Who Owns What Gets Cut

The Deliberate AI Leader — A Series for Executives Who Want to Get This Right – Part 19

The Half of Governance Nobody Finishes

In Who Owns This? Defining AI Accountability Before You Need To, we walked through the five ownership questions every AI system needs answered before it goes live. Most leaders who’ve read that piece have started naming an owner up front. Far fewer have thought through the other half of the same problem: who owns the decision to shut something down, and who actually does the work of unwinding it.

That gap matters more than it seems. A tool with no clear owner at launch causes confusion. A tool with no clear owner at retirement just keeps running, quietly costing money or creating risk, because retiring it is nobody’s job.

Who Decides

It shouldn’t be whoever built the thing. That’s not a knock on anyone’s judgment, it’s just how sunk cost works. The person who spent three months configuring an agent is the least objective person in the room about whether it’s earning its keep.

The decision belongs with whoever the pre-launch governance structure named as the owner, the same person or committee identified using the framework in Who Owns This. If that structure was never set up for a given tool, and plenty of tools in most companies predate any real governance process, the default owner is whoever owns the budget line or the business process the tool touches. Someone has to be able to say yes or no without checking three other people first.

The actual test is simple to state, even if the answer takes some digging: is this still solving the problem it was built for, better or cheaper than the alternative of not having it. If the honest answer is unclear, that uncertainty is itself the signal. Set a decision date and stick to it, rather than letting an unclear case drift for another quarter.

Who Does It

Deciding to cut something and actually unwinding it are two different jobs, and they don’t always land on the same person. Unwinding cleanly usually needs three things in the room: whoever has technical access to actually shut the system down and revoke its permissions, whoever understands what else touches it, and whoever owns the vendor relationship or contract.

That first piece matters more than it sounds. In Chatbots, AI Agents, and Automation: What You’re Actually Buying, we noted that an agent inherits the permissions of whoever set it up: access to a CRM, a shared drive, customer records. Turning the agent off doesn’t automatically revoke any of that. Someone has to go in and close those doors specifically, or the access just sits there, unused but still live, which is its own quiet risk.

What to Consider Before You Pull the Plug

A short checklist worth working through before anything gets switched off:

• What breaks downstream immediately. Run the same dependency questions from The Vendor Question Nobody Asked During the Rush: which workflows assume this exists, and who would notice first if it stopped.
• What data or history needs to be preserved before the system goes dark, and where it’s actually going to live afterward.
• Whether people quietly built manual workarounds to cover for the tool’s gaps while it was live. Those workarounds, the pattern we described in Human Middleware Was Never a Scalable Operating Model, also need to be surfaced and undone deliberately, not left to be discovered by accident later.
• Who needs to know before it goes dark, not after. A quiet shutdown that surprises a department is how trust in the whole AI program erodes, even when cutting the tool was the right call.
• What the contract and cost timeline looks like, and whether there’s a notice period that needs to start now rather than after the decision is finalized.
• Whether something replaces the tool, or the process reverts to manual, and who owns that follow-on decision. Cutting something is not the same as solving the problem it was cutting into.

The Cut Nobody Makes

The real risk in most AI portfolios isn’t a dramatic failure. It’s the quiet underperformer that never gets an actual decision either way, because no one owns making it. As we covered in What Survives the Bake-Off, the initiatives that earn their place get resourced and the rest are supposed to fall away. In practice, the ones that don’t have a named owner for the shutdown decision just linger, absorbing budget and attention that a clear cut would have freed up.

Getting the front end right is still where to start. If you’re setting up a new automation and want the shutdown conversation to be easy later instead of impossible, Your First AI Governance Win covers how to build that discipline in from day one.

Where This Leaves You

Somewhere in your AI portfolio, there is likely something that isn’t working and hasn’t been formally cut, simply because nobody has the job of cutting it. Naming that person this quarter is a smaller task than it sounds, and it will save far more than the cost of the tool itself.

If you want help sorting out what’s worth keeping and building the ownership structure for the rest, book a Strategy Call. No pitch, just clarity.

About WHIM Innovation

WHIM Innovation helps organizations harness the practical power of AI, automation, and custom software to work smarter and scale faster. We combine deep technical expertise with real-world business insight to build tools that simplify operations, enhance decision-making, and unlock new capacity across teams. From AI strategy and workflow design to custom monday.com apps and fully integrated solutions, we partner closely with clients to create systems that are efficient, intuitive, and built for long-term success.