Who Should Own AI at Your Firm?

1. A Committee Produces a Memo, Not Adoption

The instinct at most firms is to handle AI the way they handle anything important: form a committee. Pull in a partner, the COO, someone from IT, maybe a keen associate. Meet monthly. Wait for adoption to emerge.

It does not emerge. A committee is built to deliberate, and adoption is not a deliberation. It is a thousand small acts of building, fixing, and persuading that happen between meetings, and a committee does none of them. What a committee produces, reliably, is a memo: a statement that AI is a priority, a list of principles, and a plan to reconvene.

The reason is structural, not a failure of the people. When a group owns a thing, no individual does. Responsibility spreads until it is thin enough to evaporate. Everyone assumes someone else is building the setup, chasing the laggards, and answering the analyst who got a bad output on Tuesday. No one is, because no one's name is on it.

This is the same root cause behind most stalled rollouts. The missing owner is the first of the four failure modes in why AI rollouts fail at investment firms, and it is the one that quietly causes the others.

2. The One-Owner Rule

Every AI effort that goes anywhere has one person who owns it. Not leads a workstream. Owns it, the way a deal has a deal lead and the buck stops in one place.

The rule is simple and unforgiving: one named person, accountable for whether the firm actually changes how it works. Not a committee that advises them. Not a steering group that meets quarterly. A single owner who wakes up responsible for adoption and goes to bed responsible for it. The committee, if you keep one, exists to give that person cover and to settle the occasional policy question. It does not own the work.

This feels too small for something the partners have decided is strategic. That is exactly the mistake. Strategic does not mean diffuse. The most strategic thing a firm can do with AI is put it in the hands of one capable person and back them, because concentration is what turns intent into a changed workflow.

One owner is also the thing that makes everything downstream possible: a budget that can be spent, a sequence that can be run, a number someone is on the hook to produce. Without it, the strategy is a document. With it, the strategy has an engine.

3. What the Owner Actually Does

The role is concrete, not ceremonial. Strip away the title and the owner does a specific, unglamorous job, week in and week out.

Builds the setups. The shared projects, the prompts, the templates that encode how the firm works, so a person opens a tool that already knows the house style instead of starting from a blank box. Fixes what breaks. When a workflow gives a bad output or a new model changes behavior, the owner is who adjusts it, so the team's trust is repaired before it curdles.

Answers the questions. The owner is the person an analyst goes to at 9pm when the tool is not doing what they need, which is how small frustrations get solved instead of becoming reasons to quit. Drives the adoption. Picks the next workflow, runs the pilot on real work, chases the team that is dragging, and spreads the win from the people who got it to the people who have not.

Shows the partners the number. Translates all of the above into the language leadership funds: hours saved, the workflow now done the new way, the proof that the spend is working. This is the part that keeps the air cover alive.

None of this is exotic. It is the patient, hands-on work of changing how people work, and it only happens when one person is paid and expected to do it. The mistake is assuming it happens on its own once the licenses are bought.

4. Who It Should Be

The owner is chosen wrong more often than not, because firms reach for the obvious person instead of the right one. Two assumptions do the damage: that it should be the most senior person, and that it should be someone technical.

It should not default to the most senior. A managing partner is the wrong owner of the day-to-day work, because they cannot do it, and the role is daily. Seniority belongs in the sponsor, not the owner. And it should not default to IT. AI adoption is a behavior-change problem dressed as a technology problem, and the person from IT is often the worst placed to persuade a skeptical deal team, because the deal team does not take their cues on how to do the work from IT.

The right owner has three traits. They are respected, so the deal team takes the work seriously when it comes from them. They are genuinely interested, so they keep going through the unglamorous middle when the novelty wears off. And they have the time and the cover to do it, so it is a real part of their job and not a hobby crammed around a full deal load. A respected associate or VP who is fascinated by this beats a senior partner who is not, every time.

This is an operating-model decision, not a staffing one. The deeper question of whether to train that person up or hire the capability from outside is its own choice, and it is the subject of upskill or hire, the AI talent decision. The shape of how you grow the internal version is in building an internal AI champion.

5. The Three Ownership Models

There are three ways to fill the owner role, and a fourth that is really a blend of them. The right one depends on the firm's size, its talent, and how fast it wants to move.

The hybrid wins for a reason. It removes the cruel choice between moving fast (which favors outside help) and building lasting internal capability (which favors a champion). You get both: the partner supplies the experience and the pace from day one, the champion absorbs it, and ownership moves inside the firm once the champion is ready to carry it.

6. Do You Need a Head of AI Yet?

Almost certainly not yet. The dedicated head of AI is the model firms reach for too early, because it looks like seriousness. Mostly it buys an expensive seat with not enough work to fill it.

