The short answer.
An organization plans its use of AI the way it plans anything else that matters: it decides where it is going, who does what, and what comes first. For AI that means one page that lists your teams, the roles in them, the work each one repeats, and the first workflows each team should have. Every workflow on the page has a named owner and a person who decides.
Then the plan is carried out in three steps.
- Plan it.Decide, team by team, which work AI should prepare and in what order.
- Develop the workflows.Write down your best way of doing each piece of work, precisely enough that an AI assistant prepares it the same way every time, and test it on real work.
- Deploy them.Put each workflow in the hands of the people who own the work, in the AI your company already pays for, and keep it working.
That is the whole method. It works at 20 people and at 2,000. FullyPowered AI is a better way to do all three, and the rest of this guide shows how each step is done.
What most companies do instead: using AI is not a plan.
Most companies already use AI. Someone bought seats of ChatGPT, Claude, Copilot or Gemini. A few people use them every day, most use them now and then, and each person has found their own tricks. Leadership knows the money is spent and cannot say what it bought.
That is undirected AI use. Nobody chose it: it is what happens when the tools arrive before the method. Leaving AI to each person is still a decision, and after years of experimenting it has a cost. What happens when a company uses AI without a plan? sets that out, and the stages of AI at work shows where a company stands. Three things are missing.
- A list of tools is not a plan.A plan starts from the work, and a tool is what you point at it afterwards. Decide what each team should be doing with AI and every purchase after that has a place to land. A plan before agents makes the case.
- A policy is not a plan.Rules about what staff may paste into an assistant are worth having. They say what not to do. A plan says what to do.
- A pilot is not a plan.One good experiment in one team teaches you something. A plan says which team is next and what the whole company looks like when you are done.
The fix is simple and it does not need new software to begin: point the AI you already pay for at named pieces of work. 10X the return on the AI you already pay for shows that move on one piece of work.
Step one: plan it.
Planning is six decisions. Make them in this order and you have a plan by the end of a week.
- Say where you are going.One sentence the whole company can repeat: what your teams will be able to do with AI that they cannot do today. Decide where your company is going with AI shows how to write it.
- List your teams, then the roles in them.Teams first. A team shares its work, its records and its standards, so most of the plan can be settled at the level of the team. Roles come next, named for the work and not for a person. Start with your teams gives a first set for each.
- Sort each team’s work in two.Some work repeats the same way every week. Some never repeats. Repeating work becomes workflows. Ad hoc work gets better help from the assistant one question at a time. Repeatable work and ad hoc work has the sorting sheet.
- Choose what to build first.Score each candidate on what you can see today: how often it happens, whether it starts from records someone can export, whether one person already reviews it, and what a miss would cost. Choose your first AI wins has the scorecard.
- Name an owner for each workflow.One person who keeps it current and decides when it changes. A workflow without an owner is out of date in a quarter.
- Show the work between teams.The work that matters most crosses from one team to the next. Put those handoffs on the plan, because that is where work gets lost and where a prepared handoff pays off. An AI org chart should show the work between teams draws one.
How many at once? Start one workflow per team, and no more than two or three across the company. Each needs its owner’s attention while it is tested on real work, and attention is what runs out first. When the first ones are in daily use, start the next.
What to leave alone. Work that happens a few times a year. Work where the whole value is one person’s judgment in the moment. Work where a mistake cannot be caught before it leaves the building. Leave these off the first plan. There is plenty that repeats.
Step two: develop the workflows.
A workflow is your best way of doing one piece of work, written down precisely enough that an AI assistant prepares it the same way every time. Developing one is four moves.
- Capture how your best person does it.What they start from, the order they work in, what they check, what they leave out and what good looks like. From one person’s know-how to a team method shows the interview.
- Write it as instructions.The inputs, the steps, the checks, the output and the point where a person decides. From SOP to AI workflow is a worked example, and A prompt, a workflow template and a workflow shows the same task at three levels of finish.
- Test it on real work.Run it on last month’s cases, where you already know the right answer, and keep a record of every miss. What we test before a workflow is ready shows the record.
- Start from a template where one exists.The Workflows Directory holds free workflow templates by team and by industry. They are samples and a starting point, made to be customized, and we can build any workflow for your company.
You can develop a workflow yourselves with this guide and a template. We also do it with you: a 3-Day Workflow Sprint or a 2-Week Workflow Sprint takes one piece of work from how your team does it today to a tested workflow your people run. Every price is on the pricing page.
Step three: deploy them.
A workflow is deployed when the people who own the work run it every week without being asked. Four things get it there.
- It runs where your people already work.In your company’s own ChatGPT, Claude, Copilot or Gemini, on your account, run by your people. Nothing new to log in to. Your company data in ChatGPT, Claude, Copilot or Gemini covers what that means for your information.
- It rolls out role by role.The owner runs it first, then the rest of the team, with the same instructions and the same checks. The rollout checklist covers the first month.
- You measure whether it is used and useful.Installed is not used, and used is not useful. Three questions tell you which you have. Installed, used or useful? has the sheet.
