The short answer.

Train your people on the workflows their own jobs run, not on the AI assistant in general: a short set of ground rules for everyone once, then hands-on sessions for each role on its own past work, led by the person who owns each workflow. Leave a one-sheet job aid on every desk that says which workflow to reach for, how to start it and what the person still checks, and print it again whenever a workflow changes. Train managers first so they can answer questions, tell everyone plainly why the change is happening, and follow adoption by team and by workflow, never by scoring individuals. Training sticks because the work repeats the lesson every week, and the sheet is still there on the day the session is forgotten.

The articles behind this guide each settle one part of that answer. Read together, they show something none of them says on its own: every part reads from the same written record, a workflow attached to a role on the company’s plan, with one owner. The written method is the curriculum, the job aid is its shortest form, the new hire’s first week is a walk through it, and the manager’s standard lives in it. Learning and development for AI is not a programme beside the plan. It is the plan, delivered to people.

Training belongs to the third of the three steps set out in How organizations plan their use of AI: plan it, develop the workflows, deploy them. A workflow nobody runs returns nothing, which is why training is where 10X the ROI of your existing AI is won or lost.

General AI training fades. Training by role, on the job’s own workflows, holds.

General AI training fades because nothing in the trainee’s week asks for it again, and training by role holds because the workflow arrives with the work. A tour of ChatGPT, Claude, Copilot or Gemini teaches the same lessons to a scheduler, a bookkeeper and a sales lead, so nobody leaves the room knowing which piece of their own week has changed. On Monday the queue looks as it did, and the old way wins.

Training that holds has two layers.

  • Ground rules, for everyone, once.Which assistant and account, what goes in and what stays out, that a workflow prepares and a person decides, where the workflows live, and how to report a miss. The owner of the AI plan teaches them from one page.
  • Role sessions, on the job’s own workflows.Each role learns the two or three workflows it runs, hands on, on last month’s real cases, comparing each draft with what the team actually sent. Each workflow’s owner teaches it, because the owner can say whether a draft is right.

How do we train our team on AI? sets out both layers and a role training sheet. Why AI training fails when it teaches the tool and not the job covers what happens after the session.

The test for any training is one question: on which day next week will this person use it, and on what piece of work? A role session answers it before it starts.

Freedom to explore stays, and exploring was the right way to start. What it cannot supply is the piece of work to hand over, the standard for it and a person who checks it; those are the company’s to give. Is it OK to let employees figure out AI themselves? has a page to send the team.

The curriculum comes from the plan. In FullyPowered AI Planner, each team’s roles sit beside the workflows they run, each with its owner, so the learning path for a job is already written down: these workflows, for this role, in this order, taught by these people.

We offer training of this kind as FullyPowered AI Learning and Development, in four parts: hands-on training, where a team runs its own workflows on its own work; a curriculum built for the company, by team and by job, from its plan in the Planner; a job aid for every job; and change support. It is scoped to your company.

A job aid on every desk, and three habits for when it runs out.

A job aid is one sheet for one job, letter size and folded in half, that lists the workflows the job uses, when to reach for each, how to start it in the company’s assistant and what the person still checks and decides. It stands by the keyboard or goes in a bag, laminated if someone wants it to last.

The sheet is made for the person fifteen years into a finance job, who knows the work better than any trainer. That person does not need a quiet week of study, and the sheet does not teach them their job. It puts the new way within reach, in their own words: if I want to run this report, this is how I do it. A course asks a person to remember; a job aid asks them to look.

A job aid is also a file, printed again by the workflow’s owner whenever the workflow changes. The job aid is the shortest form of the workflow’s written method, so a company that writes its workflows down has already written much of every sheet.

A one-sheet job aid to copy. Fold a letter-size page in half and fill in the four panels for one job. Keep it to what fits.

