The short answer.

Every piece of work you do with AI goes through four stages: Ask, where someone types it into the chat box each time; Save, where it is written down once so it runs the same way every time; Share, where the whole team runs the same one; and Sustain, where one person owns it, updates it and checks that it still works. Nearly everyone is at Ask for nearly everything. The value of the AI a company already pays for grows at each stage above it.

The ladder describes one piece of work, such as a weekly report, a reply to a common complaint or a renewal summary, not a whole company. Different pieces of work in the same team sit at different stages, and that is normal.

The difference between typing numbers and building the model.

Using AI at the Ask stage is like using a spreadsheet to add up a column of numbers you typed in. It works, and it is gone when you close the file. Using AI at the Sustain stage is like the financial model your team runs every month: the formulas are set, the inputs arrive in the same place, one person keeps it right, and everyone trusts the output because it is built the same way each time.

Nobody learned to build that model on their first day with a spreadsheet. People moved from typing numbers to building models one habit at a time. The good news is that the same move with AI is easier to learn. A saved set of instructions is written in plain language, not in formulas, and anyone who can explain a piece of work to a new colleague can write one.

The four stages, one at a time.

Each stage keeps everything from the one below and adds one thing.

  1. Ask.Someone opens ChatGPT, Claude, Copilot or Gemini, types a request from memory and works with the answer. The result depends on who asked, how they phrased it and what they remembered to include. Ask is the right way to start and the right place for work that changes shape every time. The four asks covers how to get a better result from a single request.
  2. Save.The request that worked is written down as a workflow: the instructions, the inputs it needs, the format of the output and the checks a good result passes. The next time, the person starts from the workflow, not from a blank box, and the work comes out the same way. Much of what makes a workflow good is already in someone’s head; FullyPowered AI Studio is how we capture that know-how and write it down.
  3. Share.The whole team runs the same workflow, with the same checks, whoever is at the desk. Share is where a plan is needed, because a team has to agree which pieces of work get a shared workflow, in which order, and which version is the current one.
  4. Sustain.One person, by name, owns the workflow. That person updates it when the work changes, keeps a record of what it gets wrong, and checks on a regular date that it still does what the team needs. A sustained workflow can be designed to pull from the systems the work lives in, such as your accounting system or your ticket queue, with a person approving what goes back.

At every stage a person decides. The workflow prepares the work; the person who owns the work reads it, changes it and sends it.

One piece of work climbs the ladder.

Take a 40-person insurance brokerage, and one piece of work: the summary each account manager sends a client before a policy renews.

Illustrative example.

  • Ask.Each account manager pastes the current policy and the insurer’s renewal terms into an assistant and asks for a summary. One asks for bullet points, another for a letter. Some mention the changes in cover, some forget.
  • Save.The senior account manager writes down the version that clients liked best: which documents go in, the five headings every summary carries, the changes in cover listed first, and a check that every figure matches the insurer’s terms. She uses it for every renewal.
  • Share.The managing director puts renewal summaries on the brokerage’s plan as the account team’s first shared workflow. Every account manager runs the same one, and new joiners learn it in their first week.
  • Sustain.The senior account manager owns it by name. When an insurer changes the layout of its renewal terms, she updates the workflow the same week. Once a quarter she reads the record of corrections the team made and decides what to change.

The account managers still read every summary and decide what goes to the client. What changed is that the brokerage now has one way of doing it, and one person who keeps that way right. What AI does inside the work, gathering, checking, drafting and the rest, is covered in The five ways AI helps with a piece of work.

A one-page version to copy.

Copy this onto one page. For each stage: what it looks like, the question to ask, and the move to the next stage.

  1. Ask.What it looks like: someone types the request from memory each time. The question: does this piece of work come back, the same way, every week or month? The move: write down the request that worked best, with its inputs, its format and its checks.
  2. Save.What it looks like: one person starts from written instructions and gets the same kind of result each time. The question: would the rest of the team do this work better from the same instructions? The move: put the piece of work on the team’s plan and agree one version.
  3. Share.What it looks like: the whole team runs one workflow, and new joiners learn it. The question: who notices when the workflow gets something wrong, and who fixes it? The move: name one owner, set a date to review it, and start a record of corrections.
  4. Sustain.What it looks like: one named owner keeps the workflow current, reads its record and reviews it on a date. The question: is the workflow still doing what the team needs? The move: keep the date, and look for where it can draw from the systems the work already lives in.

Now place five of your own pieces of work. Write each one down, then the stage it is at today and the next move.

  • 1.Piece of work: ____________ Stage today: ________ Next move: ____________
  • 2.Piece of work: ____________ Stage today: ________ Next move: ____________
  • 3.Piece of work: ____________ Stage today: ________ Next move: ____________
  • 4.Piece of work: ____________ Stage today: ________ Next move: ____________
  • 5.Piece of work: ____________ Stage today: ________ Next move: ____________

Most people find all five at Ask. That is the right answer to write down, and the move from Ask to Save is the one to make this week.

One ladder at three sizes.

Ask, Save, Share, Sustain is the ladder for one piece of work. The same ladder applies to a whole company and to one person. The stages of AI at work is the same ladder for a company: from seats switched on and people experimenting, through a plan, to workflows built and managed across every team. The levels of using an AI assistant at work is the same ladder for one person: from asking one question at a time to running and keeping a set of workflows of their own.

A company moves up its stages as more of its pieces of work move up theirs.

What is different in a company of under 200 people.

In a company of under 200 people, one piece of work often climbs the whole ladder inside one team, with people who sit together. A large company needs a programme to move a piece of work from one person’s desk to every desk. You need one conversation.

  • The person who saves it often owns it.The senior person who writes the first version is usually the right owner at Sustain. Name them early.
  • Share needs a plan, even a small one.Fifteen people can still keep fifteen versions. One page that says which work is shared, and who owns each piece, is enough.
  • The AI is already paid for.The assistants your people use today are where these workflows run, on your own account, so each move up the ladder adds value to what you already have.

Share is where the plan comes in. FullyPowered AI Planner is where the plan lives: you set out your teams, their roles and the work between them, choose which pieces of work to save and share first, and your colleagues add their own teams. How organizations plan their use of AI covers the whole method.