Why your AI licences are gathering dust, and a 30-day plan to fix it
You rolled out Copilot or ChatGPT Enterprise. A few enthusiasts love it, most people opened it twice. The problem is rarely the tool. Here is what usually goes wrong and how to turn it around in a month.
The pattern is familiar. Leadership approves an AI rollout. Licences are assigned, an announcement goes out, maybe there is a lunchtime demo. Three months later, the usage dashboard shows a handful of heavy users and a long tail of people who logged in once.
The tempting conclusion is that the tool is not good enough, or that staff are resistant to change. Usually neither is true. The rollout treated AI like a software upgrade, when it is closer to a new way of working that people need to learn.
Five reasons adoption stalls
1. Nobody showed people their own use case
A demo of AI writing a poem or summarising a generic article is impressive and useless. A claims handler wants to see it draft a response to a claim like the ones on their desk. A project manager wants to see it turn last week’s meeting notes into a status update in the client’s format. Until people see AI do their work, it stays a curiosity.
2. People do not know what they are allowed to do
If staff are unsure whether they can paste in a client email, they will either avoid the tool or use it quietly and badly. Ambiguity kills adoption in regulated and client-facing businesses, which describes most firms in Luxembourg’s financial sector and many UK professional services firms. A short, clear policy does more for usage than another launch email.
3. The first experience was disappointing
Someone types a vague one-line request, gets a vague generic answer, and concludes the tool is overhyped. Good results depend on giving context: who the output is for, what it should look like, what source material to use. That is a skill, and it can be taught in an hour or two.
4. Managers are not using it
Teams copy what their managers do. If the head of department never mentions AI, never shares a useful prompt, and still asks for things to be done the old way, the signal is clear: this is optional.
5. There is no time to learn
Staff are asked to “experiment” on top of a full workload. Experiments lose to deadlines every time. Adoption needs protected time, even a little of it.
A 30-day turnaround plan
This plan assumes you already have the tool in place. It works for a single team or a whole organisation of up to a few hundred people.
Week 1: find the real use cases
- Pick two or three teams with repetitive writing, summarising or analysis work.
- Spend 30 minutes with each team lead listing the tasks that eat the most time.
- Choose three tasks per team where AI could plausibly save time and where mistakes are easy to catch.
Week 2: train on those tasks
- Run a hands-on session per team, built around their chosen tasks, using their own (non-sensitive) material.
- Teach the basics of a good prompt: role, context, task, format, and examples.
- Show where the tool fails on their work, so trust is calibrated rather than blind.
- Publish your AI policy, or a one-page interim version, before the session. People ask “am I allowed to?” within the first ten minutes.
Week 3: build a shared prompt library
- Ask each team to save the prompts that worked into a shared document or channel.
- Nominate one “AI champion” per team: not necessarily the most technical person, but someone others ask for help.
- Have managers share one real example of how they used AI that week.
Week 4: measure and decide
- Compare usage before and after. Ask people which tasks now take less time and by roughly how much.
- Collect the tasks where AI did not help. These are just as valuable: they tell you where not to push.
- Decide what to scale. Often the answer is a second wave of training for other teams, and in some cases a custom agent for one high-volume task that keeps coming up.
What good looks like after a month
You are not aiming for universal enthusiasm. A healthy result after 30 days usually looks like this:
- Most people in the pilot teams use AI at least weekly for two or three specific tasks.
- Everyone knows what they can and cannot put into the tool.
- There is a shared set of prompts that new joiners can pick up.
- You have a short list of the next opportunities, ranked by time saved.
That last list is where the real return starts. Licences are a cost. Habits are what turn them into time back.