How Boards Can Judge AI Readiness Without Self-Report

The wrong question boards keep asking

In "Why Governance Must Come Before AI in Leadership Decisions", I argued that boards should treat governance as a condition of deployment rather than a consequence of it. That was the system question. This article asks the human one. It is the last in the series, and it should be the most honest about limits.

Are our leaders AI-ready? That is the question boards keep asking, and the phrasing sounds modern. In reality, it invites the weakest form of evidence. Senior executives have every reason to sound current. They may have read widely, sponsored pilots, and spoken fluently about transformation. None of that is worthless. None of it is strong enough for the board claim that usually follows: that this person can use AI to reshape work, exercise judgment, and keep control visible in role.

Self-report can start a management conversation. It cannot carry a board decision. The same discipline this series has applied to attribution, readiness, potential, analytics, and governance applies here, at the point where the stakes are newest and the evidence thinnest.

This is the last of seven disciplines I am publishing on evidence-based leadership decisions. Each has examined a different point at which the evidence and the confidence diverge.

What the literature can and cannot support

I should be direct about the evidence base. It is the thinnest in the series, and that is part of the argument. The strategy literature gives boards a defensible frame. Teece's dynamic capabilities describe how organisations sense change, seize opportunity, and reconfigure what they have (Teece, 2007). Eisenhardt and Martin showed those capabilities are identifiable processes rather than mystique (Eisenhardt & Martin, 2000). If AI matters strategically, it matters through leaders who actually reconfigure work, not leaders who describe the possibility of doing so.

The work-design literature sharpens the unit of analysis. Jesuthasan and Boudreau argue that automation decisions only become tractable when roles are decomposed into tasks, because AI transforms tasks, not job titles (Jesuthasan & Boudreau, 2022). Acemoglu and Restrepo make the same point from economics: technologies displace some tasks and create others, and the interesting question is always which tasks, for whom (Acemoglu & Restrepo, 2019). Autor's recent work argues the stakes of getting that reallocation right, because AI can either concentrate expertise or extend it (Autor, 2024).

What none of this literature provides is a validated instrument for scoring an individual executive's AI readiness. That instrument does not credibly exist yet, and boards should be suspicious of anyone who claims otherwise. What the literature supports is narrower and more useful. If leaders matter here, they matter through observable reconfiguration of work, judgment, and control. That can be evidenced without a score.

This is where I abstain from a stronger claim. I cannot tell a board how AI-ready a leader is on a scale. I can tell a board what record would justify using those words at all.

AI Operating Agency

AI Operating Agency is a bounded, observable record of how a leader uses AI to reshape work, judgment, and control in role. Not enthusiasm about technology or fluency in a strategy session. A record.

Four lenses put a leader's claim against their record:

First, work redesign. What tasks in the leader's own span have actually changed? Which were removed, automated, restructured, or newly created? A leader who cannot point to specific task-level change is describing an intention, not a record.

Second, judgment discipline. Where has the leader drawn the line between what AI informs and what humans decide? Which decisions did they refuse to delegate, and why? A leader who cannot say where the machine's claim stops has not yet done the governance work.

Third, control architecture. What has the leader put in place so that use remains inspectable? Review points, escalation rules, challenge routes, records of override. Control that lives in one person's habits is not control the board can rely on.

Fourth, resource movement. What did the leader actually move? Budget, headcount, time, their own attention. Reconfiguration without resource movement is a slide, not a change. Across a real senior team, the record will be uneven, and that is the point. A group CFO may have redesigned the close while never touching judgment discipline. A divisional CEO may have moved budget while keeping control invisible. The four lenses give the board a way to see the unevenness instead of averaging it into a label.

The record is role-specific by design. A leader can hold a strong record in one role and an untested one in the next, which is exactly the lesson of Role-Conditional Readiness applied to a newer question. And where AI has not yet materially reached a role, the honest position is not a low score. It is abstention: not evidenced yet, and here is what would need to be true to change that.

Five questions for the record

Two things would settle this differently. A validated, role-general instrument for assessing executive AI readiness that had survived independent scrutiny. Or evidence that self-reported confidence tracks observable operating change closely enough to be relied on. Neither exists today.

Boards do not need a bigger self-assessment form. They need five questions with evidence attached.

1. What work changed? Task-level, in the leader's own span, with dates.

2. What decisions stayed human, and why? The judgment line, stated by the leader and visible in the record.

3. What controls exist, and who can inspect them? Not policy documents. Working controls with named routes of challenge.

4. What resources moved? Budget, people, time, attention. Shown, not asserted.

5. What is not evidenced yet? The abstention, stated plainly, with the conditions that would change it.

In “What Boards Still Miss About Assessing Senior Leaders”, readiness stopped being a portable label. In the following article “Why High Potential Fails Boards”, "high potential" stopped disguising evidential gaps, while in “Where Leadership Analytics Goes Wrong and How to Fix It”, system claims had to show where they stopped. Finally, in “Why Governance Must Come Before AI in Leadership Decisions”, governance had to exist before deployment. This article closes the loop at the person; the leader using the system deserves the same evidential discipline as the system itself.

Those five questions are the same discipline this series has been building from the start, pointed at the newest claim boards are being asked to accept. The board that asks them will sometimes get a thinner answer than it hoped, and that is the system working. A thin record honestly stated is a governable position, but a confident label with nothing underneath it is not.

That is where this series ends, and where the standard begins. Boards are entitled to make consequential claims about people, systems, and readiness, and they are entitled to make them only on the record.

Written by James Nash.

First published on inBeta.io. Co-published on Substack. Summer 2026.
Series: The Seven®, by James Nash. © Copyright 2026 inBeta. inBeta, Optics, Divergence and The Seven are all trademarks of inBeta Ltd