The short answer.

A company’s use of AI moves through six stages. (1) Not knowing it is possible: people use an assistant every day and do not know that a piece of repeat work can be written down once and prepared by AI the same way every time. (2) Switched on: the seats are bought and each person is left to work it out. (3) Experimenting: people try things, a few find something good, and it stays on their own desk. (4) Planned: one page says which work, in which team and role, moves onto AI first, and who owns it. (5) Built: the first workflows are written, tested on real work and run by the people who own the work. (6) Managed: every team runs current versions, with owners, a record and a regular review. Most companies are at stage two or three, and that is a fine place to have started. The move that changes everything is from experimenting to planned.

Stages one to three are undirected: AI is on, and each person decides alone what to do with it. Stages four to six are directed: the company decides, on purpose, which work moves onto AI and in what order. Our three steps map onto the last three stages. Plan it, and you reach planned. Develop the workflows, and you reach built. Deploy them, and you reach managed.

Stage one: not knowing it is possible.

At stage one, people use an AI assistant every day as a helper and do not know that whole pieces of their repeat work can be handed over.

  • On a normal Tuesday.Someone asks the assistant to tidy an email, then builds the weekly report by hand, the same way as last week.
  • What the company can say.Some of our people use AI.
  • What it cannot say.Which work AI could prepare, because nobody has asked the question.
  • What moves it to stage two.The company decides AI deserves a line in the budget and buys seats for its people. Did you know your business workflows can become something AI runs for you? shows what those seats can do.

Stage two: switched on.

At stage two, the company has bought seats of ChatGPT, Claude, Copilot or Gemini and left each person to work out what to do with them.

  • On a normal Tuesday.Everyone has a login. A few people use it daily, most open it now and then, and nobody has said what it is for.
  • What the company can say.How many seats it pays for and what they cost.
  • What it cannot say.What the seats are for, or what the money bought.
  • What moves it to stage three.Someone gives people room to try the assistant on real work and asks them to share what works.

Leaving AI to each person is a decision too: no plan is still a plan, and it has a cost. What happens when a company uses AI without a plan? sets out that cost.

Stage three: experimenting.

At stage three, people try AI on their own work, a few find something good, and the good find stays on one person’s desk.

  • On a normal Tuesday.A sales lead has a prompt she trusts for proposal drafts. Someone in finance checks supplier statements with the assistant. Neither knows about the other.
  • What the company can say.We have people doing good things with AI, and we can name a few.
  • What it cannot say.Which work across the company moves onto AI next, who owns what has been found, or whether a good method survives its author changing role.
  • What moves it to stage four.One page that says which work, in which team and role, moves onto AI first, and one person, by name, who owns each piece.

Experimenting was the right way to start, and it stops paying once it is all there is. After years of experimenting, AI is a real line in the budget and deserves disciplined, focused attention. Why plan how your company uses AI? makes the full case.

Stage four: planned.

At stage four, the company has one page that says which work, in which team and role, moves onto AI first, and who owns each piece.

  • On a normal Tuesday.The head of operations opens the plan: each team, its roles and its first workflow. Finance is developing a month-end checklist now; client service is next.
  • What the company can say.Which work moves onto AI first, in which team and role, who owns it, and what comes after it.
  • What it cannot say yet.That AI prepares the work, because nothing has been written and tested.
  • What moves it to stage five.Developing the first workflows: writing down the best way of doing each piece of work and testing it on last month’s real cases. How to build your first AI workflow walks through it.

Stage five: built.

At stage five, the first workflows are written, tested on real work and run by the people who own that work.

  • On a normal Tuesday.The accounts payable lead runs the invoice-matching workflow in the company’s own Claude, checks what it prepared and approves each match.
  • What the company can say.These named workflows exist, each was tested on real work, and each has an owner and a person who decides.
  • What it cannot say.That everyone runs the current version, or how a change reaches every user.
  • What moves it to stage six.Deploying: rolling each workflow out role by role, checking that it is used and useful, and putting a review on the calendar. How to deploy AI workflows to a whole team and Installed, used or useful? cover both.

Stage six: managed.

At stage six, every team runs the current versions of its workflows, with an owner for each, a record of changes and a regular review.

  • On a normal Tuesday.A new version of the client brief workflow went out last week. The owner recorded what changed and why, and every account manager now runs it.
  • What the company can say.Which workflows run in which teams, which version each team runs, who owns each one, and what changed and why.
  • What keeps it there.The monthly review: is each workflow used, is it useful, and which workflow from the plan comes next. How to manage AI across teams sets out the routine.

The FullyPowered Enterprise Console, for managing releases and rollout across a company, is coming soon.

Six questions that tell you your stage.

Copy these six questions and answer each yes or no. Start at the top. Your company is at the stage of the last yes before the first no.

  1. Not knowing it is possible.Does anyone in the company use an AI assistant for work?
  2. Switched on.Has the company bought seats of an AI assistant for its people?
  3. Experimenting.Can you name three people who use AI on real work, and what each one uses it for?
  4. Planned.Is there one page that says which work, in which team and role, moves onto AI first, with one owner named for each piece?
  5. Built.Is at least one workflow written down, tested on real work and run each week by the people who own that work?
  6. Managed.Does every team run the current version of its workflows, with an owner, a record of changes and a review on the calendar?

What is different in a company of under 200 people.

A company of under 200 people can go from switched on to planned in a week, because the people who decide are in one room. You can name every team from memory and choose each team’s first workflow by Friday.

Make the move to planned.

The move from experimenting to planned is the one that turns undirected AI use into directed work. How organizations plan their use of AI gives you the one-page plan and the three steps after it.

FullyPowered Planner is where that plan lives. You set out your teams, their roles and the work between them, and choose what to build first. Colleagues add their own teams. That is how the seats you already pay for start to earn their place, and how you 10X the ROI of your existing AI.