When intelligence becomes abundant: what AI means for governance
10 August 2026
Rupert Wiseman, managing director of AssumeAI, looks beyond the hype to assess the real value of artificial intelligence – if it’s managed properly
Points raised in this article:
- AI democratises intelligence, extending capability without replacing human expertise
- Effective adoption requires structure, strategy, verification and trusted foundations
- Governance must address bias, source integrity and responsible deployment
- AI elevates human judgement, accountability and trust in governance.
A few months ago, an OpenAI model disproved the prevailing conjecture on the planar unit distance problem: a question first posed by Hungarian mathematician Paul Erdős that had stood for nearly 80 years.
This was a small item in the technology press, but it tells us something that most of the coverage of artificial intelligence misses. This technology does not simply repeat back what we have already learnt; it can actively extend our understanding of the world.
That distinction matters for anyone responsible for governance. If a model can push at the edge of 80-year-old mathematics, the relevant question is no longer whether AI matters, but how an organisation adopts it responsibly and what, in the end, it changes about the work of holding an organisation to account.
I want to set out how I have come to think about that, because I believe AI is a force for good, and because the way through it is less dramatic – and more deliberate – than the headlines suggest.
Democratised intelligence
The most encouraging pattern I have seen is the democratisation of intelligence. Capability that used to sit only with those holding deep subject matter expertise is now within reach of almost anyone. In software, this shows up as work being produced through natural language far faster than was ever previously possible. In the organisations we work alongside, and in the governance teams that serve their boards, it shows up in the research and drafting that underpins good assurance. The heavy lifting of summarising, synthesising and preparing no longer scales in a straight line with headcount. A smaller team gains a much longer reach. That is not a story about replacing people. It is a story about freeing them for the work that only they can do.
Neither magic nor a fad
The most encouraging pattern I have seen is the democratisation of intelligence. Capability that used to sit only with those holding deep subject matter expertise is now within reach of almost anyone. In software, this shows up as work being produced through natural language far faster than was ever previously possible. In the organisations we work alongside, and in the governance teams that serve their boards, it shows up in the research and drafting that underpins good assurance.
The heavy lifting of summarising, synthesising and preparing no longer scales in a straight line with headcount. A smaller team gains a much longer reach. That is not a story about replacing people. It is a story about freeing them for the work that only they can do.
It would be easy, at this point, to overclaim, and the surrounding noise makes that temptation worse. In one breath the media tells you AI will change your life overnight. In the next, that it is overhyped and has no real substance behind it. Both extremes are wrong, and the polarisation itself creates a genuine public misunderstanding. The truth sits in the middle. AI is neither magic nor a fad. Like any significant technology, it is a lever, multiplying the output you can draw from a given input.
The original iPhone is the analogy I keep returning to. The watershed moment has happened, but the value does not arrive on its own. It has to be pursued in a structured way: identifying and prioritising the real opportunities, developing and deploying the right solutions, and then embedding and adopting them into how the organisation actually works. Skip that structure and you get precisely the disappointing outcome the sceptics predicted, which then feeds the very cynicism that held you back.
The role of governance
Doing this properly requires building on strong foundations, and this is where those responsible for governance have an unusually important part to play. In the short term, the dependency that matters most is the ability to verify a model's output efficiently. If checking for errors takes longer than the model saved, the gain simply inverts.
Over the longer term, the deeper risk is bias in the data these models are trained on. As more of what they consume becomes itself machine-generated, we have to be able to detect and mitigate those biases rather than quietly perpetuate them.
Jeremy Utley at Stanford writes usefully about the cognitive dissonance of working with these tools; the concern that sits beneath that, for me, is the integrity of the source itself. Dissonance can be managed. A compromised source of truth is far harder to unwind.
None of this is a reason to wait. It is a reason to build on solid ground, with clear policy, sensible habits of verification, and a considered view of where information comes from. That, after all, is what good governance has always insisted upon.
It is the same instinct that sits behind a tool we’re currently developing with GGi: Ask GGi. The approach is what is known as retrieval-augmented generation; rather than relying on a model's general training, it grounds every answer in a trusted, curated body of knowledge. In this case that means a knowledge base built up since 2009 now being integrated with a wider database of third-party institutions, so that in time it becomes the natural place to bring a governance question and receive a considered, well-sourced answer.
It is worth being honest about what a tool of this kind can and cannot do. It can retrieve, summarise and point to the right guidance in seconds, saving hours of searching and drafting. It cannot take accountability, exercise professional judgement, or sit in a boardroom on anyone's behalf. It is there to make good people faster and better informed, not to stand in for them, and I think it is a good example of a respected institution building on its own foundations rather than bolting AI on for its own sake.
The wider evidence points the same way. The Consultancy Growth Network's 2026 BenchPress research found that although most organisations now use AI in some form, only 15 per cent have fully deployed it across their operations, and that group is more than twice as likely to be a high performer on both growth and profitability. Those still stuck in trialling are the least likely to have grown at all. Where adoption is genuine, the effects are concrete: 74 per cent report increased efficiency and 56 per cent a reduced cost of delivery. Yet fewer than half have gone as far as setting an actual strategy. The distance between trying AI and truly deploying it, on purpose and on firm foundations, is where the real value sits.
Human intelligence
Which brings me to the point this has all been leading towards. If intelligence becomes abundant, the question worth asking is what remains scarce, and therefore valuable. The answer, I think, is the human part of the work.
As the routine, the administrative and the merely procedural are steadily commoditised, what cannot be automated is judgement, accountability and the trust that comes from being relied upon in a room. Those are not incidental to governance. They are the whole of it. Far from diminishing the role, abundant intelligence sharpens its value, because it clears away everything surrounding the judgement and leaves the judgement itself exposed.
That, in the end, is why I regard AI as a force for good. Used well, and on the right foundations, it does not replace the human work of governance. It returns us to it.