Mario Thomas

Capability, Not Credentials: What the Law and the Moment Now Demand

The Data (Use and Access) Act 2025 has been in force since 5 February 2026, and it gives any person subject to a significant decision taken solely by automated processing the right to contest it and to demand human review. On the page, that is a set of procedural safeguards. In practice, it is a capability requirement that most Boards have not yet recognised they have taken on.

May's articles trace that requirement through five settings. The Reasoning Gap sets out the bind directly: the law's safeguards depend on whether an organisation can interrogate its own probabilistic decisions, a capability rule-based systems possess and most approved AI systems do not. The Headroom Argument reframes the relentless flow of efficiency news as a capability signal rather than a budget saving. Ethical AI confronts the values a foundation model imports without the Board's consent. AI and the Company Secretary follows the constitutional principle of collective responsibility into the operational decisions where it silently erodes. And Ontologies and Knowledge Graphs locates the structural foundation that makes machine reasoning defensible in the first place.

The thread running through all five is that capability, not credentials, is now the binding constraint. A Board can hold the right tools, run the right awareness sessions, and still be unable to interrogate the decisions its systems make, author the values they apply, or reconstruct how a particular conclusion was reached. Tool adoption is not transformation. The gap between what an organisation deploys and what it can actually govern is where the next round of competitive, and legal, separation will form.

If your time is limited, I particularly recommend The Reasoning Gap. It is the piece that makes the others urgent: the first contestability request is the moment the absence of capability becomes visible, and by then the cost of closing the gap is several orders of magnitude higher than the cost of building it now.

How is your organisation ensuring it can interrogate, explain, and stand behind the decisions its AI systems now make on its behalf?

- Mario

The Reasoning Gap: The Capability the Law Now Demands of Boards

Published 3 May 2026 | 11 minute read

The UK regime now requires four safeguards for any significant decision taken solely by automated processing: information, representations, human intervention, and contestability. On the page these are procedural rights, but they all depend on something the law does not name — whether the organisation can interrogate its own decisions well enough for the safeguards to work. For a rule-based system that capability is built in; for a probabilistic system it is not, and most Boards have approved those systems without ever asking whether it exists.

The Headroom Argument: Why AI Efficiency Means More Compute, Not Less

Published 10 May 2026 | 9 minute read

The instinctive reading of relentless efficiency gains is that the AI budget can shrink and that hyperscaler capex will prove over-sized. That reading is the wrong way around. This article examines why architectural efficiency releases demand rather than reducing it, what every prior era of computing tells us about where the inference market is heading, and how Boards should read efficiency news as a capability signal rather than a cost saving.

Ethical AI: When the Model Imposes Values Your Organisation Did Not Choose

Published 17 May 2026 | 14 minute read

A foundation model arrives carrying a value system its provider built: what it refuses, how it frames a sensitive subject, how it resolves a question with reasonable views on either side. That standard, not the organisation's, is the one in force, and it changes with each model version without the Board's consent. This article sets out how a Board chooses, deployment by deployment, whether to accept the provider's ethics, reject the deployments where it bears on people, or build alignment the organisation owns.

AI and the Company Secretary: Operating the Boundary the Chair Polices

Published 24 May 2026 | 14 minute read

The company secretary's role was built to maintain the conditions under which directors can apply judgement and the company can meet its governance obligations. Both are now being remade — by AI inside board administration that composes the materials directors judge, and by AI inside the business that shapes the compliance position the secretary must disclose. Drawing on Cadbury, the Companies Act, the FRC's 2024 Code, and the Chartered Governance Institute, this article shows how the chair polices the boundary between agency transfer and accountability transfer, and the secretary operates it in practice.

Ontologies and Knowledge Graphs: Why Structure is the Next Data Frontier

Published 31 May 2026 | 15 minute read

The strategic value of data is no longer in question; the next frontier is data whose meaning and relationships are explicit enough for machines to reason over rather than merely retrieve. Quality tells the organisation whether information is reliable; structure tells the machine what that information means. This article argues that ontologies and the knowledge graphs built upon them have moved from technical infrastructure into Board territory, because they increasingly determine what an organisation officially knows and where durable advantage is created.

Copyright © 2026 Mario Thomas. All rights reserved.

You are receiving this email because you signed up to receive it on mariothomas.com on {{date_subscribed}} at {{time_subscribed}}. You can unsubscribe at any time.