We have AI licences. Why is everyone still doing it differently?

I keep seeing a familiar picture. One seller has a prompt they've refined for months. Another opens Copilot or ChatGPT and types whatever comes to mind. Someone else tried it twice, got rubbish back and gave up. Marketing has built an agent nobody remembers how to find. You've paid for access, but the way people work is still entirely down to the individual.

That's not a criticism of the team. If you give curious people a tool and no agreed method, they'll experiment. Some will get excellent results and others won't. The commercial problem starts when nobody captures what works, nobody owns the source information and nobody can explain what the licences have improved.

Access gets your team to the start line.

The licence itself is rarely the missing piece. What usually needs attention is the task, the information behind it and the standard you expect. Without those, even an impressive assistant can confidently produce the wrong thing.

Ask two people to prepare for the same type of customer meeting. Are they using current case studies? The right pricing? The same product claims? Do they know what a useful brief should contain? If the answers vary wildly, the inconsistency existed before AI. You've simply given it another place to appear.

Start by looking at what people actually do. Ask them to share their screen and walk through a genuine task. Find the prompts they've kept, the information they paste in and the bits they still correct manually. You'll learn more from that than another presentation about what the software could do.

1. 01Business information. Your ideal customer profile, positioning, product details, objection responses and rules about what can and cannot be said, kept in one approved place that the tool can reach and someone keeps up to date.
2. 02A process for each task. Not 'use AI for prep', but a defined workflow: what goes in, what comes out, where the output is saved, who checks it and what happens next. Written down, so a new starter can follow it on day one.
3. 03A standard for good. If you cannot describe what a good meeting brief or follow-up email looks like, you cannot tell whether the AI produced one, and you cannot coach anyone towards it either.

Choose one common job and make it work properly. Meeting preparation is often a sensible candidate. Agree the inputs, the approved information, the output format, the checks and who maintains it. Test it with the people who will use it before you call it a standard.

Now you've got something useful to share. The person who built the clever prompt gets credit for helping improve the team's method. The new starter has somewhere sensible to begin. And you can finally compare the work before and after.

Don't mistake launch day for the finish line. Products, pricing, case studies and positioning change. If nobody owns the information and the workflow, it will quietly become another out-of-date asset.

Keep room for experimentation, too. People will discover better approaches. Give them a route to suggest changes, test them and update the shared version. Governance should make good ideas easier to spread, not make everybody afraid to try.

At renewal time, I'd want to see which workflows are used, how much rework they avoid and whether the outputs meet the standard. Counting licences and logins won't tell you whether the business is getting value.

Measure adoption on real work, not logins. Are people using the shared workflow on actual accounts? Does the output meet the standard you agreed, when a manager checks a sample? Has the time for the task come down compared with the baseline you took before you started? Three honest answers to those questions tell you more than any usage dashboard, and they give you something concrete to say at renewal time.