Most small businesses do not have an AI problem. They have an ownership problem. Ask an eight-person agency who is responsible for AI and you will usually get one of two answers: "all of us" or a long pause. Both mean the same thing in practice — nobody is responsible, so nothing gets decided, and the tools pile up anyway.
You can hear it in how owners describe their own situation. "We've got four tools, and nobody knows what any of them actually do." "Someone on the team uses ChatGPT, someone else tries a different tool." "Six AI subscriptions, no coherent workflow." That is not a technology failure. It is what happens when a new capability arrives without a named owner, a budget line, or a decision process.
This guide covers who should own AI in a business of 1–100 people, what that person actually does week to week, which decisions belong to them and which do not, and how to appoint someone without creating a job nobody has time for.
Why "everyone owns AI" means nobody does
In a large company, AI ownership lands on a head of data or a transformation lead. In a 12-person business there is no such role, so the default is diffusion: everyone is encouraged to experiment, and the assumption is that good practice will spread on its own.
It does not spread, for three predictable reasons.
Nobody can approve spend. A €20-a-month tool sits below most approval thresholds, so it never gets reviewed. Multiply that across five people and you have €1,200 a year of overlapping subscriptions that no single person can cancel because no single person owns the category.
Nobody writes anything down. Your best prompt lives in one person's chat history. When they are on holiday, the capability leaves with them. Our guide on building an AI knowledge base covers the fix, but someone has to be accountable for doing it.
Nobody sets the boundary. Without an owner, each person decides privately what client data is acceptable to paste into a chat window. That is how shadow AI takes root — not through bad intent, but through an absence of anyone whose job it is to say "not that."
The four ownership models for small teams
There are realistically four options, and team size is the main thing that decides between them.
1. The owner-operator model (1–5 people)
You own it. There is no one else, and that is fine. Under five people, AI ownership is not a role — it is 90 minutes a month of your own time: review what is being used, cancel what is not, write down what worked. The failure mode here is not bad delegation; it is pretending this takes no time at all.
2. The internal champion model (6–30 people)
One existing employee takes AI on as a named part of their role, typically 10–15% of their week. This is the right answer for most SMBs. The champion is rarely the most technical person on the team — more often it is whoever is already curious, already helping colleagues, and trusted enough that their "no" sticks.
3. The split model (30–100 people)
At this size, one person cannot carry both the strategic and the practical side. Split it: a leadership sponsor who owns budget and risk (usually the founder, COO, or finance lead) and a practitioner who owns enablement and documentation. The sponsor decides what is allowed; the practitioner makes it usable.
4. The external advisor model (any size, time-boxed)
You bring in outside help to set direction, then hand execution to someone internal. This works for the initial 30–60 days and fails as a permanent arrangement, because an external advisor cannot see which workflows are quietly breaking. If you are weighing this up, do you need an AI consultant walks through when it is worth the money.
The important point: every model names a person. "The management team" is not a person.
What the AI owner actually does
Vague mandates die quietly. Here is the concrete remit, in roughly descending order of value.
- Maintains the tool register. A single sheet: tool, who pays, what it is used for, what data touches it, renewal date. Our AI tool stack audit is the starting template.
- Owns the prompt library. Five to fifteen prompts that produce your actual deliverables — proposal drafts, meeting summaries, support replies — stored where everyone can reach them, with a note on what each one is for.
- Runs onboarding. Every new hire gets a 30-minute session on which tools the business uses, for what, and what never goes into them.
- Reviews output quality. Spot-checks AI-assisted work monthly. The point is catching drift early, not policing colleagues.
- Keeps the policy current. One page, not twenty. How to write an AI policy covers what belongs in it.
- Reports one number up. Monthly: hours saved, spend, and one thing that did not work.
Notice what is absent. The owner does not build integrations, does not become your in-house AI engineer, and does not personally do everyone's AI work. The moment the role becomes "send it to Sarah, she's good with AI," it has failed — you have created a bottleneck, not a capability.
