You ask AI how to tackle a project and get an answer that sounds plausible. Before you act on it, you need to know which details hold up and what the answer takes for granted.
The lines lead from an answer to a check.
Say you want to relaunch your company's website. You tell AI what the site should cover and ask for a plan. It lists the tasks and puts them in a sensible order. You could pass it straight to your team. What it hasn't settled is who writes the copy and who signs off the finished pages.
What the plan takes for granted
The plan may have the copy finished before anyone designs the pages. That works if you already know what you want to say. If you are still debating your offer or your audience, it doesn't. The order rests on an assumption your question may never have touched on.
Go through the plan one step at a time. For each step, note what has to be settled first and whether you told the AI. You can also ask it to spell out its assumptions. Whether they fit your company is something to check with the people who know the work.
Checking the plan
01
Find the assumptions
The plan expects finished copy. That only holds if you have agreed what the site should say.
02
Confirm the details
The plan assumes images and copy arrive on time. Only the people supplying them can confirm that.
03
Weigh the consequences
You want to commit to a launch date. Make sure first that your team can keep to the plan.
The graphic uses the website example. How thoroughly you check depends on what you plan to do with the answer.
What you can verify
Next, test the details the plan depends on. For example, check whether the existing images will work on the new site and whether the copy can be ready by the proposed date. Your records and your team can tell you that; the AI can’t.
There is often more than one sensible way to run a project. So ask why the plan puts the tasks in this order, why the copy comes first and what would slip if you started elsewhere. If the reasoning matches the dependencies in your project, you can adopt that part. Where it doesn't, you and your team have more thinking to do.
If you get it wrong
If you talk the plan through with your team first, they can fill in what is missing, and a mistake costs little. If you promise a launch date, other people will count on it. So check with your team beforehand that the tasks and dates are achievable. Think about who a mistake would hit and what you could still change later.
You can move on once the key assumptions are checked and you can say what is still open. Pass those open points on with the plan, so your team knows what it can rely on.