A board that names an executive owner for its AI systems and stops there has not built accountability. It has built a name on a document, which is a different thing entirely, and the difference surfaces at the exact moment it matters most.
Every governance framework written for agentic AI in the past year arrives at the same first move: assign a person. Give the system a named executive who answers for it. This is correct, and it is also the easy part, which is why so many organizations stop there and believe the work is done. A name in a registry costs nothing. It requires no budget, no new tooling, no uncomfortable conversation about who can override whom. It can be completed in an afternoon and reported to the board as finished.
What that name does not do, by itself, is grant the owner anything. Real ownership requires a defined boundary on what the system may do without asking. It requires the practical power to suspend the system when something looks wrong, not the theoretical authority to request a suspension through channels. It requires access to the information that would tell the owner a problem exists before the problem becomes an incident. None of that arrives with the title. All of it has to be built, deliberately, by someone willing to give up a piece of control to the person they just named.
This is the same failure Glenn has watched leaders make for thirty years, wearing a new label. A leader assigns work without stating what authority comes with it, what success looks like, or what happens if the assignment is not met, and then wonders why the outcome was wrong. The failure was never the assignment. It was the absence of the conversation that should have come with it: what needs to be done, what power has actually been transferred, what the standard is, and what the timeline is. Call it what it is: an accountability contract, not a name plate. Most leaders were never taught the difference, and most boards approving AI governance policy this year are about to make the identical mistake at the system level.
The failure is more dangerous here than it was with a human report. A person handed a title without real power will usually say so. They will push back when asked to answer for outcomes they cannot control, because self-preservation makes people vocal about the gap between their name and their authority. An executive named as the owner of an autonomous system rarely gets that warning instinct in time. The system keeps running at whatever speed it was built to run. The owner discovers the gap between their name and their actual power only when an action has already executed and someone asks why they did not stop it, and the honest answer is that they were never given a way to.
This is worse than having no owner at all, and boards should sit with that fact before they check the box. An unowned system is at least visibly ungoverned, which tends to attract attention and get fixed. A system with a named owner who was never given eligibility, envelope, or evidence looks governed on paper. It satisfies the audit. It clears the quarterly review. The board reports to its own shareholders that accountability has been assigned, and every sentence of that report is technically true, right up until the incident that proves the ownership was ceremonial rather than structural.
Activity is not accountability, and most organizations still cannot tell the difference. Filling a governance registry with names is activity. An executive who has formally accepted eligibility, who holds a version-controlled authority envelope, who can suspend the system without asking permission first, and who knows exactly what happens to them if they do not use that power when the moment calls for it, is accountability. The first produces a document. The second produces a person who will actually act.
This argument, and the specific eligibility standard, authority envelope, and escalation architecture that turns a named owner into an accountable one, is developed in full in Touch Stone Publishers' Executive Leadership Playbook on agentic AI governance and its companion research.
The name was never the governance. It was the promise of governance, made in a single afternoon, waiting to be kept or broken by everything the board does next.