Paying for AI seats is the easy part. Getting five busy people to use them on real work, the same way, every week, is the part nobody plans for.
To train a small team to use AI, pick one real job everyone does every week, give everyone the same tool, write down what must never go into it, and practise on that one job together until it is routine. Then add the next job. Courses and long policy documents come later, if at all.
Most owners do the opposite. They buy a few seats, send a link, say "have a play with it", and a month later one person uses it every day, two tried it once, and nobody knows what the others are doing. That is not a training problem. It is a starting point problem, and it is easy to fix in a team of five.
Because "use AI" is not a task. Nobody opens a blank box at nine in the morning with nothing to ask it. People reach for a tool when it saves them from a specific chore they dislike, and until someone names that chore, the tool sits in a browser tab and the subscription renews anyway.
There is usually a second reason, said less often: people are not sure what they are allowed to paste into it. When the rules are unclear, careful staff do nothing, and less careful staff paste in things they should not. Both problems go away once the owner makes two decisions out loud: which job first, and what stays out.
Three things, each one a short conversation, not a project. Get them on one page that everyone can find, and you have done most of the work a formal policy would do.
Whatever the tool drafts, a named person reads it before it reaches a customer. The customer sees your business, not the tool, so the words are still yours. Say this once, clearly, on day one.
Six steps, each one doable in under an hour, spread over two or three weeks. None needs a trainer or a budget meeting. They work best with the whole team in the room, or on one call, for the first two.
Take a real piece of last week's work, say an enquiry email, and do it with the tool on a shared screen. Show the first draft, show what is wrong with it, show how a better instruction fixes it. Twenty minutes is plenty.
Each person does that one job with the tool for a week, on their own real work. Ask them to keep the instruction that gave the best result, word for word.
One document, one heading per job, with the instructions that worked and a line on when to use each. This page is the training. New staff read it on their first day instead of starting from nothing.
Write a short paragraph the tool can be given every time: who you are, the area you cover, how you sign off, words you never use. Drafts stop sounding generic the moment this is pasted in.
Same time each week. What saved time, what went wrong, what got added to the prompt page. Keep it short or it will stop happening.
Only once the first job is routine for everyone. Move to the next most repeated chore and run the same loop. Two or three jobs done well beat ten tried once.
Not necessarily the owner, and not necessarily the most technical person. The best lead is whoever already does the chosen job most often and complains about it most. They have the strongest reason to make it work, and the rest of the team trusts their judgement on that job.
The owner's part is smaller but matters more: picking the job, setting the red lines, protecting the fifteen minutes each week, and using the tool visibly themselves. Staff copy what the owner does, not what the owner says. If the owner never touches it, the team quietly decides it is optional.
When the same job is being done by hand the same way every day, a person pasting the same instruction into the same box, that job is a candidate for an automation that runs on its own. The prompt page you built tells you exactly which jobs those are. Our guide to small business automations covers what that next stage looks like.
That is also the moment a wider plan starts to pay off. If you want to see how one well chosen job leads into the next, our AI strategy guide for small businesses lays out a 90 day plan, and our automation service shows what it looks like when we build the next step for you.
Not to start. A general course teaches what the tools can do in theory, while your team needs to know how to do their own weekly jobs with it. Practising on real work with a shared prompt page teaches more in two weeks. A course can make sense later, once the habits are there.
Ask which part of their week they would most like to get rid of, and start there. Resistance usually comes from being handed a tool with no clear job, or from worry about what it means for their role. Make it clear the tool drafts and the person decides, and let them see a colleague save time on the same task.
It is better if they do not. A business account keeps work in one place you control, lets you remove access when someone leaves, and usually comes with clearer terms on how your data is handled. Read those terms for the plan you choose, and write the answer into your one page of rules.
For one well chosen job, a few weeks of the weekly check is usually enough for it to become the normal way that job gets done. The sign is when people add to the prompt page without being asked. If that has not happened, the first job was probably too rare or too vague, so pick a more frequent one. If a more frequent job does not stick either, cancel the seats rather than push harder.
When a customer searches nearby and books the competitor with a faster site and instant replies, it isn't because they were better. It's because they were easier to find and faster to answer. That's fixable, and it starts with a 15-minute chat that costs nothing.