How to Plan a Project With AI (Even If You Have Never Run One)

Most people who run projects were never trained to run projects. You were good at your job, something needed organizing, and now you own a thing with a deadline, five people, and no plan. The scary part is not the work. It is the suspicion that you have forgotten something important and will find out in week six.

AI is unreasonably good at exactly that suspicion. It has seen how thousands of projects are structured, it never gets bored of listing dependencies, and it will happily tell you what usually goes wrong. Here is the full flow, from blank page to weekly updates.

Step 1: turn a vague goal into a real scope

Projects fail at the start, not the middle, and almost always for the same reason: nobody wrote down what done looks like. Fix that before anything else:

I'm running a project: [describe it in two or three sentences]. Deadline: [date]. People involved: [roles]. Budget or constraints: [anything real]. Before planning, interview me: ask up to 8 questions that a good project manager would ask and I probably have not thought about. One at a time. Then write a one-paragraph scope statement, plus an explicit list of what is NOT in scope.

The out-of-scope list is the most valuable paragraph you will write all project. It is what you point at in week six when someone asks for "one small extra thing".

Step 2: break it into steps with real dependencies

Break this project into phases and tasks. For each task give me: a one-line description, who should own it (by role), a rough time estimate, and what must be finished before it can start. Then give me the critical path: the chain of tasks where any delay pushes the whole deadline. Flag any task where my timeline looks unrealistic and say why.

Ask for the critical path by name. It is the single most useful concept in project management and it takes one line to request: the chain where slipping one day slips everything. Most first-time plans have three tasks quietly sitting on it.

Step 3: find the risks while they are still cheap

This is where AI earns its place. Ask it to be pessimistic, on purpose:

Act as an experienced project manager reviewing my plan with healthy skepticism. Give me the 10 most likely ways this project goes wrong, ranked by a combination of how likely and how damaging each is. For each: an early warning sign I could watch for, and one thing I can do now to reduce it. Include risks that come from people and communication, not just tasks.

Half of the list will be obvious and half will make you wince. The wince is the point: it is cheaper to handle a dependency now than to discover it the week it blocks you. This same skeptic move works on presentations and on any plan you are too close to.

Step 4: write the kickoff so everyone starts aligned

Most project confusion is communication debt taken out in week one:

Write a kickoff message for the team: why this project matters in two lines, what done looks like, who owns what, the key dates, and how we will communicate (where updates go, how often, who to ask when blocked). Under 250 words, friendly and clear, no corporate filler. Then give me a 30-minute kickoff meeting agenda with a question per item that forces a decision.

Run the meeting notes through the minutes prompt and your project has documentation from day one, which nobody has ever regretted.

Step 5: make the weekly update boring

A good status update takes five minutes and prevents most escalations:

Turn my notes from this week into a project update: progress since last week, what is on track, what is at risk (and what I need from whom), and what happens next week. Use a simple status per workstream: on track, at risk, or blocked. Be honest about slippage, do not bury the blocker at the bottom, and keep it under half a page. My notes: [paste]

Consistency beats polish here. The same shape every week means people can scan it in twenty seconds and notice the one thing that changed.

What AI will not do for you

It does not know that the design team is already underwater, that this client always changes their mind at the last review, or that two of your people do not get along. Project management is mostly judgment about humans, and that stays yours. Treat the AI plan as a strong first draft from a competent stranger: complete, sensible, and missing everything only you know. Any dates or figures it produces still need checking, per our fact-checking habit.

Planning your own week is the smaller version of this same skill, and it is lesson one of our free course Everyday Productivity: Get Hours Back Every Week. The office prompt library has the agenda and update prompts you will reuse every week.