A small business AI rollout usually follows the same curve. Week one, you buy ten seats and everyone logs in. Week two, three people are genuinely interested and show the rest what they found. Week four, it comes up less in meetings. By week six, two people still use it daily, everyone else works the way they did before, and you are paying around $300 a month for software two people open.
Nothing broke. The tool works fine. The rollout had no job attached to it.
Handing out licenses is not a rollout
Most companies start by buying seats. Microsoft 365 Copilot runs about $30 per user per month. Business tiers of ChatGPT and Claude land in the same $25 to $30 range. Ten seats is roughly $3,600 a year, cheap enough that nobody builds a plan around it. The licenses arrive, staff get a link, and the instruction is some version of "have a go with it."
People who already had spare time and curiosity will find a use. Everyone else has a full day. A blank chat box at 9am competes with forty unread emails, and the emails win.
Before you buy a single seat, write down the task you expect it to do. If you cannot name one, you are not ready to buy yet.
Week six is when the novelty runs out
The first few weeks run on interest. Someone drafts a difficult email, someone summarizes a forty page PDF, and it feels fast. None of that is tied to a deadline, so none of it survives a busy week. Around week five or six the first real crunch arrives, people fall back on the method they trust, and the habit does not come back on its own.
Anything that happens only when someone remembers to do it will stop happening. Attach the tool to work that already has a due date and an owner.
The fix: one recurring task, one owner, one number
Pick a single task that happens every week whether anyone feels like it or not. Monthly client reports. Tuesday quote follow ups. The Friday staff schedule. Summaries of the week's intake calls. Something with a deadline somebody is already responsible for.
Then do three things before the first login.
- Name the owner. One person, by name, responsible for that task being done with the tool for the next 30 days. Not a committee, and not "the team."
- Write the prompt once and save it. Staff will not invent a good prompt under time pressure. One person writes it, tests it on three real examples from last month, and saves it somewhere everyone can reach in two clicks. That is about an hour of work, and it decides whether the rollout lasts.
- Write down today's number. Hours the report takes now. Days between a quote going out and the follow up. Get the real figure first, even if it is an estimate from the person who does the job.
Put the owner, the task, the saved prompt, and the starting number on one page and send it to the whole team on day one.
What this looks like in practice
Say a ten person firm spends four hours a month assembling client reports from notes and spreadsheets. That becomes the task. The office manager owns it. Someone writes and tests the prompt that turns raw notes into the firm's standard report format, then saves it in the shared drive. Four hours is the number to beat.
At the end of 30 days you know whether it took two hours or still took four, and you know what the remaining time is spent on. That is a decision you can make. "The team seems to like it" is not.
Write that paragraph for your own business this week. If you cannot fill in the number, measure it first.
Book the six week check now
On the day you buy the licenses, put a 30 minute meeting on the calendar six weeks out. Three questions: did the task get done with the tool every time it came up, did the number move, and what got in the way. The answers are usually small and practical. The prompt needed the client name formatted a certain way. The output needed one extra paragraph of human editing.
If the number moved, add a second task and a second owner, and run the same 30 days again. If it did not move, either the task was wrong or the tool was wrong, and both are fixable for the cost of one more month.
Know when to stop
If you have run two 30 day cycles on two different tasks and nothing measurable changed, cancel the seats. Keep one or two for the people who do get value from them. Small businesses can stop cheaply, which is a real advantage over a company locked into a three year enterprise agreement.
If you are about to buy seats, or you are at week six wondering why the enthusiasm faded, the first call is free. Tell us how many people you have and which recurring task is eating the most time, and we will tell you what we would attach the tool to first, including the cases where the answer is to leave it alone. Get in touch and we will find a time.