AI and the future of governance – part one, the role of the machines

In the first of a two-part article, Daniel Taylor looks at the power of artificial intelligence to enhance governance practice

“There are more things in heaven and earth, Horatio,

Than are dreamt of in your philosophy.”

William Shakespeare: Hamlet

Governance has always been an odd discipline. It sits somewhere between law and philosophy, between the visible artefacts of compliance and the invisible forces of power, culture, and judgement. It borrows the language of control, assurance, and accountability, yet its real work is about outcomes: whether institutions behave wisely, ethically, and in the long-term public interest.

Into this already ambiguous space comes artificial intelligence, accompanied by the usual mix of excitement, anxiety, and hyperbole. Depending on whom you ask, AI will either automate governance reviews into irrelevance or finally rescue a chronically under-resourced public sector from regulatory overload. The truth is more interesting and more demanding than either extreme.

What AI offers governance is not replacement, but amplification. Used well, it can dramatically accelerate and deepen the technical dimensions of governance assessment: mapping compliance, testing alignment, surfacing gaps, and comparing practice against fast-moving regulatory and sector standards. Used badly, it will produce a comforting illusion of assurance while missing precisely the things that matter most.

The idea of perfection

Governance has always been an exercise in managing imperfection: imperfect people, information, structures and incentives. The idea of a perfect governance system has always been a theoretical ideal, not an achievable endpoint. But AI challenges that assumption by offering tools that could continuously diagnose weaknesses, model consequences, and propose near optimal structures at a speed and scale human committees cannot hope to match.

What this really changes is visibility. AI can make the compliance position of an organisation far more legible; instantly discovering not just whether required documents exist, but how they connect, whether they align, and where they contradict each other.

This matters because good governance impact is rarely achieved through mechanics alone. It emerges from the interaction between formal structures and informal dynamics, between what boards intend and what actually happens.

At its best, governance is not reactive but anticipatory. AI opens up the possibility of more proactive governance: identifying emerging risks earlier, stress-testing arrangements before they fail, and supporting continuous improvement rather than episodic review. As observed in related work on intelligent systems, the real opportunity lies not in one-off diagnostics, but in creating feedback loops that allow governance arrangements to be tested, refined, and improved as standards evolve and contexts change. Governance becomes less about periodic inspection and more about learning.

Speed, scale, and the compliance floor

Governance reviews are labour-intensive. They involve trawling through sprawling policy libraries, committee papers, schemes of delegation, regulatory correspondence, and assurance reports. In many public purpose organisations, these documentary universes have expanded rapidly in response to increasingly interventionist regulatory regimes. The challenge is that many organisations need more rapid assessments, maintaining the rigour and quality of analysis.

AI is exceptionally good at this kind of work. It can ingest large volumes of documentation, cross-reference them against statutory duties, regulatory conditions, and sector codes, and identify inconsistencies, omissions, and outdated provisions in minutes rather than weeks. It can test whether policies are internally coherent, whether approval routes align with constitutional documents, and whether required elements are present, missing, or contradicted elsewhere.

This matters because compliance remains the necessary floor of governance. Control is not governance, as the King codes rightly remind us, but without it, governance collapses into wishful thinking. AI can help organisations get this floor right more quickly and more consistently than traditional methods allow.

Higher education as a governance stress test

UK higher education provides a sharp illustration of both the opportunity and the limits of AI in governance work. Over recent years, universities have had to respond simultaneously to Care Quality Commission activity in areas of student support and wellbeing, amendments to the Office for Students’ conditions of registration, and the Higher Education (Freedom of Speech) Act 2023 with its new statutory duties, complaints routes, and enforcement powers. The freedom of speech regime in particular has exposed structural weaknesses in governance frameworks across the sector.

The Act requires universities and students’ unions to maintain detailed, legally compliant codes covering academic freedom, external speakers, protest management, complaints handling, and escalation routes. For many institutions this has meant wholesale restructuring of policy libraries, approval processes, training regimes, and audit trails. Governance, legal, and compliance teams have faced a sharp increase in workload, often without a commensurate increase in capacity.

The enforcement risk is no longer theoretical. The Office for Students’ £585,000 fine (originally set at £1 million) imposed on the University of Sussex sent a clear signal that freedom of speech compliance is now a material financial and reputational risk. Unsurprisingly, this has driven a more defensive posture across the sector. One vice-chancellor remarked to me, only half-jokingly, “I never know when I might wake up and be on the front page of the Daily Mail.”

Behind the humour lies a serious point. Compliance officers and academic leaders are being asked to balance competing legal duties around free speech, equality, safeguarding, and health and safety without clear bright-line tests. Anxiety about regulatory findings and personal accountability has become a feature of the system rather than an exception.

How AI can help

This is precisely the kind of environment in which AI, deployed intelligently and ethically, can add real value.

Used by an expert governance consultant, AI can rapidly assess whether a university’s freedom of speech policies fully reflect statutory requirements and regulatory guidance, are consistent with institutional statutes and schemes of delegation, align with related policies on equality, disciplinary processes, and complaints, and contain clear operational routes from principle to practice.

It can highlight where language is ambiguous, where obligations are implied but not operationalised, and where approval or review cycles are misaligned with regulatory expectations. It can benchmark policies against emerging sector norms and flag where institutions are drifting out of step.

Crucially, this can be done quickly enough to support live decision-making rather than post hoc reassurance. Speed matters when regulatory interpretation evolves faster than governance cycles.

The NHS: scale, scrutiny, and systemic risk

If higher education is a stress test for governance capability, the NHS is a stress test at scale.

Few sectors combine such vast organisational complexity with such intense regulatory scrutiny. integrated care boards, NHS trusts, and foundation trusts operate within a dense web of statutory duties, national policy, regulatory oversight, and political expectation. The governance burden is immense: quality and safety regulation through the Care Quality Commission, financial oversight and intervention powers from NHS England, explicit expectations around population health and inequalities, and a constant cycle of national priorities and guidance.

In this environment governance teams are stretched thin. Boards face hundreds of pages of papers each month, assurance frameworks proliferate, and policies multiply in response to inspections, inquiries, and high-profile failures. As in higher education, compliance activity expands faster than organisational capacity to absorb it.

AI has clear potential to support this system. Deployed well, it could rapidly analyse board assurance frameworks, committee terms of reference, risk registers, and policy suites against regulatory requirements and recognised good practice. It could test alignment between quality governance, financial governance, and workforce oversight, and identify where assurance is duplicated, fragmented, or missing altogether.

For organisations under regulatory pressure, speed matters. AI-enabled assessment could allow experienced consultants to form a system-level view of governance maturity far more quickly than traditional review methods permit, supporting earlier intervention and more targeted improvement.

AI can’t help with dysfunctional dynamics

Yet the NHS also illustrates the hard limits of automation. Many of the most serious governance failures of recent years have not arisen from missing policies or poorly drafted terms of reference, but from dysfunctional dynamics: boards overwhelmed by operational detail, blurred accountability between organisations and systems, cultures that suppress challenge, and a persistent gap between assurance on paper and reality on wards.

No algorithm can judge whether a board truly understands the risks it is carrying on behalf of patients and the public. No system can tell when reassurance has quietly displaced scrutiny. These remain profoundly human judgements.

In the second part of this article, I will dig a little deeper into the need to strike a balance between AI analysis and human judgement, drawing on Judge Mervyn King’s insistence that good governance is not a mechanical compliance exercise, but an ethical and strategic practice that only experienced practitioners are properly equipped to carry out.