The honest test is whether there is a full-time job's worth of AI work. At a ten or twenty-person firm, there is not. There is real work, but it fits inside a capable person's role alongside their other responsibilities, and forcing it into a full-time hire either over-scopes the spend or invents busywork to justify the title. A head of AI with too little to do tends to build complexity for its own sake, which is worse than no head at all.

The threshold is about workload and scale, not prestige. A dedicated head of AI starts to make sense when the firm is large enough that adoption across many teams and a growing portfolio genuinely exceeds what one person can do part-time, when there are standing systems that need full-time stewardship, and when the AI work is continuous rather than a series of projects. That is a real point for some firms. It arrives later than the urge to hire does.

Until then, an internal champion with real time, or an outside partner, or the hybrid of the two, covers the role at a fraction of the cost and with none of the premature overhead. Hire the head of AI when the work demands it, not when the title sounds right.

7. Where IT and Compliance Fit

IT and compliance are essential to an AI effort. Neither should own it. Confusing enable with own is how firms end up with AI that is secure, governed, and unused.

IT enables. It provisions the right tiers, sets the access controls, handles the integrations, and keeps the data inside the perimeter. That is necessary work, and it is infrastructure work, not adoption work. Hand IT the ownership of adoption and you get a tool that is correctly configured and that no deal team feels any pull to use, because the people who decide whether the work changes do not take their lead from IT.

Compliance sets the boundary. It decides what data can go where, what the firm will tell an examiner about how it supervises AI, and how to answer the LP who asks in diligence. That boundary is what lets the owner move fast inside it without fear. It is a guardrail, not a steering wheel. A compliance team that owns AI will optimize for the absence of risk, which means the absence of use.

The clean division: compliance draws the lines, IT builds the rails, and the owner drives adoption inside both. Each is necessary. Only one of them is the owner, and it is not the two whose job is to say where the car cannot go.

8. The Reporting Line and the Air Cover

The owner does the work, but the owner cannot grant themselves authority. That has to come from above, and where it comes from decides whether the role survives contact with a busy firm.

The owner needs a senior sponsor, ideally a managing partner, who is visibly behind the effort. The sponsor does not do the daily work. The sponsor does something the owner cannot: makes it unambiguous that this is a priority, protects the owner's time from being eaten by deal work, and tells the partners to actually use the thing when they would rather not. That is the air cover, and without it a respected associate trying to change how partners work gets quietly overruled by the next live deal.

The reporting line should be short. The owner reporting straight to a managing partner or the COO sends the signal that this is real and gives them a fast path when a partner needs to be nudged. Bury the role three levels down and the signal inverts: the firm reads it as optional, and optional is fatal for something that asks people to change a habit.

The pairing that works is a junior-enough owner to do the daily work and a senior-enough sponsor to make it stick. Get either wrong, a sponsor too distant or an owner too senior to do the work, and the role hollows out from one end or the other.

9. When to Bring an Outside Owner

Sometimes the right owner is not on the payroll yet, and pretending otherwise stalls the whole effort. The signal is specific: you have the budget and the will, but no internal person who is both able and available to own it.

That gap is more common than firms admit. The respected, interested person you would pick already has a full deal load and no spare hours. Nobody in the building has run an AI adoption before, so even a willing owner is learning on your time and your risk. The work is needed now, and growing the internal owner from scratch is a slow way to start. When those are true, an outside owner is not a failure to build internally. It is the fastest way to get a competent owner in the seat today.

This is exactly the role an AI Operating Partner plays: fractional AI leadership that does the owner's job from outside, the setups, the fixes, the adoption, the number, without adding a full-time seat. Done well, it is also the front half of the hybrid: the partner runs it while an internal champion learns the role alongside them, then ownership moves in-house. You get the experience now and the internal capability later, instead of choosing between them.

The test is not whether you could eventually grow your own owner. Almost any firm could. It is whether you can afford to wait while you do, and for a firm with real deal flow and a real adoption gap, the answer is usually no.

10. Where to Start

Name the owner this week. One person, accountable for whether the firm changes how it works. If you cannot fill that line with a name, that is the first problem to solve, before any tool, because every other decision depends on it.

Then give them two things: real time, carved out of their other work and protected, and a senior sponsor who makes the priority unmistakable. An owner without time is a title. An owner without air cover is a volunteer who gets overruled. With both, you have an engine. The fuller version of the operating model sits inside the AI strategy and roadmap for investment firms.

If the honest answer is that no one inside is both able and available right now, that is the case for an outside owner. An AI Operating Partner does the owner's job from day one and, when you want it, trains your internal champion to take the keys. An AI Readiness Sprint is the fastest way to name the first workflow and decide which ownership model fits your firm.