The short answer.
An AI plan for a company is one page that lists your teams, the roles inside each team, the work each role repeats, the workflows that will prepare that work and the order they are built in, with one named owner and one person who decides for every workflow. One person owns the whole plan: an existing leader who already runs the work, usually the head of operations, the chief of staff or the owner of the business, who keeps the page true, settles which team is next and runs a monthly review. Each workflow on the page has its own owner inside the team that runs it, and a person in that team decides on everything the workflow prepares. A short strategy sits above the plan and a two-page policy sits beside it, and of the three documents only the plan tells a team lead what their team does next.
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 question a leader asks about AI, from the policy to the org chart to next year’s budget, is answered by the same five facts written once for each workflow. The five facts are the team, the role that runs the workflow, the piece of work it prepares, the owner who keeps it current and the person who decides on its output. The strategy is the sentence above those rows. The policy is the margin around them. A goal is a row with a quarter added. The budget line is the owners’ time on those rows. The org chart is the same rows drawn under your reporting lines. Write the five facts once and every other AI document is a view of the plan.
Planning is the first of three steps set out in How organizations plan their use of AI: plan it, develop the workflows, deploy them.
An AI plan is one page of teams, roles, workflows, order and owners.
An AI plan holds six things: the direction in one sentence, the teams, the roles in each team, the workflows each role runs, the order they are built in, and an owner and a person who decides for each workflow. A list of tools to buy is not part of it.
- The direction.One sentence on what your teams will do with AI that they cannot do today.
- Teams, named for the work.Each with its lead. A team of one is still a team.
- Roles inside each team.Named for what the job does, never for the person holding it.
- Workflows under each role.Two or three per role, each one your best way of doing a piece of repeating work, written so the assistant you already pay for prepares it the same way every time.
- The order.Now, next and later, across the whole company.
- Two names per workflow.The owner, who keeps the workflow current, and the person who decides on what it prepares.
A plan missing the names is a wish list; a plan missing the order starts every team’s first workflow at once, and none of them finishes. FullyPowered AI 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 put an owner on every workflow. Colleagues add their own teams, and the plan prints and exports when the leadership team wants it on paper.
A one-page AI plan to copy. Fill in the top once, then one block for each team. Write “Unknown” where nobody knows yet, with the name of who will find out.
- Where we are going.One sentence: ______
- Owner of the plan.One name: ______
- Team and lead.______
- Roles in the team.Named for the work: ______
- Workflows, in order.For each: the role that runs it, what it starts from, what it produces: ______
- Owner of each workflow.One name per workflow: ______
- Who decides.For each workflow, the role that approves what it prepares: ______
- Work that crosses to another team.What is handed over, and to which team: ______
- Status.Now, next or later, for each workflow: ______
- Review day.The same day each month: ______
How is an AI plan different from an AI strategy and an AI policy?
An AI strategy says why the company uses AI and where it should help first, an AI policy says what everyone may and may not do with it, and an AI plan says which role in which team runs which workflow, in what order, owned by whom. The three answer different questions and change at different speeds: the strategy and the policy change once or twice a year, and the plan changes every month as workflows are built and new ones are chosen.
A quick test settles which one you are holding. Ask what the document lets a team lead do on Monday. If it tells the lead what matters, it is a strategy. If it tells the lead what is allowed, it is a policy. If it names the next workflow for the lead’s team and the person who owns it, it is a plan. AI strategy, AI policy, AI plan: what is the difference? puts all three on one page.
A small company does need a written AI policy, and two pages are enough: which assistant is approved, on the company account; what information may and may not go into it; that a person checks and decides before anything is sent, paid or changed; and one name to ask. Banning AI while the policy is written only moves the work onto personal accounts the company cannot see. Does a small company need an AI policy? has the two pages to copy.
Together the two articles make a point about order: the policy is the fastest of the three to write and the easiest to mistake for progress. Write the strategy and the policy quickly, then put the effort into the plan. A company that bought seats, wrote rules and stopped is the company in Why companies get so little from the AI they already pay for.
