A question is being asked in senior leadership meetings that did not need to be asked five years ago. If one of our AI systems produces an outcome we did not intend — a hiring decision, a credit assessment, a service interaction the regulator notices — who in this team answers for it? Around the table, nobody disagrees with the question. Nobody names a specific person.
In Brief
- 62% of organisations cannot produce a comprehensive inventory of the AI applications already running across their business — the footprint has outrun the governance map.
- 58% of senior leaders cite unclear or fragmented ownership as the primary barrier to measuring AI performance, and only 25% have fully implemented AI governance programs.
- The accountability gap persists because AI was deployed faster than the leadership team negotiated who would answer for unintended outcomes.
- The first move is not redrawing the governance chart. It is assessing where AI accountability genuinely sits across the team today.
The question receives an answer-shaped silence. The gap between asking and answering is wider than any single executive in the room realises.
The split that looks like ownership
The ownership gap is not unique to any one sector or organisation type, and the evidence for it is now substantial. Pearl Meyer’s 2026 survey of corporate leaders found a four-way split on who owns AI that is instructive precisely because it maps so cleanly onto the fragmentation visible inside most senior leadership teams (Pearl Meyer, 2026, as cited in Gerut, 2026). 32% said the leadership group as a whole owns AI. 22% said the layer below. 27% pointed to individual business leaders. 17% named functional heads in HR, legal, or finance. Four roughly equal answers, with no majority among them. McKinsey’s 2025 State of AI survey of organisations across sectors found a similar picture: only 28% said their CEO takes direct responsibility for AI governance oversight, and in many cases governance is jointly owned — with on average two leaders reported to be in charge (McKinsey & Company, 2025). Two leaders nominally in charge of the same function is not shared ownership. It is the condition that produces no owner when a decision needs to be made.
A distribution like that, inside a single leadership team, is structurally indistinguishable from no owner. When ownership is held in roughly equal quarters by four different bodies, the question of who is accountable becomes the question of whose turn it is, and that question is decided after the incident, not before. The question of who owns AI is not receiving a wrong answer from the team; the team is offering several answers that cannot all be true at once. McKinsey’s finding that high-performing organisations are three times more likely to have senior leaders who demonstrate clear ownership of and commitment to AI initiatives reinforces what the fragmentation data implies: the difference between accountable and unaccountable AI is not technical sophistication — it is a leadership team that has resolved the question before it is forced to (McKinsey & Company, 2025).
This gap was not caused by any one executive. It is the predictable result of how AI entered the organisation: in pilots, in productivity experiments, in vendor demonstrations that ran ahead of the governance conversation. By the time the AI footprint became material, the accountability conversation that should have preceded deployment had already been overtaken by deployment itself. Larridin’s 2026 study of senior leaders found that 62% of organisations cannot produce a comprehensive inventory of the AI applications already in use across their business (Larridin, 2026). The footprint outruns the map.
Why leadership teams fail to see the gap in accountability
The Pearl Meyer survey reveals a second pattern that explains why this ownership gap goes unnoticed at the leadership level. 100% of the senior governance stakeholders surveyed believed their executive team operated as a cohesive unit. Only 66% of the executives themselves agreed. A third of those executives reported that their team did not actually work well together (Pearl Meyer, 2026, as cited in Gerut, 2026).
Brad Jayne, the principal at Pearl Meyer who led the study, observed that AI is not creating new organisational problems — it is illuminating ones that were already there (Gerut, 2026). The gap is not capability. Each executive is individually accomplished, individually accountable for their function, and doing what their role requires. What is missing is the practice of negotiating who answers when a system spans every function at once — which is exactly what an enterprise AI deployment now does. Grant Thornton’s 2026 AI Impact Survey of nearly 1,000 senior business leaders makes the cross-functional nature of this gap concrete: inside many organisations, COOs overseeing AI-affected operations are discovering governance gaps that CFOs are not funding and that CIOs and CTOs are not surfacing (Grant Thornton, 2026). Each function sees its own piece. Nobody sees the whole.
The result is a quiet asymmetry. Senior leadership and governance bodies hear “we’ve got this” because each executive can speak with authority about their own domain. Internally, the team knows that the question of who would answer for an AI system producing an unintended outcome has never actually been resolved. Executives tell their committees and governance stakeholders the AI question is handled, then turn inward without a clear answer of their own. The 58% of leaders in the Larridin study who cite unclear or fragmented ownership as the primary barrier to measuring AI performance are describing the same condition (Larridin, 2026). The measurement problem is a downstream symptom of the ownership problem. Grant Thornton’s 2026 survey puts a precise number on the consequence: 78% of senior business leaders lack strong confidence that their organisation could pass an independent AI governance audit within 90 days (Grant Thornton, 2026). The confidence reported upward and the defensibility of the position are not the same thing.
