision 29 set a board-level AI accountability standard for listed companies from January 2026. Public bodies face the same expectations, increasingly reinforced through the Cabinet Office Algorithmic Transparency Recording Standard and sector-specific guidance. Boards are accountable now, with or without the tools.
AI CLEAR is working first with corporation boards in UK Further Education, where the governance gap is well evidenced. Most colleges have no AI literacy training at board level, no AI on the risk register, and no AI-specific legal advice on retainer; seventy-one per cent of FE leaders cite AI as their number one challenge for the second consecutive year (Jisc 2025).
Most accounts of AI governance focus on tools, models and technical controls. CLEAR™ focuses on whether the governing body has the capability to direct, scrutinise and hold AI deployment to account. The thesis behind the framework is that governance maturity, not technical sophistication, determines whether AI adoption serves the people the institution exists to serve.
AI CLEAR Ltd owns and commercialises the CLEAR™ AI governance framework, developed by Stuart Rimmer MBE.
The operating environment, regulatory landscape, and AI deployment surface in which the institution makes AI decisions.
Strategic intent, culture, and where governance responsibility for AI sits at board level.
Operational deployment, controls, tooling, and the implementation discipline that translates governance into practice.
Oversight, transparency, audit trail, and ethical assurance — including who is responsible when an AI system underperforms or causes harm.
Measurement of impact and outcomes, with feedback that drives continuous improvement of governance over time.
Twenty years' C-suite leadership across UK regional economic development and further education. Sector-recognised institutional reformer and most recently Interim Chief Executive of Thames Freeport. MBE 2021; TES FE Leader of the Year. Lead author of the forthcoming Ready or Not: AI Leadership Readiness in UK Further Education. Founder of Civentis Advisory.
Former Senior Civil Servant with two decades at the centre of UK industrial and trade policy. Senior leadership roles at the Department for Business and Trade, UK Overseas Network, and Thames Freeport where he led investment work, including the Freeport's AI Growth Zones bid. Most recently Strategy and Public Affairs Director at an AI venture.
AI CLEAR draws on a small bench of senior practitioners and academics. Our experts contribute scholarship, sector insight and senior practitioner experience to the work of the firm.
Dame Sally Dicketts is an experienced chief executive, chair and non-executive director, with nearly thirty years leading further education colleges and developing people and organisational culture. She now chairs three boards spanning further education, family learning and health charities, and holds governance roles at Royal Holloway, University of London, as a trustee, and the University of Bradford, as Pro-Chancellor and senior independent member. She is senior independent director of the Sport and Recreation Alliance and an executive coach specialising in leadership, strategy, governance and change. She was appointed CBE in 2013 and DBE in 2023 for services to education.
Mark Fagan is a Lecturer in Public Policy at the Harvard Kennedy School, where he teaches operations management, systems thinking, supply chain management and policy development, and lectures on service delivery and artificial intelligence in executive education. He leads the School's Autonomous Vehicles Policy Initiative and advises organisations on adopting AI to improve operations. He has advised public and private sector management on strategy for more than thirty years and was a founding partner of Norbridge, Inc. He is the author of Supply Chain Management: A Public Sector Perspective (Elgar, 2024) and Governing With AI (Bloomsbury, 2026).
Orla Cooper is a Chartered Company Secretary and Chartered Governance Professional with over twenty-five years' experience in corporate governance and regulatory compliance. She spent fifteen years with DCC plc, a FTSE 100 company headquartered in Dublin, supporting the board, its committees and subsidiaries in meeting complex regulatory requirements and embedding good governance practice. Earlier in her career she held managerial roles at Deloitte, JPMorgan, Mazars and EY, specialising in regulatory compliance and corporate law advisory.
A selection of work informing the practice of AI CLEAR and the development of the CLEAR™ framework.
Rethinking what AI readiness means for public institutions. Argues for a shift from compliance-oriented frameworks to genuine institutional sovereignty — the capacity to understand, direct and govern AI adoption on an organisation’s own terms.
“The institutions that will lead — that will adopt AI responsibly, govern it intelligently, and adapt as it evolves — will not be those that have achieved the highest compliance scores. They will be those whose leaders have done the deeper work of understanding what they are doing and why.”
There is a distinction that the current discourse around AI readiness has largely failed to draw, one that, in its absence, is quietly distorting how public institutions are approaching the most consequential technological shift of our era. It is the distinction between compliance and sovereignty.
