Role-based AI training builds each session around the work one team already does, and it is the version that sticks. Generic AI training teaches the tool to everyone at once, then leaves each person to find where it fits their own job. That mapping step is the hardest part of adoption, and it is the exact part a generic course hands to the audience on the way out the door. A company can buy awareness for everyone in a single session. The capability that reshapes daily work gets built one team at a time.
Why does a company-wide AI lunch-and-learn wear off in a month?
It usually goes like this. A capable speaker, a strong demo reel, questions from the floor, and a few people open accounts on the spot. Then the work piles back up: the queue returns, the chatbot tab goes cold, and a month later the only trace of the session is the calendar invite.
The easy reading says the team lacked follow-through, or the tool was overhyped. We read it differently. The session taught what the tool can do in general, then assigned everyone the same homework: figure out what it can do for your job in particular. That translation takes reps, and nobody in the room was given any. The busiest people were left to manage it alone, on top of the job itself, and that is how most of these sessions quietly come apart.
Can one AI course serve marketing, ops, and finance at once?
The useful version of AI is different in each of those rooms, and the differences are the whole curriculum.
Marketing runs on voice. The valuable work there is a first draft in the brand’s register, or one brief spun into a dozen variants, and the standing fear is off-voice copy getting out with the company’s name on it, so a marketing track spends serious time on review and reuse.
Ops carries a different worry: a confident wrong answer slipping into a workflow nobody rechecks. The valuable work there is getting the SOP out of one veteran’s head into a first draft and taking routine steps off people’s plates, so an ops track leads with verification habits before it automates anything. Finance comes down to controls. The boundary that data must never cross decides everything, so a finance track starts with what is even allowed into a prompt.
In one shared room the course can only teach the intersection, and the intersection is the tool itself: what a context window is and why the model sometimes invents things. That hour is useful, but it attaches to nobody’s particular work and fades fast. Every example lands with a third of the room while the other two thirds check email, and the guardrails one team needs read as someone else’s paperwork to the rest.
Mixing functions also forces the course to teach to the middle, which serves no one well. The self-taught power user is bored, and the skeptic hears nothing specific enough to change their mind.
There is a reason the generic course keeps getting booked anyway. It is easy to buy, and one date with one vendor gives a leadership team a line item they can point to later.
What does role-based training look like in practice?
It starts before the session, with the manager. We ask for a short call and a look at the team’s recurring work, in whatever shape that work takes. Four things drive the curriculum:
- the deliverables that repeat, and the ones that always run late
- the inputs those deliverables start from, in whatever messy form they arrive
- the rules the team already follows, from approval chains to data boundaries
- the tasks the manager most wants off the team’s plate, where the aim is to remove the work and keep the people
The session then works that list, one track per team, each drawn from the team’s own work. Finance shows the shape of it most clearly. A reviewer tears apart a variance-commentary draft during the session, because review belongs inside the routine, where problems get caught as they form. The other tracks run the same way on their own material: an ops track on the tangled threads people there already untangle by hand, and a marketing track on the briefs the team is handed. Every prompt gets written against the team’s own inputs, which is why they hold up when the next week of work arrives. Nobody has to carry the training home and apply it later, because the applying already happened in the room, with the manager present to settle guardrail questions as they came up.
That format is our role-based AI workshops service. The scope is fixed, and the tracks come from the work your people already do. You get the numbers on the first call. Workshops can also add 30 days of support afterward as a paid add-on, because the sharpest questions tend to arrive once people are back at their desks and a task in front of them refuses to match the pattern from the room.
How does a role-based track differ from a generic course?
It is fair to ask whether this is only the generic course with fresh examples. In tool training the examples are much of the curriculum, so swapping them already changes what people can do once the session ends. Even so, the examples are the smallest of the differences. Structurally, a generic course is organized around the tool’s features, while a role-based track organizes around the team’s deliverables. Guardrails shift too, since a rule that is existential in finance is clutter in marketing, and with the manager in the room those calls get made on the spot. The last difference is the one that counts most. A generic course ends when the slides run out. A track keeps going until every task the team named up front has a working routine attached to it.
How do you know it worked?
You picked the tasks up front, so you check the tasks afterward. That’s the entire trick, and generic training cannot perform it. A session about AI in general leaves two things to measure, attendance and how people felt walking out. Both were probably excellent at the lunch-and-learn, and neither one pays for anything.
A role-based track leaves behind a short list of named routines per team. A few weeks later a manager can put a direct question to each one: still in use, or quietly abandoned. A routine that got dropped is worth knowing about, because it points straight at the next hour of training. Measuring ROI gets much easier once each team has something this concrete to check against.
The manager also stops being a spectator here. Generic training is something that happens to a team and then recedes. In a role-based track the manager helped pick the tasks and sat in on the guardrail calls, so once the routines are running, the follow-up questions belong to them.
Is generic training ever the right call?
There is one job a single wide-audience session does well, which is building shared vocabulary. It can settle a leadership group on direction and surface the people who were already experimenting quietly. We give talks like that ourselves and enjoy them. The trouble only starts when a talk is asked to do a workshop’s job, since an hour of awareness has never installed a habit on its own.
So the order matters. You want the wide awareness session first, in one room, and the per-team tracks after it, once people know enough to ask pointed questions. The lunch-and-learn that burned you was most likely a decent talk wearing a training badge, and the part that went missing is the part nobody scheduled.
That missing part is what we build. The first call runs 30 minutes and costs nothing. Come with the tasks your teams keep repeating, and we will map out what the tracks would cover.