  1. Front: the job.Role and team, named for the work: ______. The line “The assistant prepares. You check, and you decide what goes out.” Printed on: ______. The one person to ask: ______
  2. Inside left: the workflows this job uses.For each one, by name: when to reach for it (“when the month closes”, “when a new order arrives”): ______. What to have ready: ______. How to start it in the company’s assistant: ______. What you check and decide: ______
  3. Inside right: the rest of the job’s workflows.The same four lines for each, three to five workflows in all on the two inside panels: ______
  4. Back: when something goes wrong.What went wrong: compare what you gave it with what came back: ______. Stuck: what to try once more, then whom to ask: ______. Where to look: the workflow’s written steps, at: ______
  5. Back: who owns each workflow.One name per workflow: ______
  6. Back: the latest version.Where the current sheet is kept, and the line “Printed again whenever a workflow changes”: ______

The back panel carries three habits that hands-on training teaches alongside the steps, because no sheet covers every case.

  • Working out what went wrong.When a draft is off, look at what the workflow started from before blaming the workflow. A wrong export, a missing file or last month’s date is often the reason.
  • Keeping going when stuck.Try once more with the step spelled out, then ask the owner. Being stuck is part of learning the work, and nothing goes out on a guess.
  • Knowing where to look.The workflow’s owner, its written steps in the plan, the team’s shared place. A person who knows where answers live does not have to carry them all.

Where does each person start, and how does a team move up?

Each person starts somewhere on a ladder of AI use, and a team moves up when the method of the person who has climbed becomes a workflow everyone in the role runs. Using an assistant at work has six levels: asking it questions, giving it your files, setting standing instructions, running reusable workflows as a team, connecting those workflows to your systems on a schedule, and having several assistants work together. The levels of using an AI assistant at work describes each level and what it takes to move up one.

Few teams start at zero. Picture fluency on a scale: some people go from 0 to 60, one colleague is at 25 and could be at 60, another is at 40 and could be at 80. The numbers are a picture, not a measured scale: good training meets each person where they are and moves them up through their own work.

A fluent answer from the chat box feels like knowing AI, which is how a company stays at the surface. Everyone can chat. That is not the same as knowing AI. sets out what sits past it, and Basic Excel and advanced Excel: the same gap exists with AI gives the comparison every leader recognises. The higher levels are written in plain language, so the gap is easier to close than the spreadsheet one, and the hard part, knowing the work, is the part your people already have. Why companies get so little from the AI they already pay for follows the same gap across the whole company.

Here is what joins the levels to training: a team does not need every person to become a builder. One well-written workflow, owned by one person, moves the whole role to level four on the day it is deployed, and the role session plus the job aid is how each person arrives there.

What do we tell employees, and how are managers equipped?

Tell employees four things, once, from the person who runs the company: what is decided, what each team is asked to do, what stays the same, and where to bring ideas; then have each manager say it again to their own team that week, with the answers to the questions their people will ask. The right day is the day the plan names a first team, a first workflow and its owner. What do we tell employees about AI? has the script to read.

Unease about AI usually comes from three open questions each person carries alone: am I allowed to use it for this, is my way the right way, and what does it mean for my role. A page per team answers all three in writing, and the unease eases once people can read the answers. Why is my team uneasy about AI, and what helps? has that page.

Change support carries those answers to people; an HR reader knows the wider discipline as change management. It does four things: says plainly why the change is happening and why it matters, to the people it affects; equips managers to answer questions; gives people a place for concerns to be heard; and follows adoption by team and by workflow, so help goes where a team is stuck. It asks three questions of each team: do people understand the change, do they know why it matters, are they taking it up.

Managers are trained first, because after the sessions end the manager is the person people ask.

A manager’s briefing sheet to copy. Fill it in for one team before the announcement, and keep it to hand for the questions that follow.

  1. Why we are changing.In two sentences, in the team’s own terms: ______
  2. What stays the same.Who signs off on each piece of work, and the standards the team keeps: ______
  3. “Am I allowed to use it for this?”What goes into the assistant, and what stays out: ______
  4. “Is my way the right way?”The shared workflow, and who checks its output: ______
  5. “What does it mean for my role?”For each role: what the workflow prepares, what the person decides: ______
  6. “Am I being watched or rated?”Nobody is scored on AI use, and nothing records screens or keystrokes.
  7. Where concerns go, and when training happens.One person, by name; the role sessions and who leads each: ______

How AI changes a manager’s job.