- Someone keeps it working.Procedures change, exports change and the assistants ship new models. The owner’s routine keeps the workflow current. Keep it working follows one workflow through six months.
A person decides at every step. The workflow prepares, checks and drafts. Anything that sends, pays or changes a record waits for someone to approve it. The FullyPowered Enterprise Console, for managing releases and rollout across a whole company, is coming soon.
A one-page AI plan to copy.
Copy these headings into a document. Fill them in with the leaders of each team in one sitting. One page is enough for a first plan.
- Where we are going.One sentence: what our teams will be able to do with AI by the end of the year.
- Our teams.The three to six teams the plan covers first, and the lead of each.
- For each team: the roles.Named for the work, such as accounts payable or intake, not for a person.
- For each team: the work that repeats.Five pieces of work the team does the same way every week or month.
- For each team: the first workflow.The one piece of work to develop first, what it starts from, what it produces and who decides.
- The owner of each workflow.One name per workflow.
- The handoffs.The two or three places where work passes between teams and gets stuck.
- The AI we use.Which assistant the company has approved, and what information may go into it.
- The order.What is being developed now, what is next, and what waits.
- The review.Who looks at the plan each month, and what they check: is each workflow used, is it useful, what is next.
Illustrative example.
Take a 40-person services firm. Where it is going: every client gets an answer the same day, and month end closes in the first week. Its first three teams are finance, client service and operations. Finance starts with a month-end close checklist, owned by the controller, who decides every posting. Client service starts with a prepared brief before each client call, owned by the head of accounts. Operations starts by writing down how a new client is set up, owned by the operations lead. The handoff that gets stuck is from sales to client service when a new client signs, so that goes on the plan next. The managing director reviews the page on the first Monday of each month. A plan before agents shows a filled-in page in full.
What is different in a company of under 200 people.
Most of what is published about planning AI is written for large enterprises: steering committees, centers of excellence, a platform team. A company of 20 or 150 people plans differently, and it has real advantages.
- The plan fits on a page.You can name every team and most roles from memory. A large company needs a program to find out what its people do. You can ask them on Tuesday.
- The owner is an operator, not an IT department.The person who owns the plan runs operations or the business itself. IT helps with accounts and access. Do you need an IT person to start? answers that directly.
- You start with the assistants you already bought.No platform decision comes first. The plan decides what the seats you have are for.
- The rules are short.Which assistant is approved, what may go into it, and who decides. Two paragraphs, written by the person who owns the plan.
- You can cover the whole company.A large company pilots in one division for a year. You can give every team its first workflow in a quarter, because the people who decide are in the same room.
Small is the advantage here. This is where 10X the ROI of your existing AI comes from: the same seats, pointed at the work that matters, by a company small enough to decide quickly.
Who owns the plan.
One person, by name. In most companies under 200 people that is the head of operations, the chief of staff or the owner. Their job is to keep the page current, settle the order of work between teams and run the monthly review. Each workflow then has its own owner inside the team that uses it. Who owns an AI workflow after the pilot? sets out the four roles around a workflow, and Build your AI plan with the people who do the work shows how to bring each team into it.
Where the plan lives.
A first plan can live in a document or a spreadsheet. It works for the first page. It strains when three teams edit it, when one workflow belongs to several teams, and when you want to see every role that touches one piece of work.
FullyPowered Planner is where the plan lives after that. You set out your teams, their roles and the work between them, and choose what to build first. Your colleagues add their own teams, and everyone sees their part of the same plan. The plan is yours and stays with your company.
Decide where your company is going with AI. Then bring your people there, one workflow at a time.
Seven-day trial with a card, everything unlocked, nothing charged for seven days. See every plan and price.
Go deeper on each part.
- Why plan at all.Why plan how your company uses AI? and Did you know your business workflows can become something AI runs for you?
- Who owns it.Who should own AI in a company your size? and How to manage AI across teams without everyone doing their own thing.
- The rules.Does a small company need an AI policy? It includes a two-page policy to copy.
- Building the first workflow.We have a plan. How do we build the first AI workflow?
- Rolling it out to everyone.How to roll out ChatGPT, Claude, Copilot or Gemini across a company, From a few AI pilots to the whole company and How to deploy AI workflows to a whole team, and keep them working.
- Whether it is paying back.How to tell whether AI is paying back.
- The year ahead.How to put AI in next year’s plan and budget.
- Software for the plan.Is there software for planning how a company uses AI?
- Where to start.The readiness checklist and The first 90 days of an AI adoption plan.
- What to do first.Choose your first AI wins and Start with your teams.
- Planning by team and role.An AI org chart should show the work between teams and Add AI workflows to your chart of seats.
- Developing a workflow.From SOP to AI workflow and What we test before a workflow is ready.
- Rolling it out.The rollout checklist and Installed, used or useful?
- What it is worth.What an AI workflow is worth and What it costs to put AI to work.
- Doing it yourselves or with help.Should we hire an AI consultant, and what does it cost?.