Every business needs an AI department. Most don’t have one.

AI works. That part is settled. Most businesses have a shelf of unused licences, a nervous IT team, and staff who’ve quietly decided it isn’t for them. The tools work. Adoption is the constraint.

It does not compete with any of those departments. It makes each one do what it already does faster and cheaper, which is the whole point of it. It sits with your leadership team and works out specifically what’s stopping this business, not businesses in general, from using AI. It sits with your staff and gets them past the nervousness of trying, on their own real work, until they’ve built something themselves and stopped being afraid of it. It keeps a running view of what’s changed this month, because the tools move faster than anyone has time to track. And it does this without belonging to any one department, because the moment it does, every other department treats it as someone else’s initiative.

Name it plainly. Not a transformation programme, not a workstream parked inside IT, not an innovation lab with a whiteboard and no mandate. The way work gets done is changing, and a business that cannot say that out loud in the name of a department will struggle to say it anywhere else.

Folding it into an existing function is the quiet way to make sure nothing happens. Transformation teams are built for three-year programmes with a steering committee. IT is built to keep things running and to be careful about anything that might break them. Both are doing the job they were set up to do. Neither of those jobs is this one, which is to work out each month what these tools can newly do and have people using them by Friday.

So pull the plaster off and give it its own name. Half the permission people need is hearing the business say the thing out loud.

An AI function tangles fast if it sits inside the org chart. It gets read as one department’s pet project, or worse, as a threat to another’s headcount. Run from outside, it can move faster, say what nobody internal is politically free to say, and hold no stake in who wins the argument. That’s not a permanent state. Plenty of businesses will eventually want this in-house. But it’s the right way to start, and often the right way to stay.

The part nobody budgets for is people wanting to. You can buy licences for the whole company, run a training morning, publish the policy, and still watch everybody go back to the way that already works for them. Told to use it, most people use it once.

So the first job is getting them interested. We sit with your staff on their own real work rather than a demo, and we stay there until something they do every week takes ten minutes instead of two hours. That moment does the convincing, and it does it better than any strategy slide.

Then we get out of the way. People who have seen it work start finding uses we would never have thought of, and they show each other, which spreads faster than anything we could deliver. Taking a room from wary to curious is the part we have spent years learning. What people do once they are keen tends to be better than the plan.

1. 01AI’s biggest problem is getting people to use it.
2. 02The blockers are people, not tools. Usually it is the gatekeepers who feel threatened by it: someone whose authority rests on being the one who knows, or a team whose headcount looks exposed. Until you deal with that openly, no amount of licences or token spend changes it.
3. 03IT should support the AI function, not gate it. In most businesses today, it’s the other way round.
4. 04An AI department works best run from outside the org chart, at least at first. That’s what keeps it honest.
5. 05Judgement and nuance still come from a person. The tools improve every month; that part doesn’t change.
6. 06Being open about how it’s done costs nothing and proves everything. The tools change monthly, so a secret list of them was never worth much. We would rather hand over what we know and be judged on what we can do with it.
7. 07Enthusiasm beats instruction. Once staff are keen, they take it from there, and they take it further than we would have.

Two ways, and which one fits comes down to the company.

Out of the departments it serves. An AI function makes Finance, Legal, HR, Operations, Sales and Marketing cheaper to run, so it takes a slice of what it saves them. No new budget has to be found, and new budget is always the hardest kind to find. The aim is capacity: the same people getting through more of the work that matters to them.

Or as its own line. Plenty of companies would rather fund it cleanly and skip the internal argument about who contributes what. That works too, and it is usually quicker to get moving.

Either way, sitting outside the building is what makes it workable. We have no headcount to protect and no empire to build, so we can say a department should stop doing something without it turning into a fight over territory. An internal team funded out of everyone else’s budget gets read as a land grab on day one, and then spends its first year defending itself instead of doing the work.

We have sat inside this, not read about it: fifteen years on the business side of large corporates, an investment bank, a global consultancy and a university, watching good ideas stall on nobody’s decision in particular. The pattern repeats. Staff want to use tools they already have licences for. Somebody senior is waiting on a rule that has never been written down. Months pass.

That’s the gap TasksmithAI sits in: an AI department you can start using this month, without building one, hiring for it, or waiting for the internal argument to finish.