PromenautStep 3 of 5 · Wave 2 to wave 3
From asking an assistant to working with one
Give a team a chat assistant and the early results are easy to see. Drafts arrive in seconds, long documents shrink to a page, and questions get answered without waiting for a colleague.
Then the gains level off. The assistant sits in a separate window. It knows nothing about the system the person is working in, forgets the last conversation, and has no way to act on what it suggests. Each use starts with a person explaining the situation again.
The fourth wave moves the assistant into the tools people already use. A copilot sits beside the document, the code, the spreadsheet or the inbox. It sees what the person sees and suggests as they work: the next line, the formula, the reply.
This is a real step. Less copying and pasting, less explaining, and help at the moment it is needed. For many organisations it is where the measurable value of AI shows up first.
But look at who is doing the work. The person still drives. They decide what happens next, accept or reject each suggestion, and own the result. A copilot speeds a task up; it does not take one on. The gains add up person by person, which is why they stop where the person stops, and why the next step is a different kind of change.
