How to roll out AI to your team
The standard AI rollout is: buy licences, send an announcement with a link, wait. Six weeks later two people are using it constantly, everyone else tried it once, and someone concludes the tool was overhyped.
The tool was fine. The rollout asked people to invent their own use case, which is a much harder task than it sounds when you're already busy.
Start with tasks, not tools
"Here's an AI assistant" gives someone a blank page. "Here's how to turn your site notes into a client report in ten minutes instead of an hour" gives them a reason.
Pick two or three tasks that are:
- Frequent — done weekly at least, so the habit has room to form
- Disliked — the tedious ones convert people fastest
- Low-stakes — internal first, so early mistakes are cheap
- Text-shaped — writing, summarising, reformatting, drafting
Start deliberately narrow. Breadth comes on its own once people can see the point.
Protect the practice time
This is where most rollouts quietly fail. Learning a new way of working while fully loaded means defaulting to the old way every time, because the old way is reliable and the deadline is real.
An hour a week for the first month is usually enough — but it has to be genuinely protected, not "fit it in where you can". If it isn't in the diary, it doesn't exist.
An Australian small-business example
A Gold Coast property group buys licences for eighteen staff and runs one training session. Two months later, usage is three people.
They restart differently. One team, one task: turning inspection notes into tenant reports. One person who does that job works out a good approach and writes down the prompt. Thirty minutes a week is set aside, in the diary.
Within a month the whole team uses it for that one task, because it removes the part of the job they liked least. Within three, they've extended it to two more tasks themselves — without being asked, which is the signal you're looking for.
Make good prompts shared property
The single biggest waste in a rollout is ten people separately discovering the same thing. When someone gets a good result, the prompt should go somewhere everyone can find it — a shared document is enough.
Better still, promote the recurring ones into a system prompt so nobody has to remember them at all.
What to expect, week by week
| Weeks | What it looks like |
|---|---|
| 1–2 | Awkward. Slower than the old way. Some people conclude it doesn't work. |
| 3–4 | First real wins. One or two people become internal advocates. |
| 5–8 | Habit forms for the target tasks. People start asking about others. |
| 9–12 | Extension happens without prompting. This is the point it's stuck. |
Knowing weeks one and two look like failure is most of what stops a rollout being cancelled during them.
The three things that decide it
- A specific first task, not a general tool
- Protected time, in the diary, for the first month
- Someone who does the work leading it, not only sponsoring it
Get those right and the licences almost look after themselves. Get them wrong and no amount of budget or enthusiasm at the top will move the middle.
Frequently asked questions
How long before a team is genuinely using it?
What about staff who don't want to?
Should we train everyone at once?
Who should lead it?
What if people use it badly?
Put this to work
Ad On Group runs AI training and enablement for Australian teams through Ad On AI — a three-month, self-paced program that takes non-technical staff from their first prompts to working AI agents.