AI needs an owner, not a committee

IT, operations, and the business units all have a fair claim. That’s the problem.

From Monty’s Desk

Ask five leaders at a mid-size company who owns AI, and you’ll often get five answers. IT says it does, because it holds the tools and the data. Operations says it does, because that’s where the work gets done. The business units say they do, because they’re the ones already using it.

All three are partly right. That’s exactly why AI stalls.

When everyone has a claim, no one has the job. Pilots pile up, tools multiply, and questions about data and approvals get a different answer in every team. Spend goes up. How the business runs stays the same.

The case for each

IT. IT knows security, access, vendors, and integration. Without IT, AI turns into a pile of unapproved tools and data leaving the building. But when IT owns AI alone, it tends to become a technology project. The tools get picked well and then go unused, because nobody tied them to real work.

Operations. Operations knows the processes, the handoffs, and where time is wasted. It’s often the best place to find the first workflows worth fixing. But operations can’t set company-wide rules on data or vendors by itself, and it rarely has the budget to scale.

The business units. Finance, HR, sales, and service teams feel the pain most and move the fastest. That energy matters. Left alone, though, each unit builds its own version, with its own tools and its own rules. You end up with ten small experiments and no shared way to judge any of them.

Why a committee doesn’t fix it

The usual answer is an AI steering committee. It feels fair, since every function gets a seat.

In my experience running enterprise IT and shared services, committees are good at review and bad at decisions. They meet monthly. They defer hard calls. And when something goes wrong, no single person is accountable for fixing it.

A committee can advise. It can’t own.

What works: one owner, shared rules, many contributors

The fix is simple to say and harder to do:

Name one owner. One person is accountable for the company’s AI plan, its rules, and its results. Where they sit on the org chart matters less than whether they have the authority and the time.
Write shared rules. Set them once, for everyone: what data can be used, which tools are approved, who signs off on what, and when a person has to review the output.
Treat functions as contributors. IT secures and integrates. Operations finds and redesigns the work. The business units run it day to day, and each workflow gets a named business owner.

This is how good shared services have always worked. A central owner sets the standard, and the functions deliver against it.

Where ownership fits in the bigger path

Ownership is one stage of a staged approach I use with mid-size and enterprise companies:

Assess and pick. Map where AI is already in use, what it costs, and choose two or three workflows worth doing first.
Set guardrails and ownership. Name the owner, write the shared rules, and assign a business owner to each workflow.
Run real workflows. Put the chosen workflows into daily use with the people who do the work.
Scale across functions. Take what worked to other teams with a repeatable playbook.
Measure and adjust. Track time, cost, quality, and adoption, and let the evidence drive the next round.

Most companies try to start at stage three or four. Without stage two, though, every step after it rests on goodwill.

A quick test

Ask yourself three questions:

Can you name the one person accountable for AI across your business?
Does every team follow the same rules on data and approved tools?
Could that person tell you what AI is worth to the business today?

If any answer is “not really,” you’re in good company. You’re also in the best spot to fix it, before the habits set.

I lead this work myself, with AI and agentic tools for support. If you’d like an outside view, my fixed-scope AI readiness review starts there: where AI is in use, who should own it, and which workflows to start with.

Learn about the AI readiness review →

Human-led · AI-supported. Prepared with AI and agent support. Reviewed and approved by Monty McCoy, principal, McCoy Ven. LLC.