Blog

Thinking in progress

Essays and daily insight on Responsible AI, governance and the quantum era — written between flights, boardrooms and policy drafts.

Loading the latest articles…

March 2026
6 min

Compliance Is Not a Strategy

Regulation is rarely the thing that kills a company; the posture a company adopts towards regulation usually is. Organisations that read the EU AI Act as a documentation exercise will assemble the binders, pass the audit and discover — too late — that the market has quietly repriced trust as a product attribute rather than a legal formality. The Act's real content is architectural: it asks who is accountable for a decision, what evidence exists that the system behaves as claimed, and how quickly harm can be detected and reversed. Firms that answer those questions in their engineering and operating models rather than in their legal appendices gain something compliance cannot buy — the ability to deploy consequential AI at speed, because the controls are already load-bearing. Compliance is the floor. Strategy is what you build on top of it.

February 2026
5 min

The Boardroom AI Literacy Gap

Boards have spent a century learning to interrogate money. They can read a balance sheet under pressure, sense when a forecast is being flattered, and ask the one question that unravels a rehearsed answer. Very few can do the same to a model. That asymmetry has now crossed from awkward to fiduciary: directors are accountable for systems whose failure modes they cannot articulate, briefed by executives who are themselves relaying vendor language. Closing the gap does not require directors to become engineers. It requires a small, durable vocabulary — provenance, drift, human oversight, contestability — and the confidence to keep asking until the answer stops being a slogan. The boards that acquire this literacy will not merely avoid scandal; they will approve better technology faster, because they will finally understand what they are approving.

January 2026
7 min

Who Builds the Future Decides Whose Future It Is

Representation in artificial intelligence is routinely filed under culture, which is precisely why it keeps failing. Treat it instead as a risk control and the argument becomes uncomfortable rather than aspirational: homogeneous teams share blind spots, shared blind spots become untested assumptions, and untested assumptions become production defects that land on the people least able to challenge them. Every well-documented AI failure of the past decade — misread faces, mispriced risk, misallocated care — was legible in advance to somebody who was not in the room. Diversity, in this framing, is not an act of generosity towards underrepresented groups; it is a cheap and reliable mechanism for surfacing failure modes before deployment. Who builds the future decides whose future it is, and governance frameworks that ignore this are governing on paper only.

December 2025
4 min

The Quiet Power of Documentation

Documentation is the least glamorous requirement in AI regulation and the single best predictor of whether an organisation will survive scrutiny. Not because regulators enjoy paperwork, but because a system whose data lineage, design choices, evaluation results and known limitations can be reconstructed on demand is a system somebody actually understands. The act of writing it down changes the engineering: ambiguities that were tolerable in conversation become intolerable on the page, and decisions that were never really made become visible as gaps. Companies that build this discipline early treat an incident as an investigation; companies that skip it treat an incident as an archaeology project conducted in public. In the AI era, the ability to explain yourself precisely, quickly and in writing is not bureaucracy — it is the durable form of institutional memory.

Enquire

Bring These Ideas to Your Stage

Every essay here has a keynote behind it.

Book Dr. Neda Enquiry Form