Boards across regulated industries are approving AI investments at a pace that would be unthinkable for almost any other capital allocation. Yet one simple question often goes unanswered: if this decision proves wrong in three years, who answers for it? In conversation after conversation, the answer is rarely clear. Not because leaders are careless, but because the question has not been asked with enough seriousness to force preparation.

KPMG's Global AI Pulse Survey, referenced in its April 2026 board governance work with INSEAD, found that nearly three quarters of boards are perceived to have only moderate or limited AI expertise.

Three quarters. That is not a niche capability gap. It is most boards in the world governing a technology they cannot yet challenge with fluency.

The consequences are already visible. Senior executives describe AI programmes being approved with a speed, optimism and ambiguity that would never pass scrutiny in credit, risk, or technology investment committees. And the advisory market, despite its scale, has not closed the gap.

Two recent studies make the mechanism clear. Harvard Business School and Boston Consulting Group tested GPT-4 against realistic consulting work carried out by 758 consultants. Where a task sat inside AI's genuine competence, work graded by human reviewers improved by more than 40 percent. Where a task sat outside that competence, work came out nineteen percentage points worse than consultants working without AI at all, because the consultants using it could not reliably tell the difference between the two kinds of task. The failure mode was not the model. It was judgment.

A 2024 meta-analysis in Nature Human Behaviour found that human–AI combinations performed better than humans alone on average, but worse than whichever performed better independently: the human or the AI. Results varied by task and by the relative capabilities of the human and the system. The leadership implication is that effective collaboration must be designed and tested, rather than assumed.

This is the fault line running through the market itself. Plenty of firms offer AI advice. Very few are structurally capable of providing independent, senior judgment.

Where the Market Falls Short

Structural Conflicts

Large consulting firms and hyperscalers can deliver competent AI governance workstreams. But many also sell, build, or licence the systems a board may later need to assess. That is not a moral failing. It is a structural conflict, and it means the same firm recommending your AI strategy may have a commercial interest in expanding it.

Tooling Without Judgment

Governance software vendors solve real problems, from mapping systems against ISO 42001 to preparing for the EU AI Act, and we regularly point clients toward them when the gap is technical. But a control mapping tool is not board judgment. It can tell you whether a box is ticked. It cannot tell you whether the person accountable for that box would withstand scrutiny from a regulator, or from a journalist who has already made up their mind.

Independence Without Delivery Pressure

Nonprofit and academic centres bring independence and excellent thinking. What they often do not bring is a commercial delivery model tested under P&L pressure, the kind that comes only from having run something yourself and being accountable for how it turned out.

The Six Fault Lines

Diagnosing an AI programme properly means assessing six separate vectors, because AI risk rarely appears as a single problem.

• Governance: who is accountable, by name, when an AI-driven decision goes wrong, and whether they can explain that decision under real scrutiny.
• Economics: whether the investment case pays back, on what timeline, and who captures the value — the institution, its customers, or the vendor.
• Workforce: what happens to the talent pipeline once AI absorbs the entry-level work that used to train future leaders.
• Sustainability & Infrastructure: whether AI strategy and climate strategy are being decided together, or whether a net-zero pledge and an AI expansion plan coexist untested in the same annual report.
• Resilience & Sovereignty: what happens operationally if a critical AI provider goes down, is acquired, or withdraws service in a jurisdiction the institution depends on.
• Public Trust & Fairness: whether an AI-driven credit decision, claim or underwriting call would survive being explained to the person affected, and to a public that expects better.

Each fault line will be explored in its own article. Together, they form a map of where AI programmes actually break, drawn from people who have watched them break in real institutions, not in a lab.

The Collective

Faultlines was built to close this gap: senior practitioners, each carrying deep expertise in one discipline, working as a single team with no AI system to sell and no licence to protect. Its members bring operational, regulatory, academic and ESG expertise, and every one of them has carried real accountability — P&L ownership, board responsibility, or regulatory exposure — not simply advised on it.

The biggest risk in AI is rarely the technology itself. It is the gap between what the technology can do and what the people governing it understand well enough to challenge.

The members of the Faultlines Leadership Council did not join for another logo on a slide. They joined because they believe something simple and hard earned: judgment beats tooling, and judgment only counts if the person offering it has carried the weight of being wrong.

What Clients Actually Receive

Not a generic framework with a logo pasted on the cover. Not another slide deck. You get senior people who have run businesses, written policy or sat on boards, spending real time inside your specific decisions. You walk away with an assessment built to withstand challenge from a regulator, a journalist, or your own audit committee.

In practice, the AI Leadership Review examines a consequential AI decision, identifies its dependencies and stress-tests its assumptions through the relevant fault lines. It brings independent senior judgement to the people accountable for the outcome, providing a clearer basis to proceed, change course, delay or stop.

AI capability is accelerating faster than most boards' capacity to govern it. Closing that gap is not a compliance exercise. It is the difference between leading the next decade and becoming the case study everyone else studies afterwards.

If you want to stress-test your institution's position across these six vectors, we will walk you through it directly. Email confidential@faultlines.group.

Sources

• KPMG and INSEAD, Board Governance and AI: Global Survey, April 2026.
• Fabrizio Dell'Acqua et al., Navigating the Jagged Technological Frontier, Harvard Business School Working Paper 24-013, September 2023.
• Michelle Vaccaro, Abdullah Almaatouq and Thomas Malone. When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 28 October 2024.