Someone has to own AI. It probably isn't who you think.
Ask a team of twenty who is in charge of AI and you will usually get a pause, then a name. The name belongs to whoever set up the first chat account, or the person who keeps forwarding articles, or the owner who mentioned it at a staff meeting once. Nobody chose them. The job landed on them because nobody else picked it up.
How the job gets assigned by accident.
In a smaller organization, AI tends to arrive through side doors. One person tries a chatbot to tidy up a proposal. Someone in the front office connects a form tool to the shared calendar. Within a few months there are several tools running and no single person who can say what they all do or why.
At that point the questions start finding someone. Can we paste client notes into this? Who has the login? Why did the intake summary come out wrong on Tuesday? Whoever answers the first few becomes the de facto owner, usually without time, authority or any change to the rest of their job. The AI work ends up squeezed into their evenings.
An accidental owner is better than none, but it has two problems. The work gets done unevenly, because it competes with that person's real duties. And when they leave or burn out, everything they were holding goes with them: the logins, the settings, the reasons things were set up the way they were.
Why the most technical person is often the wrong pick.
The reflex is to hand AI to whoever is best with computers. In a town office that might be the person who resets passwords. In a professional firm it might be the analyst who writes spreadsheet formulas nobody else can read. They will have no trouble with the tools, and the tools are the smallest part of the job.
Most of what an AI owner does is about people and process. They need to know which tasks eat the team's week, which ones carry risk, and who will resist a change and why. They have to say no to a tool a senior colleague likes, and explain the reason in plain words. They have to notice when a shortcut is saving an hour for one person while costing two hours for the person downstream.
The person who understands how the work moves through the building will learn the tools faster than a tools person will learn the work.
Someone who sits where work changes hands is often a better fit: an operations lead, an office manager, a practice administrator, a program coordinator at a nonprofit. They already know where things get stuck. They talk to most of the team every week. And they are trusted by people who would never ask the IT person a question that makes them look slow.
What the owner does in an ordinary week.
The role is smaller and more practical than the title suggests. For a team of, say, twelve to forty people, it comes to a few hours a week. Most of that time goes to keeping a short list of what is in use, answering questions before they turn into workarounds, and watching for the moment a useful habit should become a shared one.
A typical week might include checking which accounts are active and who has them, reading through a couple of outputs from anything running on its own, and spending twenty minutes with one person who has a task they think AI could help with. Once a month the owner reports to leadership in a page or less: what changed, what it cost, what went wrong and what they want to try next.
Everyone else keeps using AI and suggesting ideas. The owner's job is to make sure good ideas get tried properly and bad ones get stopped early, and that someone can always answer the question "what are we running, and is it safe?"
Give them time, a budget line and the right to say no.
Naming an owner does nothing if the role is unfunded. The common failure is a title with no hours attached, so the work slides back to evenings and the person resents it by the second quarter. Take something off their plate before you add this to it, and say what came off in front of the team.
They also need a small amount of real authority. That means a modest budget line they can spend without a meeting, the standing to pause a tool that handles client or resident data badly, and a direct line to whoever signs off on new software. Without that, the owner becomes a note-taker for decisions other people make.
Finally, write the role down. A spoken assignment fades within a month, while a written one survives a reorganization and can be handed to the next person.
A one-page job description for an AI owner.
If a line doesn't apply, cross it out; if nobody on staff could do most of it, that tells you what to train for first.
AI owner: role summary
- Purpose: makes sure the organization's use of AI is useful, known and safe, and that new uses get a fair trial.
- Reports to: the executive director, owner or department head, monthly, in one page.
- Time: a set number of hours each week, protected on the calendar, with named duties moved elsewhere to make room.
- Keeps: a current list of every AI tool and automation in use, with who uses it, what data it touches and what it costs.
- Decides: which tools may be used with client, patient, student or resident information, and can pause any tool that fails that test.
- Spends: a small monthly budget for trials and training without separate approval, up to an amount leadership sets.
- Supports: holds a standing slot each week for anyone to bring a task, a question or a mistake.
- Reviews: checks the output of anything that runs unattended at least once a month and signs off on it.
- Hands over: keeps logins, settings and the reasons behind them written down, so a successor could take over within a week.
- Is good at: listening, explaining things without jargon, and saying no politely. Comfort with software helps but comes second.
