An idea rarely arrives with instructions. It sits in a note, a message or a conversation. The starting point of this method is simple: don’t try to produce more straight away. First pin down what needs to move forward, then let AI prepare a version you can review.
IN SHORT
- Name a result you can review, not an activity.
- Choose a clearly defined context: a few useful sources, not your whole space.
- End with an explicit decision, separate from the preparation.
Name the result, not just the activity
“Working on the launch” describes an activity. “Preparing a progress update to share with the team” describes a deliverable. The difference matters: you can review the second, correct it and decide what happens next.
Write one sentence that states the expected result and who it is for: a summary for your team, a proposal for a client or a list of questions for a meeting. That frame remains your reference point throughout the preparation.
Choose what is really useful
A clear result starts with a clearly defined context. Select the notes, documents and constraints that serve this request. The aim is not to hand over your whole space, but to choose what helps explain the situation.
For a “Shop launch” project, the project plan and the latest meeting notes are enough to start with. A deadline and a decision that is still open can complete that context.
Example request: “Using this plan and these notes, prepare a progress update in three parts: done, blocked, next step.”
Prepare a version you can discuss
The first version is not a decision: it is a draft, something to work from. Ask for a short structure, the points to check and a clear distinction between what is in the sources and what is being proposed.
You can then ask for a specific change: shorten it, change the audience, bring out the open points. A result is more useful when it stays easy to correct than when it looks final.
End with an explicit decision
Review the content, check the recipient and choose what comes next. Keeping a draft, creating a task and sending a message are three different actions: don’t lump them together in a single blanket approval.
The right end point isn’t one more answer. It’s a next step clear enough for everyone to know who does what, and with what expected result.
In Inkore
This method matches the Inkore journey: a request in the conversation, sources chosen for that conversation, a first version in the project, then “Review” for any external action. The document can be exported to Word, PDF or PowerPoint once you are happy with it. See how Inkore prepares the next steps of a project.
Frequently asked questions
How many sources should you give an AI?
As few as possible, but enough to understand the situation: often one reference document and some recent notes. Leave out old versions that should no longer serve as a basis. If something important is missing, ask the AI to say so rather than fill the gap with a guess.
What if the first version isn’t right?
Ask for a specific change rather than a whole new version: shorten it, change the audience, bring out the open points. Each correction keeps a trace of your intention, and the result stays easy to compare with what came before. Small, precise requests usually get you there faster.