Role-based AI training beats all-hands demos: a practical playbook for enterprise cohorts

The all-hands is easy to schedule. One hour, one demo, every function on the same call, and a recording for the people who could not attend. The chat fills with thanks. The recording gets twelve views. On Monday the finance manager and the customer lead are staring at the same generic prompt, and neither of them can see their job in it.

A demo is a fine way to show that a tool exists. It is a poor way to change how a role works. Enterprise cohorts are large enough that one session feels efficient, and varied enough that one session teaches nobody's actual task. Role based training is more work to design. It is the version that survives the week after.

Why the all-hands fades

People remember what they tried, not what they were shown. In an all-hands they try nothing, or they try a party trick. The examples are chosen to be impressive to a mixed room, which means they are specific to no one. Risk is waved at in a final slide, because the risks are not the same for a recruiter and a claims handler.

There is also a status problem. Senior people attend to be seen attending. Junior people hesitate to ask a basic question in front of the whole company. The recording then becomes the programme, and a recording cannot sit with someone while they try the task and tell them whether the output is fit to send.

Pick roles, not a crowd

Start with three to five roles where the work is frequent, the data rules are knowable, and a manager cares about the output. A mid-market firm does not need a pathway for every job title in the HR system. It needs a pathway for the jobs that touch customers, money, code, documents or decisions.

Write one page per role before you book a room. Say what they produce in a week, what a good one looks like, what they must not paste into a tool, and which steps are draft, decide or send. If you cannot fill that page, you are not ready to train the role. An all-hands will still be easy to schedule, and it will still leave each person without a task they recognise.

A cohort, not a class

A useful cohort is an intact group: the people who do the role, plus the manager who will see the work, in a group small enough that everyone builds something. Eight to twenty is a sane range. Larger than that and you are back to a demo with breakout rooms that nobody visits.

Run it as practice. Bring last week's real task, with sensitive data swapped for a stand in if you must. Each person produces an output they could defend. The manager says whether they would send it. The group names the check that was missing. That is the session. Slides are there to set the data rule and the shape of the hour, not to carry it.

The AI training workshop is built this way on purpose. Vendor neutral, on the team's work, with a short list of use cases for the role rather than a tour of everything the product can do.

The playbook

1. Choose the role with a sponsor who will look at the outputs two weeks later, or do not start.
2. Write the one page brief: products of the week, the quality bar, the data that is in bounds and the step a human still owns.
3. Collect five real tasks from the cohort, stripped of anything the policy forbids, so the exercises are not invented by the facilitator.
4. Teach the smallest set of moves that role needs, and ignore the features that belong to a different job.
5. Make every person ship one output in the room, and make the manager mark it against the quality bar before they leave.
6. Agree one workflow they will run for ten working days, with an owner and a place the outputs live.
7. Meet again for an hour. Keep the workflow, change the check, or stop. Write the decision where the next cohort can see it.

The seventh step is how you stop collecting pilots. A workflow that is still a pilot after ten days is either not real or not owned. Both are expensive if you keep them on a slide.

What to do with the all-hands

You can keep a short all-hands for the story: which roles are in motion, what is allowed, what is not and where to ask for a cohort. Twenty minutes, led by someone who has done the work, with no live prompt theatre. Point people to the role sessions. Do not pretend the all-hands is the training.

Sequence the cohorts so they teach you. The second role should be easier to brief than the first, because you learned what the one pager missed. Share the data rules across roles so you do not invent a private policy per team. Share the workflows only when the work is actually similar. A lawyer and a marketer do not need the same exercise just because they both use words.

Measure by role, rather than with a company wide confidence score. For each cohort, ask whether the named workflow ran for ten days, whether a manager rejected at least one output, and whether there is a sentence that says what happens next. If you cannot answer those, you have a demo. The cohort was supposed to leave a workflow someone owns.