AI makes a manager the owner of how the team’s work is done, in writing: the manager decides which pieces of the team’s work AI prepares, keeps the team’s method current as the work changes, and looks at prepared work and at how people check it, more than at who looks busy. Setting the standard, developing people and deciding the hard cases stay exactly where they were. How does AI change a manager’s job? has a weekly checklist that replaces the status round.

A manager’s changed job is also a training job, and the articles meet here. The exception log, where people note each case they corrected, is the manager’s best teaching material: a correction that appears twice is a gap in the written method, and a case the workflow flagged as unclear is a lesson to walk through with the person who checked it. Each fix to the method reaches the job aid at its next printing. So the manager who reads the log on Monday is running the company’s training without calling it that, and coaching moves up to judgment: the exception, the difficult client, the case the method does not cover.

How do new hires learn the workflows their role runs?

New hires learn the workflows their role runs by running them, with one named colleague checking how they checked: on the first day they get the team’s method in writing, a draft the workflow has prepared on a live case, and the job aid for the role. The plan already lists which workflows come with the role, so the manager does not rebuild that list from memory. Review steps down on purpose, from every piece, to a sample, to the exceptions, as the new person’s checking matches the team’s standard. How do new hires learn the AI workflows their role runs? has a first-week sheet.

A new hire learns one way to do a job only if the company has one. When five colleagues each prompt an assistant their own way, the new person inherits five habits and none of the reasons. The fix is to collect what people actually run, choose one version for each piece of work, build it as a workflow and attach it to the role with an owner. How do we share what works with AI across the company? has the collection form. The rules a new hire’s questions bring to light go into the workflow and onto the job aid, so the next hire gets them too.

How do we bring a skeptical team along?

A skeptical team comes along when the ask stops being vague: start from the work the team itself names as most tedious, let the team write what good output looks like, run the workflow on cases the team has already finished, and make the team’s lead the owner with the final say on going live. Skepticism is usually a reasonable answer to a vague “use AI more”. How do we bring a skeptical team along with AI? has the questions to work through in one meeting.

A willing team that has not changed its habits needs the same thing from the other side: a named piece of its own work, done the way its best colleague does it, with the manager running it first. How do we get people to actually use AI at work? has six questions to ask before anything is built.

For training, make the most careful skeptic the checker and a teacher. The rules a skeptic writes become the “what you check” lines on the job aid, the cases a skeptic doubts become the past cases the role session trains on, and the misses a skeptic catches become fixes in the next version. The person fifteen years into the job is often that skeptic, and is often right; their standard is exactly what the workflow and the sheet should carry. How to roll out AI across a company, and tell whether it is working sets the same move inside the wider rollout.

Follow adoption by team and by workflow, never by person.

You follow adoption by asking, for each team, which of its workflows are running and where it has stalled, so help goes where a team is stuck; nobody is scored on how fast they took to it. A team that is stuck on one workflow gets another session on that workflow, a clearer job aid or a fix to the method, not a mark against anyone. Installed, used or useful? gives the three questions to ask in order.

People who know they are not ranked report misses freely, and a reported miss is how a workflow and its sheet improve.

A team adoption sheet to copy. One sheet per team, reviewed each month by the team’s manager and the owner of the AI plan.

  1. Team and manager.______
  2. Workflows on the plan for this team.Each by name, with its owner: ______
  3. For each workflow: running?In daily use, used some of the time, or not yet: ______
  4. Do people understand the change, and why it matters?What the team asked this month, and what is still unclear: ______
  5. Where has the team stalled?The workflow, and the reason as the team gives it: ______
  6. Help sent.Another session, a reprinted job aid, a change to the method, a word from the manager: ______
  7. Job aids current?Date each was last printed: ______
  8. What the gained time goes to.Named work, decided with the team: ______

The last line matters more than it looks. When people see the time a workflow gives back go to work they helped choose, they ask for the next workflow themselves. What will your team do with the capacity it gains? has the worksheet. A monthly review keeps the sheet current; How to run a monthly AI review gives the agenda.

One company trains its teams, from the plan to the desk.

Illustrative example.

Take an 85-person regional bakery with four office teams beside its production floor: wholesale orders, production planning, finance, and people and office. The operations director owns the AI plan, and the people lead owns training. The plan in the Planner gives wholesale orders two workflows to start: confirming each café and grocery order against the standing order and the delivery schedule, and drafting the reply when a customer changes an order after the cut-off.