Who should own the AI plan, and where does AI go on the org chart?
One person owns the AI plan, by name, and that person is an existing leader who already sets the order of work between teams: the head of operations, the chief of staff or the owner of the business. Not a committee, which defers; not a new head of AI, who must first learn the teams; and not IT, which owns the systems rather than the work.
Ownership of AI runs in three layers, and confusion about “who owns AI” comes from collapsing them into one.
- The owner of the plan.One person for the company. Owns the page, the order of work between teams and the monthly review.
- The owner of each workflow.One person in the team that runs the workflow, usually whoever knows the work best. Keeps the workflow current and decides when it changes.
- The person who decides.The role that approves each output before it is sent, paid or changed. Often the person who reviewed the same work before any workflow existed.
Who should own AI in a company your size? has a one-page role description for the owner of the plan.
The same three layers answer the org chart question. AI gets no box on the chart, because a box means a person with a reporting line, and workflows report to nobody. AI goes beside each role, as the two or three workflows that role runs, each with its owner and the person who decides written next to it. The reporting lines stay as they are, and owning the plan is a responsibility added to a job someone already holds. Where does AI go on the org chart? shows how to mark up the chart you already keep.
Both articles reach one conclusion: ownership sits where the knowledge already is. The leader who sees work move between teams owns the order, and the person who does the work best owns the method.
What the owner of the AI plan does each week and each month.
Each week the owner of the AI plan checks that the workflows marked “in use” are still run by their roles, clears what is blocking the workflow in development, hears one team in rotation and then updates the plan with what those three turned up. Each month the owner runs the review with the workflow owners, sets the next month’s order, confirms every workflow still has an owner and sends one paragraph to leadership.
The order of the weekly jobs matters. Updating the page last means the page records what is true, rather than looking current while the work behind it drifts. When a workflow drifts, the owner of the plan names the problem and the workflow owner fixes it. What does the owner of the AI plan do each week? has a weekly checklist, and How to run a monthly AI review covers the meeting.
A new owner starts differently. The first month goes to mapping where work passes from one team to another, getting two decisions from the CEO (the direction and the public naming of the owner), and getting one owner and one piece of work from each team lead. The handoff map is the part of the plan only the owner can write. What should a head of operations do about AI? has the first-month checklist.
The owner of the plan does not write every workflow, approve every use of the assistant or decide inside a team’s work. Once the plan is written and the first workflows reach their teams, How to roll out AI across a company, and tell whether it is working takes over.
What the company decides centrally, and what belongs to each team.
A company decides five things about AI centrally: the assistant it pays for, the ground rules, the one plan, the order of what is built and the owner of each workflow. Each team decides three things: which of its work to propose, the detail of each workflow and who checks what each workflow prepares. Central decisions are made once and apply to every team; team decisions are made by the people who do the work, inside the central five.
Without that split, every team does its own thing: the same work is solved several ways, and nobody can say what the company runs. What should we centralize about AI, and what should we leave to teams? draws the line item by item, and How to manage AI across teams gives the workflow record that holds the centre together.
A new question needs a test, because no list covers everything. A question goes to the centre when a wrong answer in one team creates work or risk for another team, or when two teams would otherwise answer the same question twice.
The articles together show that the split and the plan are the same thing. Each central decision is a line on the plan, and each team decision is filled in on the team’s own block, by the team. A company that keeps one plan has already drawn the line between centre and teams.
Plan by team first, then by role.
Plan AI by team first, because a team shares its records, its standards and the person who approves its work, and then by role, because a workflow belongs to a role and stays when a person moves. A role is a kind of job named for the work, such as renewals, intake or billing.
For each role, write four things: the work it repeats, the workflows it runs, what each workflow starts from, and what the person in the role still decides. How do we plan AI by role? has the role sheet. A role with no workflow on the plan is a role whose repeating work nobody has looked at yet.
