The scenario we see most often: a company buys a dozen ChatGPT licences, sends a “we use AI now” email, and three months later a handful of people use them — mostly to polish emails. The investment fails not because the tool is weak, but because there was no plan.
Start with a task audit, not the tool. List the activities that consume the most time and repeat weekly: reports, proposals, similar email replies, meeting notes. These are your first AI use cases — specific, measurable and visible to everyone.
Step two is ground rules: what may be pasted into ChatGPT and what may not, how to verify answers, who owns the result. A simple one-page AI policy removes the biggest adoption blocker — employees’ fear of doing something wrong.
Step three: training on your team’s real tasks, not generic examples. A salesperson should leave with prompts for their proposals, an accountant with prompts for their reports. That is when AI stops being a curiosity and becomes a work tool.
Finally, measure: time saved per task, weekly usage, output quality. What gets measured grows — and management sees a return on investment instead of another IT cost.