An organisation that is AI-compliant has done what is required of it. It has reviewed its data protection obligations under the UK GDPR, considered the guidance issued by its sector regulator, and perhaps completed a self-assessment against ISO/IEC 42001. It can, if asked, produce documentation. In a world of tick-box governance, this passes for readiness.
An organisation that is AI-sovereign is doing something categorically different. It understands its own values well enough to know which uses of AI are consistent with them and which are not. It has developed the institutional confidence to say no to a compelling technology solution when it conflicts with its ethical commitments, and the confidence to say yes, with appropriate governance, when the conditions are right. It does not wait to be told what readiness looks like.
The policy frameworks currently shaping AI adoption in the public sector, valuable as many of them are, are configured for compliance, not sovereignty. This paper argues that this configuration is a strategic error.
The language of AI readiness, as it has developed in UK policy discourse, is the language of compliance. The AI Opportunities Action Plan speaks of assurance ecosystems and adoption hubs. ISO/IEC 42001 offers a management system framework. The Data (Use and Access) Act 2025 relaxes certain data protection obligations while maintaining guardrails for the riskiest use cases. The direction of travel is clear: establish the framework, communicate the expectations, measure compliance.
This is not wrong. But it is insufficient. The problem with compliance-oriented readiness is not that it asks too much of institutions. It is that it asks too little. It asks for procedures where it should be asking for culture. It asks for documentation where it should be asking for understanding.
Schein’s foundational work on organisational culture identifies three levels at which culture operates: visible artefacts (policies, procedures); espoused values (what the organisation says it believes); and underlying assumptions (the taken-for-granted beliefs that actually drive behaviour). Compliance frameworks operate primarily at the first level. Only leadership — sustained, culturally embedded, and intellectually serious — can reach the underlying assumptions.
The GoLLM and Heriot-Watt University AI Readiness of UK Local Authorities report (2026) maps AI maturity across 280 local authorities. The institutions demonstrating the most robust governance practices are not, in the main, those that have invested most heavily in compliance infrastructure. They are those whose senior leadership teams have developed genuine conceptual literacy about AI governance.
Argyris and Schon’s distinction between single-loop learning — correcting errors within existing frameworks — and double-loop learning — questioning the frameworks themselves — maps precisely onto the distinction between compliance and sovereignty. Only sovereignty enables double-loop learning: the institutional reflection that allows an organisation to interrogate its own assumptions about what AI is for.
AI sovereignty would look like an institution that has done the deeper work of understanding itself — its values, its risk appetite, its data landscape, and the specific ways in which AI systems might serve or subvert its obligations to the communities it serves.
It would look like a board that has developed genuine literacy about AI governance — not the ability to explain a neural network, but the ability to ask the right questions of the executives presenting AI propositions, to identify the gaps in those propositions, and to hold the organisation to account.
It would look, above all, like an institution that has developed what the CLEAR™ Framework calls a ‘leadership identity’ in relation to AI: a settled understanding of the kind of AI-using organisation it wants to be, robust enough to guide decision-making in conditions of uncertainty.
External frameworks should be treated as tools, not as ends in themselves. ISO/IEC 42001 is a valuable starting point but cannot substitute for the leadership conversations that determine what an AI management system is actually for.
The investment that UK public institutions most urgently need to make is not in compliance infrastructure. It is in the kind of leadership development that builds genuine conceptual literacy about AI, genuine ethical seriousness about its implications, and genuine institutional confidence in navigating its complexities.
The institutions that will lead will not be those that have achieved the highest compliance scores. They will be those whose leaders have done the deeper work of understanding what they are doing and why. They will be, in the precise sense I have been using, sovereign.
Selected References
Why place-based AI infrastructure is the overlooked layer in Britain’s innovation strategy. Examines the structural gap between national AI ambition and the regional and sub-regional governance capacity required to make adoption real.
“Britain's AI economy will not be built from the centre down. The places that lead will be those that build sovereign AI leadership capacity at the regional and institutional level — before, and not only because, the policy architecture catches up with them.”
Britain’s AI ambition is expressed, in the main, in national terms. The AI Opportunities Action Plan speaks of computational infrastructure, sovereign capability, and international positioning. The DSIT’s January 2026 progress report records a sixfold increase in supercomputer capacity, six billion pounds in domestic AI venture capital, and the ambition to equip ten million workers with AI skills by 2030.