The order desk lead has taken the bakery’s orders for many years and knows every account’s habits. She doubted an assistant could tell a standing order from a one-off, so she wrote that rule first, and her hard cases became the past orders the role session trains on. She owns the order confirmation workflow and teaches it.

The operations director makes the four-part announcement at the Monday meeting. The managers were trained first and each filled in a briefing sheet, so the wholesale manager answers the order desk’s questions that afternoon. Everyone attends the ground rules session together. Only wholesale orders has role sessions, run on last month’s real orders; finance and production planning wait, with their workflows on the plan for later.

Each person on the order desk gets a job aid for their own job. The order desk lead’s inside left panel lists the order confirmation first, with “when an order arrives” beside it and “standing order and delivery schedule” under what to have ready. The order confirmation started from a free workflow template in the Workflows Directory, adapted to the bakery’s own accounts and schedule before anyone trained on it.

The people lead and the wholesale manager fill in the team adoption sheet each month. When the cut-off reply workflow stalls because nobody is sure which customers may be offered a later delivery, the team gets a second session on that one workflow, the rule goes into the method, and the job aid is printed again.

What is different in a company of under 200 people.

In a company of under 200 people, the trainer is a colleague, not a department, and that makes training faster and closer to the work. Published advice tends to assume a training team and a course library; you have a lead for each team who already knows the work.

  • One session covers the ground rules.The whole company, or the whole office, learns them together from the person who owns the plan.
  • Train as each workflow is ready.A team with no workflow yet learns the ground rules and waits, rather than sitting through training it will not use.
  • Change support is said in person.The leader reads each team its page and hears its questions the same week, and a promise that nobody is rated holds because everyone knows who made it.
  • One role can be one person.Attach the workflow and the job aid to the role anyway, so both survive a holiday or a resignation.

Start from the plan, and the training, the job aids and the change support follow from it. The Planner is where the plan lives: you set out your teams first, then their roles and the work between them, choose what to build first, and your colleagues add their own teams, so every role’s learning path starts from the workflows on the plan.

Questions people ask.

For short answers across the whole subject of planning AI, Answers takes the questions leaders ask in order.

Is a general AI course for the whole company worth it?

A general course teaches the assistant, and the lesson fades because nothing in a person’s week asks for it again. Keep the general part short: one session on the ground rules, for everyone, once. Spend the training on role sessions, where each person runs the workflows their own job uses on their own past work.

What is a job aid for AI workflows?

A job aid is one sheet for one job, letter size and folded in half, that sits on the desk or goes in a bag. It lists the workflows the job uses, when to reach for each, what to have ready, how to start it in the company’s assistant and what the person still checks. It is printed again whenever a workflow changes, so it is always current.

How do we train people who have done the job for years?

Start from the expertise they have, not from the tool. Train them on the workflows their own job runs, on their own past cases, and give them a job aid that says which workflow to reach for and how to start it. Their standard is what the workflow should carry, so ask them to write the checks.

What do we tell employees about AI?

Tell them four things, once, from the person who runs the company: what is decided, what each team is asked to do, what stays the same, and where to bring ideas. Say it on the day the plan names a first team, a first workflow and its owner. Each manager then repeats it to their own team that week and answers the team’s questions.

How do new hires learn the AI workflows their role runs?

They run them from the first day, with one named colleague reviewing how they check each draft. The company’s plan lists the workflows that come with the role, and the job aid for the role is on their desk. Review steps down from every piece, to a sample, to the exceptions, as their checking matches the team’s standard.

What do we do with a team that is skeptical of AI?

Make the ask specific. Let the team choose its most tedious repeating work, write what good output looks like and test the workflow on cases it has already finished, with its lead deciding when it goes live. Put the most careful skeptic in charge of checking, because the misses they catch make the workflow better.

How do we know whether AI training is working without rating people?

Follow adoption by team and by workflow: which workflows each team runs, and where a team has stalled. A stalled team gets another session, a clearer job aid or a fix to the method, never a mark against a person. Nobody is scored on how fast they took to it.