The short answer.

You go from a few AI pilots to the whole company by making three things true. The pilot is written down well enough that someone else runs it the same way. It has an owner who is not the person who built it. And a plan names which team is next, which workflows that team gets and who owns each one. A pilot that lives in one person’s head stays a pilot; a pilot with these three things becomes the first page of a company plan.

Every company plan for AI has three steps: plan it, develop the workflows, deploy them. A pilot covers the second and third steps for one team. Going company-wide means doing the first step for every team.

Why good pilots stall.

Good AI pilots stall for three reasons, and none of them is the AI.

  • The enthusiast.The pilot was built by the one person who loves the assistant. That person knows which prompt to paste, which file to attach and which output to ignore. Nobody else does, so the pilot works only when the enthusiast is in the room.
  • The missing owner.The builder moves on to the next experiment. Nobody keeps the pilot current when a procedure or an export changes, and within a quarter the team quietly stops using it.
  • No next step.Leadership saw a good demo and said “great”. Nobody decided which team goes second, so the pilot became a story people tell instead of the start of a plan.

Each reason has a plain fix: write it down, hand it to an owner, put the next team on a plan.

Turn the pilot into a workflow others run.

A pilot becomes a workflow when a second person, working from the written version alone, gets the same result the builder gets. Four moves take it there.

  1. Write down what the builder does.What the work starts from, the order of the steps, what gets checked, what the output looks like and where a person decides. Sit with the builder for an hour and write it as they talk.
  2. Put it in the company’s own assistant.The workflow runs in your company’s ChatGPT, Claude, Copilot or Gemini, on your account, not in the builder’s personal one. Anyone on the team opens it and runs it.
  3. Test it with someone else at the keyboard.A colleague who did not build it runs it on last month’s cases, where the right answer is already known. Every miss is a gap in the written version. Fix the instructions, not the colleague.
  4. Name the owner.One person in the team that uses the workflow, by name, who keeps it current and decides when it changes. Who owns an AI workflow after the pilot? sets out the four roles around a workflow.

The builder stays the expert to call. The owner is the person accountable. Keep those two jobs in two people, and the workflow survives the builder’s next project. The rollout checklist then takes the workflow from working to daily use in its first month.

Choose the second and third teams.

The second and third teams are the ones that receive work from the first. The first workflow already improved what leaves the pilot team, so the team on the other side of that handoff sees the difference first and has the most reason to go next.

Follow the work. If the pilot prepares a brief for each new client, the team that onboards the client goes second. The handoffs between teams are where work gets stuck, and where a prepared handoff pays off most.

For each new team, choose one workflow to start. Choose your first AI wins has the scorecard: how often the work happens, whether it starts from records someone can export, whether one person already reviews it, and what a miss would cost.

The one-page case for leadership.

Make the case for a company-wide AI plan on one page, in your own terms, with names. Leadership approves a plan with owners; it does not approve enthusiasm. Copy these headings into a document and write one or two lines under each.

  1. What the pilot was.The piece of work, the team, and what the workflow starts from and produces.
  2. What the pilot showed.In the team’s own words: what is now prepared that used to be done by hand, and what the people who decide still decide.
  3. Who runs it now.The owner by name, and the people in the team who run it without the builder.
  4. The next three workflows.One line each: the team, the work, what it starts from and what it produces.
  5. The owner of each.One name per workflow, in the team that will use it.
  6. What it asks of each team.The hours from the owner and from the person who knows the work best, and when.
  7. The AI we use.The assistant the company already pays for, and what information may go into it.
  8. The review.Who looks at the plan each month, on which day, and what they check.
  9. The decision we need.One sentence: what leadership is asked to approve today.

Illustrative example.

Take a 60-person distributor. The customer service lead built a pilot that turns each incoming order email into a checked order summary. She wrote it down with the operations manager, moved it into the company’s own assistant, and a colleague ran it on the previous month’s orders. The service supervisor became its owner. The order summary goes next to the warehouse, so the warehouse’s pick-list check became the second workflow, owned by the warehouse lead. The third is the dispatch note that finance reads when it invoices, owned by the finance lead. The operations manager put all three on one page, with each owner’s name and the hours asked of each team, and took it to the next leadership meeting.

The review that keeps it moving.

A monthly review of thirty minutes keeps a company AI plan moving. One person runs it, by name, usually the head of operations or the owner of the business. It asks the same questions every month.

  • Is each workflow used?The owner says who ran it this month and how often.
  • Is it useful?The owner says what changed in the work and what the exception log shows.
  • What needs fixing?Anything that broke or drifted, and who fixes it by when.
  • What starts next?When a workflow is in daily use, the next one on the plan starts. No more than two or three in development at once across the company.

What is different in a company of under 200 people.

In a company of under 200 people, the path from pilot to whole company is short. Most published advice assumes a large enterprise, with a steering committee, a center of excellence and a year-long pilot in one division. A smaller company works differently.

  • The case is one page, not a program.The people who approve it sit in one meeting.
  • The owner is an operator.The head of operations or the owner of the business runs the plan. IT helps with accounts and access.
  • Every team is a short walk away.You can trace the work from the pilot team to the next two teams in an afternoon.
  • You can cover the whole company in a quarter.One workflow per team, started in turn, reaches every team while a large company is still choosing its pilot division.

That speed is where 10X the ROI of your existing AI comes from: the seats you already pay for, pointed at named work, team after team. Make an AI pilot your next quarterly priority shows how to set the first one up as a quarterly goal.

Where to go next.

The whole method, from the first plan to keeping workflows working, is in How organizations plan their use of AI. When the plan outgrows a document, FullyPowered Planner is where it lives: you set out your teams, their roles and the work between them, and choose what to build first. Your colleagues add their own teams. We also develop the next workflows with you, in a 3-Day Workflow Sprint or a 2-Week Workflow Sprint.