These are not unimpressive figures. But they describe a strategy with a significant structural gap. The gap is not at the top — the national policy architecture is reasonably well developed. Nor is it at the bottom — individual organisations are increasingly engaged with AI. The gap is in the middle: at the regional and sub-regional level where institutional infrastructure for AI adoption either exists and functions, or does not.
The European Commission’s 2025 Regional Innovation Scoreboard identifies a structural asymmetry at the heart of the UK’s regional innovation challenge. Mayoral Combined Authorities have negotiated powers over local economic development, transport, and skills. But industrial and research and development policy remains largely centralised, delivered on terms set in Whitehall.
This creates what might be called the devolution paradox: Mayoral Combined Authorities have sufficient responsibility to be accountable for regional economic outcomes, but not sufficient power to shape the innovation ecosystems that determine those outcomes. They are responsible for delivery without being equipped for design.
The Devolution Priority Programme offers a genuine opportunity to reimagine how subnational governance operates. But it carries a significant risk: that during the transition, the governance capacity required to make devolved AI decision-making legitimate and effective is not built at pace with the transfer of responsibility.
Thames Freeport offers one instructive illustration of how place-based economic governance can provide a more enabling institutional environment. The AI Institute and Freeport Innovation and Venture Studio model being developed in the Thames Estuary corridor represents precisely the kind of place-based institution-building that the national strategy needs at the meso level: a vehicle for connecting capital, knowledge, and governance capacity in a specific place.
Filling the missing middle requires investment in three areas that the current national strategy underweights.
Britain’s AI economy will not be built from the centre down. The places that lead will be those that build sovereign AI leadership capacity at the regional and institutional level. That requires investment not only in compute and capital, but in the institutional infrastructure and leadership capacity that make adoption sustainable.
Selected References
Why the UK’s AI governance gap is a leadership problem, not a legislative one. Analyses the five cross-sector AI principles and the conditions under which they can or cannot translate from policy aspiration into institutional practice.
“Governance frameworks are only as effective as the leaders charged with implementing them. Until we close the leadership readiness gap, no bill, however comprehensive, will take hold.”
In January 2026, the Department for Science, Innovation and Technology published its AI Opportunities Action Plan: One Year On, reporting that 38 of 50 key commitments had been met. By conventional metrics of policy delivery, the picture looked encouraging.
And yet something essential is missing. The UK’s AI governance architecture has been built on five cross-sector, non-statutory principles: safety, security and robustness; transparency and explainability; fairness; accountability; and contestability and redress. They are intellectually coherent. They are also, in their current form, largely inert — because the leaders responsible for applying them, in hospitals, universities, colleges, local authorities, and enterprise boardrooms, are not equipped to do so.
This is the governance gap that no forthcoming AI Bill will resolve. It is not a legislative problem. It is a leadership problem.
The UK’s pro-innovation, principles-based approach to AI regulation emerged from the 2023 White Paper. Where the European Union opted for the legislative weight of the AI Act, the UK chose to work with existing regulators and voluntary standards. Flexibility without institutional capacity, however, is simply ambiguity. And in conditions of ambiguity, it is leaders — not frameworks — who determine whether principles become practice.
The February 2026 AI Readiness of UK Local Authorities report finds consistent patterns of governance infrastructure deficit: authorities deploying AI tools without the accountability frameworks, bias monitoring processes, or board literacy required to govern those deployments responsibly. This is not a finding about technology. It is a finding about leadership readiness.
Consider what it would actually mean for a senior leader in a further education college, or a director of adult social care, to operationalise the UK’s five AI principles in their day-to-day decision-making.
These are the questions that arise the moment a leader moves from reading a policy document to making an actual procurement decision. The principles must become capacities. Capacities cannot be legislated into existence.
The anticipated UK AI Bill will introduce accountability mechanisms for frontier models and more rigorous testing requirements for high-risk systems. These are genuine advances. But legislation can mandate transparency. It cannot manufacture the leadership capacity to interpret what transparency means in a given institutional context.
The UK has produced a governance framework of genuine sophistication. What it has not done, at anything like the required scale, is invest in the leadership layer that sits between policy aspiration and institutional practice. That is where governance actually happens — where the UK’s five AI principles will either translate into practice or remain what they currently are: principles without power.
Selected References
AI governance, human judgement, and the limits of the machine. Examines automation bias in senior leadership, the conditions under which human oversight becomes performative, and the governance design that meaningful oversight actually requires.
