Use AI to manage a small project: tasks, owners and evidence

A three-person team preparing a workshop rarely lacks ideas. The missing pieces are often who owns each task, which work must wait, and what proves completion. AI can draft the first plan, but the team must confirm its commitments.

This guide uses a fictional online workshop with a registration page, confirmation email and participant materials.

🎯 1. Define the outcome before splitting tasks

Write a short brief covering the event date, intended audience, required deliverables and items outside this release. Mark missing dates or approval owners as unconfirmed.

“Launch the workshop” is difficult to verify. “Registration works, applicants receive the correct email, and the speaker's slides are approved” gives the team observable outcomes.

Do not let AI turn assumptions into commitments. An attractive schedule based on unavailable people creates rework.

🧱 2. Give every task a completion signal

Use these minimum fields: task ID, deliverable, owner, proposed deadline, dependencies, status and evidence.

Three example tasks:

• J01 — Confirm event details: An owns it; complete when the approver confirms the topic, date and description.
• J02 — Registration page: Binh owns it; the layout can begin early, but final copy depends on J01; complete after a successful test submission.
• J03 — Confirmation email: Chi owns it; complete when a test registration receives the right message and links.

“Email copy drafted” and “registration flow tested” are different states. They should lead to different decisions.

👥 3. Assign work against available capacity

Tell AI how much time each person can contribute and when they are unavailable. Let it suggest owners, then ask each person to confirm.

If Binh has two hours available, five extra tasks should not appear simply because he is technically skilled. When demand exceeds capacity, reduce scope, change ownership or revise the deadline. Record the decision on the affected task.

💬 4. A prompt for the initial plan

Draft a plan for an online workshop.
Required deliverables: [list].
Event date and approval owner: [fill in].
People, skills and available time: [fill in].
For each task provide:
ID, deliverable, proposed owner, proposed deadline,
dependencies, completion criteria and evidence.
Separate parallel work from work that must wait.
Do not invent availability or commit the team to dates.
Flag missing information and ask confirmation questions.

Store the approved plan in a shared spreadsheet the team already uses. If work is already in GitHub, Projects supports tables, Kanban boards, roadmaps and custom fields. GitHub Projects documentation. Choose the place people will actually update.

🚧 5. Report changes and blockers

Each update should identify newly completed work, blocked tasks, decisions needed and new evidence. Avoid “80% done” unless the team has agreed what that means.

Compare the approved plan with the latest activity log.
Draw conclusions only from the supplied information.
Return: completed work with evidence; blocked tasks;
decisions for the approver; proposed date changes
with reasons.
Do not mark work complete without acceptance evidence.
Do not send notifications or edit the board.
Draft the update only.

✅ 6. Finish with an end-to-end rehearsal

Before the workshop, have someone act as a participant: open the page, submit details, receive the email and access the materials. Turn problems into owned tasks instead of scattered chat messages.

AI makes the structure of work easier to see; owners confirm what actually happened. A small plan with regular updates and completion evidence is more useful than a long list nobody maintains.