Decision rights: what the owner decides, and what they do not
This is the part most businesses skip, and it is the part that determines whether the role works. Write it down explicitly.
The AI owner decides: which tools go on the approved list; what the standard prompts are; what training happens; whether a given workflow is ready to run with AI in it.
The AI owner recommends, leadership decides: anything above a spend threshold you set (€100 a month is a reasonable line for a small team); anything touching customer data in a new way; anything that changes a client-facing process; any headcount implication.
Nobody decides unilaterally: putting AI in front of customers without a human review step, and using AI in regulated or legally sensitive output. Those need a second pair of eyes regardless of who owns the category.
Pair that with a clear veto. The owner must be able to say "we are not using that tool" and have it hold. Ownership without authority is just extra admin, and it is the single most common reason the role lapses after two months.
How much time it takes — and how to pay for it
Be honest about the cost, because an unfunded role gets abandoned the first busy week.
For a 10–30 person business, budget roughly 4–6 hours a month in steady state, front-loaded to 8–12 hours a month for the first quarter while the register, prompt library, and policy get built. That is about 3% of one person's time. It is not a secret extra; it comes out of something else, and you have to say what.
Three ways small businesses fund it, in order of how well they work:
- Reduce the champion's other commitments. Take one recurring task or meeting off them. Cleanest, and it signals the role is real.
- Attach it to an existing ops or enablement role. Whoever already owns tooling, onboarding, or internal process is the natural fit.
- Add a small stipend or review it at the next salary cycle. Works, but only alongside option 1 — money does not create hours.
What does not work is adding it to a full plate and hoping. If you cannot find four hours a month, you do not have an AI ownership problem yet; you have a capacity problem, and AI will not fix it for you.
How to choose and appoint your AI owner
Pick on three criteria, weighted in this order: judgement (will they say no when it matters), communication (can they teach a sceptical colleague without condescending), and curiosity (are they already experimenting). Technical depth is fourth and often fourth by a long way.
Avoid two tempting but poor choices. The first is your most technical person, if their instinct is to build rather than to standardise — you will get an impressive internal tool nobody uses. The second is your most AI-enthusiastic person, if their enthusiasm has no brake on it; this role needs someone comfortable turning things down.
Then appoint properly, in writing. A short note to the team covering: who the owner is, what they decide, what comes to leadership, how much time they have for it, and when it gets reviewed. Set the first review at 90 days — long enough to see whether it is working, short enough to change course.
If the honest answer is that nobody on your team fits, that is useful information. It points at hiring for AI skills on your next role, or at the time-boxed external model while you grow into it.
Where AI ownership goes wrong
The title with no time. Someone is named "AI lead" at an all-hands and given no hours. Nothing changes, and the team concludes AI is theatre.
The bottleneck. The owner becomes the person who does all AI work rather than the person who makes AI work usable. Fix it by making the deliverable documentation and training, not output.
The gatekeeper. Ownership calcifies into a procurement process where every request waits three weeks. In a small business, approval should take days, not weeks — otherwise people route around it and you are back to shadow AI.
The orphan. The champion leaves and the role is not reassigned. Everything they knew was in their head. This is why the tool register and prompt library are the first two deliverables, not the last.
Ownership without strategy. Someone diligently manages a stack of tools that is not connected to any business priority. The role only pays off inside a direction — our walkthrough on how to create an AI strategy for small business covers setting that first, and training your team to use AI covers the enablement side the owner will run.
The bottom line
AI ownership in a small business is not a new hire, a job title, or a committee. It is one named person, four to six hours a month, a short written list of what they decide, and a real veto. Under five people that person is you. Between six and thirty it is an internal champion with 10–15% of their week protected. Above thirty, split the strategic and practical halves.
The businesses pulling ahead on AI are rarely the ones with the best tools. They are the ones where somebody is accountable for the answer to "which tool, for what, with whose data" — and has the time and the authority to keep that answer current.
Where does your business stand on AI?
Take the free 3-minute AI Readiness Quiz and get a personalised score with your next steps.
Take the Free Quiz →