The line between people and AI inside a role is the same on every team: AI prepares, people decide. Anything that sends, pays or changes a record waits for a person. People and AI on one team: who does what? turns that split into five lines per workflow.
The two articles meet at one rule for the plan: every workflow on the page names two roles, the role that runs it and the role that decides. A plan that names only the first has written down the preparation and left the decision to whoever happens to read the draft. The plan describes work and never performance; nobody is rated or ranked by it.
How do we write the first plan?
The first AI plan is written in two moves: one-to-one conversations with each team lead about the team’s work, then one meeting with all the leads that turns the notes into the page. The one-to-ones come first because a lead says more alone than in front of peers, and because a lead asked about the team’s week, not about AI, names the first workflow.
What should a CEO ask each team lead about AI? has the full set of questions for the one-to-one. The short form below labels each question with the part of the plan it fills.
Team lead questions to copy. Ask them in the same order in every conversation, and write the answers in the lead’s own words.
- For the workflows.Which pieces of work does your team do the same way every week or month? Name five: ______
- For what each starts from.For each, what does someone open to begin, and can it be exported today? ______
- For who decides.Who checks each piece before it leaves the team? ______
- For the order.Where does work wait, or get done twice? ______
- For the method.What does your best person do that others do not? ______
- For the work between teams.What arrives from another team, and what do you send on? ______
- For what stays with people.Which work should AI stay away from? ______
- For timing.What is in the way this quarter, and when does it end? ______
- For the owner.Would you, or someone on your team by name, own the first workflow? Yes, not now, or a name: ______
The meeting then answers six questions in a fixed order: where are we going, which teams go first, what repeats in each team, which piece comes first and why, who owns each one, and when do we review. A note-taker writes the page as the room answers. How to write your company’s AI plan in one sitting has the agenda to copy.
Put the two articles side by side and the meeting turns out to be mostly a ranking exercise. The one-to-ones supply nearly every field on the page; the room adds the direction and the order. Skip the one-to-ones and the meeting collects facts instead of deciding.
The year’s AI goals and the budget line hang from the plan.
An AI goal for the year is one row of the plan with a quarter added: a named workflow, run by a named role in a named team, in daily use by the end of a named quarter, with one owner. “Adopt AI” and a count of seats are not goals, because nobody can check them on the last day of a quarter. Four to eight goals, one or two per quarter, with the surest first, is a year.
Each goal is checked with a yes-or-no question at the end of its quarter: does that role run that workflow on every case now? How do we set AI goals for the year? has a blank goals page and three vague goals rewritten.
AI goes into the annual plan the same way: as work, not as a tool line. Under each company goal, name the team that carries it, the work that repeats behind it and the workflow that will prepare that work. The AI budget line then has three parts: the assistant seats the company already pays for, the time of the people who own the workflows, and any outside help to develop them. How to put AI in next year’s plan and budget has the one-page sheet; every price for our own help is on the pricing page.
Taken together, the two articles remove two documents: there is no separate AI goals deck and no separate budget paper, and the monthly review of the plan is also the progress check on the year’s goals.
One plan, from the first conversations to the first review.
Illustrative example.
Take an 85-person civil engineering firm with four teams on its plan: project delivery, proposals, finance and the office. The managing principal names the director of operations as owner of the plan at a leadership meeting, so every lead knows whose page it is.
The managing principal writes the direction in one sentence: every client receives the same careful paperwork whichever engineer leads the project. The director of operations writes the two-page policy the same week: one approved assistant, no confidential client drawings or staff records in it, a person reads everything before it leaves the firm, and questions go to her.
She then holds a one-to-one with each lead and asks the team lead questions. Proposals rebuilds the same project descriptions and staff summaries by hand for every bid. Finance prepares invoices from the timesheet system, and the finance manager checks every one. Project delivery wants help with design calculations; the lead says the calculations stay with licensed engineers, so the team’s candidate becomes the notes and action list after each client meeting. The office manager is a team of one and is moving the firm to a new file system, so the office says “not now”.