The question executives should ask
When an AI system produces an outcome the organisation did not intend, the distributed ownership pattern produces a specific kind of response. The first move is to find whoever was nearest: the team that deployed the model, the vendor that supplied it, or the function that owns the affected process. The question becomes who was closest to this incident rather than who was accountable for it. Those are different questions, and external scrutineers know it. Regulators, senior leadership, and the public asking about an AI outcome are asking the second one. Grant Thornton’s finding that 78% of senior leaders could not confidently pass an independent AI governance audit within 90 days is what the proximity-not-accountability pattern accumulates to at scale (Grant Thornton, 2026). The organisation has been deploying AI. It has not been building the accountability structure that would let it defend what the AI has done.
This is where the growing cost becomes visible. Each incident handled by proximity rather than accountability widens the gap between what senior leadership has been told and what managers and their teams can actually answer. Larridin found that only 25% of organisations have fully implemented AI governance programs; most have AI policies on paper, but few have the prior negotiation that would let any of those policies attach to a person who can be asked the question and held to the answer (Larridin, 2026). The longer that gap holds, the harder it becomes to close without an external prompt. The external prompt, when it arrives, arrives as a regulatory inquiry, an audit finding, or a public incident the team is required to explain in real time.
Start where accountability should sit
The instinct, when this gap is highlighted, is to redraw the org chart and appoint specific roles like a Chief AI Officer, create a steering committee, and rewrite the AI policy. Any of these may eventually be the right move. None of them is the first move.
The first move should have been to undertake an assessment before deployment. This decision needs revisiting. Run the question through the leadership team as it stands today. If our AI system produces an outcome we did not intend, in this specific context, who answers for it? Then, separately, who could? The gap between those two answers is the diagnostic finding. In most leadership teams, the people who could answer are not the people the current structure says would. That mismatch is the thing to address. The structure itself is downstream of where accountability actually sits. McKinsey’s 2026 State of AI Trust research makes the performance case for closing it: organisations that assign clear ownership for AI governance exhibit materially higher governance maturity scores than those without a clearly accountable function — with a measurable gap between the two groups that widens as AI systems become more embedded in critical operations (McKinsey & Company, 2026). The gap is not just a governance risk. It is a performance drag that grows worse over time.
An executive who runs that assessment quietly, before the regulator, the auditor, or the incident forces it, is choosing the timing of the conversation. The team gains the chance to negotiate where accountability genuinely belongs while the question is still hypothetical. The leader who waits is inheriting a different conversation — one in which the negotiation happens after a specific incident has named a specific person, with the framing the incident sets rather than the framing the team chooses.
The leaders who currently report confidence upward are not misrepresenting their teams. They are reporting the confidence that each individual function genuinely holds. What is absent is the structural answer that would make that confidence defensible when a specific AI system produces a specific unintended outcome and a specific person is asked to explain it. The executive who closes that gap before it is closed for them is operating with a different organisation eighteen months from now than the leader whose answer is given by the incident that exposes the question.
References
- Gerut, A. (2026, April 22). Boards say the C-suite owns AI strategy. The C-suite doesn’t agree. Fortune. https://fortune.com/2026/04/22/ai-ownership-c-suite-board-disagree-pearl-meyer-survey-brad-jayne/
- Grant Thornton. (2026, April). 2026 AI impact survey. Grant Thornton Advisors LLC. https://www.grantthornton.com/services/advisory-services/artificial-intelligence/2026-ai-impact-survey
- Larridin. (2026, February 3). New study shows C-suite leaders highly confident in AI ROI even as 58% claim there’s no clear ownership of AI and 75% lack AI governance. BusinessWire. https://www.businesswire.com/news/home/20260203918939/en/New-Study-Shows-C-Suite-Leaders-Highly-Confident-in-AI-ROI-Even-as-58-Claim-Theres-No-Clear-Ownership-of-AI-and-75-Lack-AI-Governance
- McKinsey & Company. (2025, November). The state of AI: How organisations are rewiring to capture value. McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- McKinsey & Company. (2026, March). State of AI trust in 2026: Shifting to the agentic era. McKinsey & Company. https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/tech-forward/state-of-ai-trust-in-2026-shifting-to-the-agentic-era
- Pearl Meyer. (2026). Boards say the C-suite owns the AI strategy. The C-suite doesn’t agree. Pearl Meyer. https://pearlmeyer.com/in-the-news/boards-say-the-c-suite-owns-the-AI-strategy-the-c-suite-doesnt-agree
