A chatbot is useful to a project manager exactly in proportion to how well you ask. A weak prompt returns a generic textbook list that takes longer to rewrite than to have written yourself. A good prompt returns a draft you can actually use. Below are ten that work — and, at the end, an honest answer about where a chatbot's usefulness stops.
The four parts of a good prompt
- Role and context — who you are, which project, what sector, how large a team.
- Task — exactly one clear output, not three at once.
- Format — table, list, email, length.
- Constraints — what must not be assumed, what is already decided, what the budget is.
Without the fourth, the model invents whatever is missing. One extra sentence up front beats ten corrections afterwards.
Ten prompts that save hours
1. Making sense of the project
"I am a project manager in [sector]. I will describe a project informally. Ask me the 10 questions this project cannot be planned without — ordered by which one is most expensive to leave unanswered."
2. Work breakdown structure
"Break this project into a deliverable-based WBS with three levels. Every work package must be 8–80 hours. Return a table: level 1, level 2, work package, estimated effort." (See the WBS guide.)
3. Finding risks
"List the 12 risks most likely to derail this project. For each: probability, impact, an early warning sign and one concrete mitigation. Do not include generic risks such as 'the schedule may slip'."
4. Challenging estimates
"Here is my schedule with durations. Play devil's advocate: where are the estimates most likely too optimistic, and why? For each, name the assumption that has to hold for the estimate to be correct."
5. Status report
"Draft a one-page status report from the notes below. Structure: overall status plus reasoning, milestones, budget, completed deliverables, top 3 risks, what I need from management. Maximum 250 words."
6. Meeting notes into decisions
"Extract three separate lists from these notes: (1) decisions made, (2) actions with owner and deadline, (3) open questions. Add nothing that is not in the text."
7. A difficult email
"Write an email telling the client about a two-week delay. Tone: direct, not apologetic. Include the cause, the new date, what we have already done and what we need from them. Under 150 words."
8. Budget sanity check
"Here are the project budget and the work plan. Find inconsistencies: costs with no activity, activities with no cost, costs arising before the activity that causes them. Return only the inconsistencies found."
9. Stakeholder map
"List the likely stakeholders for this project, including the ones usually forgotten. For each: interest, influence, what they want from the project, what would upset them, how often to update them."
10. Closing retrospective
"From these notes, formulate 5 lessons learned. Each must be a concrete rule that fits on a checklist, not a general recommendation. Format: situation → conclusion → rule for the next project."
Three mistakes that ruin prompts
| Mistake | Result | Fix |
|---|---|---|
| Question too broad | A generic textbook list | Give context: sector, size, constraints |
| Several tasks at once | Everything answered shallowly | One prompt, one output |
| No format specified | Essay-style prose | Demand a table, columns, a word count |
Where a chatbot's usefulness stops
Every prompt above produces text. But project management is not text — it is connected data. A chatbot does not know what changed yesterday, does not update the budget when a deadline moves, does not hold risk status and does not remember why something was decided three months ago. Every answer starts from zero and every output has to be copied somewhere by hand.
That is where the saved time leaks back out: the draft takes five minutes, connecting it to the real project takes an hour. More on that gap in why ordinary AI struggles with project management.
How Projektiassistent closes the gap
In Projektiassistent the AI is not a separate chat window but part of the project. Plan, schedule, budget and risks come from the same input and stay connected — change one and the related parts change too. Decisions and changes stay in a log, so context is not lost. And What-If lets you see what moving a deadline or scope does to everything else before you commit. See: AI in project management.
Summary
Good prompts give a project manager fast drafts — use them. But a draft is not a project. If you want AI output to land in the plan, the budget and the risk register without copy-paste, you need a tool that actually knows your project.
Try it: projekt2.projektiassistent.ee.