At the planning meeting the leads agree the order. The proposal first draft comes first, run by the proposals coordinator role, starting from the firm’s past proposals and staff summaries, owned by the proposals coordinator and decided by the principal who signs each bid. The invoice preparation workflow comes next, owned and decided by the finance manager. The meeting notes workflow waits a quarter, owned by a senior project engineer, with the project manager deciding what goes to the client. The office joins after the file move. Two of those become goals for the first half of the year, each with its quarter. The handoff the director of operations names is the one from proposals to project delivery, when a won bid becomes a project.
At the first monthly review she moves the invoice workflow from “next” to “now”, and the goals and the budget line are read from the same page.
What is different in a company of under 200 people.
A company of under 200 people plans AI by conversation, and its plan, its goals and its budget line can be one page kept by one person. Much of the published advice is written for large enterprises, with steering committees and a chief AI officer; a smaller company needs neither.
- The centre is one person.The head of operations, the chief of staff or the owner holds the plan alongside a current job. No new department, no new box.
- Every lead is in reach.The one-to-ones fit inside a normal week, and the order of teams is settled in one meeting.
- One person often wears several hats.The owner of the plan may own a workflow too, and a team lead may run, own and decide on the same piece of work. Write each hat on its own line, so everyone can tell which one a decision comes from.
At 20 people the plan is smaller still: three or four teams, six to ten roles, and five pieces of work across the whole company rather than five per team, because the same few people develop and test every one. What larger companies add (a committee, a scoring sheet for every task, a map of every handoff) stays out until the company grows into it. What does an AI plan look like for a 20-person company? has a half-page plan to copy. The headings stay the same as the company grows; the lists get longer.
That is where 10X the ROI of your existing AI comes from in a smaller company: the seats you already pay for, pointed at named work, with an owner on every workflow. When a workflow on the plan is ready to build, the Workflows Directory holds free workflow templates to start from, and we build any workflow your teams need.
Questions people ask.
For short answers across the whole subject of planning AI, Answers takes the questions leaders ask in order.
- What actually goes in a company AI plan?
A company AI plan holds one sentence of direction, the teams, the roles in each team, the workflows each role runs and the order they are built in. Every workflow has an owner who keeps it current and a person who decides on what it prepares. A monthly review date closes the page.
- Isn’t our AI policy the same thing as an AI plan?
No. An AI policy says what everyone may and may not do with AI: the approved assistant, what information may go into it and that a person decides. An AI plan says which team and role gets which workflow, in what order and owned by whom. A company with only a policy has permission and no direction.
- Who should own AI in a company of 50 or 100 people?
One existing leader who already sets the order of work between teams, usually the head of operations, the chief of staff or the owner of the business. That person owns the plan, not every workflow. IT handles accounts and access, and each workflow has its own owner in the team that runs it.
- Where does AI go on our org chart?
AI gets no box on the org chart, because workflows report to nobody. AI goes beside each role, as the two or three workflows that role runs, with the owner of each workflow and the person who decides written next to it. The reporting lines stay as they are.
- How long does it take to write an AI plan?
A company of under 200 people writes its first AI plan in one meeting with the leader of each team. The groundwork is a one-to-one with each lead beforehand about the work the team repeats, who checks it and who would own a workflow. The page is finished when the meeting ends, and the first workflow starts the next morning.
- How do we set AI goals that mean something?
Write each goal as one named workflow, run by one role in one team, in daily use by the end of a named quarter, with one owner. Check it on the last day of the quarter by asking whether that role now runs that workflow on every case. “Adopt AI” and seat counts are not goals, because nobody can check them.
- Does a 15-person company need an AI policy?
Yes, and two pages are enough: the approved assistant on the company account, what information may and may not go into it, that a person checks before anything is sent, paid or changed, and one name to ask. Do not ban AI while you write it, because people will use personal accounts instead. Write the two pages first, then the plan that says what each team does with the assistant.