Skip to article

Every AI Agent Needs an Owner and an Expiration Date

A named human for each AI agent, a date when it comes up for review, and an off switch designed in advance. What PwC's Jenn Kosar described, and the first list to write.

By Jacqueline V. TwillieNovember 20, 20263 min readGuide
Quick answer

Give each AI agent a clear human owner and a date when it comes up for review. At an online enterprise AI summit on November 12, 2026, PwC's Jenn Kosar described expiration or re-approval dates, monitoring for dormancy, recertification, and an off switch designed in advance.

Can you name the person who owns each AI agent your company runs? Jenn Kosar, PwC's US AI Assurance Leader, spoke about ownership and lifecycle at an online enterprise AI summit on November 12, 2026.

Every AI agent starts with one human owner

Kosar said she would start with each AI agent having a clear human owner, responsible for its outcomes, its performance and the monitoring. She said maybe they will not get into the debate over whether AI agents are semi-human, and described a workforce mindset instead: roles, responsibilities and access rights. In practice she listed a verifiable identity, clear task-specific roles and responsibilities, and well-defined boundaries on where autonomy exists and where it does not.

BNEDai treats an AI agent as a system that a person owns, runs and answers for. Start with a list of every AI agent you run, with one name next to each. Then keep one page per AI agent, with its identity, its role, its limits and its owner's name, where the owner can find it.

Expiration and re-approval dates bring the owner back

Kosar said innovation proliferates, and she gave some techniques. Those include expiration dates or periodic re-approval dates on AI agents, which trigger the business owner who is accountable to confirm the AI agent is still useful. You can also monitor for an AI agent going dormant, she said, and run a recertification when something is not working.

Design the off switch in advance

Kosar said to design an off switch in advance, with the automations or interventions that would turn off an AI agent that starts performing outside its boundaries.

BNEDai's note: write the off switch into the page for each AI agent, next to the owner's name, so it is decided before it is needed.

A review rhythm on the calendar

BNEDai's published example comes from the October 7, 2026 session with Women Applying AI. Jacqueline V. Twillie described a 15 minute meeting on her calendar that recurs every three weeks, where she reviews her AI agents' instructions. She said most of her AI agents run a couple of times a week. A three week cadence works for her, and someone using an AI agent daily may want to check every two weeks. The details are in how often to review an AI agent's instructions.

BNEDai's reading: a date on a calendar with a named owner works the same way as the re-approval Kosar described: someone has to come back and confirm the AI agent is still useful.

Where the owner and the limits sit in a build

BNEDai's default is that an AI agent drafts and a person decides what goes out. The page on every AI agent starts with the same rule shows where each limit is set.

Frequently asked questions

What does Jenn Kosar say each AI agent needs?

A clear human owner, responsible for its outcomes, its performance and the monitoring.

What triggers the owner to confirm an AI agent is still useful?

Expiration dates or periodic re-approval dates, according to Kosar.

What should happen when an AI agent performs outside its boundaries?

Kosar said to design an off switch in advance, with the automations or interventions that would turn it off.

Source: remarks by Jenn Kosar, PwC's US AI Assurance Leader, at an online enterprise AI summit, November 12, 2026.

Reviewed on November 20, 2026.

More on Ownership for Corporate Executives