“The form of leadership persists. The substance has been quietly evacuated. The synthetic leader does not know they have become synthetic. They continue to read and sign the documents, chair the meetings, and take the decisions — but the decisions are increasingly shaped by analytical outputs they have not interrogated.”
There is a failure mode in AI governance that the current regulatory discourse has not yet learned to name. It is not the rogue algorithm making decisions without human oversight. It is the emergence of what I want to call the synthetic leader: a human being occupying a position of authority who has, through the habitual consumption of AI-generated outputs and the gradual surrender of independent analytical effort, become something less than the fully reasoning, independently judging agent that governance requires.
This is witnessed in offices across the country. In a recent project involving a Big Four consultancy, the deliverable was clearly originated in a large language model and the subsequent recommendations were largely untouched by human leadership — the exchange between client and consultant had itself become, in substance, an exchange between language models. This is a predictable consequence of the trajectory of AI adoption in institutional contexts.
The scientific literature on automation bias is now extensive. Parasuraman and Riley’s foundational 1997 paper defined automation bias as the tendency to over-rely on automated systems and to under-weight one’s own judgement when automated recommendations are available. Subsequent research in aviation, healthcare, military command, and financial decision-making has confirmed the basic finding: when automated systems appear authoritative, human decision-makers systematically reduce their independent analytical effort.
Kahneman’s work on System 1 and System 2 thinking provides a useful theoretical frame. The effortful, deliberate reasoning of System 2 is metabolically costly, and human beings will default to System 1 whenever conditions permit. The presence of a sophisticated AI system that has produced a confident, well-formatted, internally consistent recommendation is precisely the kind of condition that permits it.
A governance process that requires a human to review and approve AI outputs before they take effect is not, in itself, meaningful oversight if the human doing the reviewing has neither the conceptual grounding to evaluate what they are reviewing nor the institutional standing to challenge what they do not understand. The form of oversight is present. The substance is absent.
Meaningful human oversight is a leadership capacity, not a procedural requirement. It must be developed.
The CLEAR™ Framework’s Synthetic Leadership Matrix maps two dimensions: the degree to which leaders actively direct the AI systems they use — shaping prompts, challenging outputs, interrogating assumptions — and the degree to which they retain independent analytical capacity in domains where AI is most relied upon.
| HIGH AI DIRECTION | LOW AI DIRECTION | |
|---|---|---|
| HIGH INDEPENDENT JUDGEMENT | Critical Collaborator — Uses AI as a genuine tool. Challenges outputs routinely. The governance ideal. | Reflective Sceptic — Cautious adoption, retaining the independent reasoning capacity governance requires. |
| LOW INDEPENDENT JUDGEMENT | Active Director — Engaging with AI deeply but at risk of domain-specific over-reliance. | Synthetic Leader — Neither directing the AI systems used nor maintaining independent analytical capacity. The governance risk this paper names. |
At the individual level: sustained development of AI literacy, strategic orientation, governance judgement, and ethical conviction — the capacity the CLEAR™ Framework’s Individual Sensemaking Model describes. This requires protected space for deliberate reasoning, explicit practice in challenging AI outputs, and cultures of psychological safety.
At the institutional level: boardrooms and executive teams need new governance rituals around AI — structured moments of collective deliberation explicitly designed to preserve rather than shortcut independent judgement.
The governance frameworks we have built assume human leaders who are genuinely in charge. The synthetic leader is the figure who occupies that governance role without, in the substantive sense, fulfilling it. Naming this failure mode is the first step towards addressing it. Addressing it requires sustained investment in the leadership development that governance, in the age of AI, actually demands.
Selected References
An examination of how the public sector can adopt artificial intelligence to improve service delivery, drawing on Mark’s teaching and research at the Harvard Kennedy School. Elements of CLEAR™ build on this scholarship on the efficiency, quality and equity dimensions of public-sector AI.
Find the book →A study of supply chain management in public-sector settings, with implications for procurement, service delivery and operational oversight relevant to the institutional context in which AI is deployed.
Find the book →A working paper setting out how organisations can make the assessment of risk explicit in decisions to adopt and deploy AI, complementing the governance questions the CLEAR™ framework asks of boards.
Read the paper →We work with senior leaders and governing bodies across the UK public sector on AI governance. Our first engagements are with corporation boards in UK Further Education. To start a conversation, email us.
hello@aiclear.uk