The short answer.

Yes, companies have been here before. Every office tool arrived the same way: bought for everyone, used well by a few, and worth far more once the company decided what it was for. The spreadsheet paid off when a team agreed on one model, email when a company agreed how to use it, and business software when the process was written down first. AI is at the same point, and the lesson is the same: the tool arrives first, the method follows, and the companies that write the method down get the value.

What to do differently this time is to skip the long wait. You already know the method comes second, so decide it on purpose now, team by team, instead of letting each person find their own.

Every office tool arrived the same way.

Each new office tool followed the same three stages, and AI is following them too.

  1. Bought for everyone.The tool was new, cheap enough and plainly useful, so the company gave it to every desk. Nobody wrote down what it was for, because nobody knew yet.
  2. Used well by a few.A handful of people found what the tool was good at and built their own habits around it. Everyone else used a small part of it, or used it the way they had worked before.
  3. Worth far more once the company decided.The value arrived when someone in charge said: this is what the tool is for here, this is how we use it, and this is who keeps it right. The tool did not change at that moment. The method did.

Most companies are in the second stage with AI today. Seats of ChatGPT, Claude, Copilot or Gemini were bought, a few people use them every day, and each person has their own tricks. The stages of AI at work describes where a company sits and what moves it on. Experimenting was the right way to start. After years of experimenting, the next stage is a decision.

The spreadsheet paid off when a team agreed on one model.

The spreadsheet paid off when a team stopped keeping its own versions and agreed on one model everyone worked from. Before that, the spreadsheet made each person faster and the team slower. Every analyst had a copy with their own formulas, the numbers did not match, and meetings went on arguing about whose figures were right.

The fix was never a better spreadsheet. A team agreed on one model: one set of inputs, one structure, one owner who changed the formulas, and a rule that the shared copy was the true one. Then the spreadsheet made the whole team faster.

The lesson for AI: personal prompts are the private spreadsheets of today. Each person’s way of asking an assistant to draft a proposal or check an invoice gives a different answer. A shared workflow, written once, tested and owned by one person, is the one model the team works from.

Email paid off when a company agreed how to use it.

Email paid off when a company agreed how to use it: what goes by email, who is copied, how fast a reply is expected and where decisions are recorded. Before those agreements, email made sending easy and working hard. Inboxes filled, threads split, decisions were buried in replies, and each person set their own rules.

The companies that got the value from email wrote a few plain conventions and kept to them. The tool was the same in every company. The agreements were the difference.

The lesson for AI: an assistant on every desk without agreed conventions produces the same noise. Agree which work AI prepares, what a person checks before anything leaves, and which information goes into the assistant. What happens when a company uses AI without a plan? sets out what undirected use costs.

Business software paid off when the process was written down first.

Business software, the systems for accounting, orders, customers and stock, paid off when the company wrote its process down before it set the system up. A company that bought the system first and hoped the process would follow got software that matched nobody’s real work, and people kept their side spreadsheets.

The companies that did well mapped the work first: who does each step, in what order, with what checks, and who approves. The system then carried a process people already agreed on.

The lesson for AI is the closest of the three. 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. Writing the method down is the whole job, and it comes before any new tool.

What should we do differently this time?

This time, decide the method on purpose and early, instead of waiting for it to emerge. With earlier tools, companies waited for the few good users to spread their habits on their own. That wait was long, and many habits never spread. With AI you can shorten it.

  • Start from the work, not the tool.List your teams, the roles in them and the work each one repeats. Then point the assistant you already pay for at named pieces of that work.
  • Capture the few good users now.The people who already use AI well have a method in their heads. Write it down as a workflow while they are still in the role.
  • Give every workflow one owner.One person, by name, keeps each workflow current and decides when it changes, the way one person owned the shared spreadsheet.
  • Keep a person deciding.A workflow prepares, checks and drafts. A person approves anything that sends, pays or changes a record.
  • Use what you already have.Workflows run in your company’s own ChatGPT, Claude, Copilot or Gemini, on your account, run by your people. How do we get more value from the AI we already pay for? shows that move.

Five lessons as questions to copy.

Copy these five questions into a document and answer them with your team leads. Each one carries a lesson from an earlier tool.

  1. What is AI for here?Which pieces of work, team by team, should AI prepare, and which stay with people?
  2. Where is our one model?For each piece of work, is there one shared, tested workflow, or does each person use their own prompts?
  3. What are our conventions?What does a person check before AI-prepared work leaves, and what information may go into the assistant?
  4. Is the method written down?For the work we want AI to prepare, have we written down how our best person does it, before choosing any new tool?
  5. Who keeps it right?Which one person, by name, owns each workflow and decides when it changes?

What is different in a company of under 200 people.

A company of under 200 people can move from bought for everyone to decided by the company in weeks, not years. Most published advice on adopting a new tool is written for large enterprises, with committees and a long rollout.

  • You can find the few good users by asking.Your team leads know who uses AI well. Ask them this week.
  • The conventions fit on a page.One page of agreements covers the whole company, the way a short email etiquette once did.
  • The owner is an operator.The person who owns the plan runs operations or the business itself, and knows the work first-hand.
  • One decision reaches everyone.When the leadership agrees what AI is for, every team hears it the same week.

Write the method down.

Every earlier tool rewarded the companies that wrote the method down, and AI rewards them faster. How organizations plan their use of AI sets out the three steps: plan it, develop the workflows, deploy them. FullyPowered AI is a better way to do all three. Free workflow templates give you a starting point, and we develop workflows with your team in a 3-Day Workflow Sprint or a 2-Week Workflow Sprint.

FullyPowered Planner is where the plan 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.