KAMI PM, your AI project manager
KAMI PM is the AI built into KAMI Tasks. It reads your projects and does the project-management legwork — answering questions, drafting whole project plans, flagging risks, writing the weekly report — while you keep the decisions.
Where: KAMI PM in the main menu You need: access to the projects you ask about; token balance for AI features
When to use this, and when not to
| You want | Use |
|---|---|
| An answer about your projects right now | Ask KAMI PM |
| A brief turned into a ticket plan | Create a Project Plan |
| A regular written summary for stakeholders | The weekly report |
| To be told the moment something happens | An automation — cheaper and immediate |
| A decision made | A person |
Automations and KAMI PM look similar and are not. An automation fires on an event; KAMI PM reasons over the whole picture. Use automations for "tell me when X"; use KAMI PM for "what is going wrong and why".
Ask it things
The chat is the front door: Ask KAMI PM about your projects and sprints — what is at risk, what is stale, what changed. It answers from your actual tickets, not from general knowledge.
Create a project plan
KAMI PM can turn a brief into a ticket plan:
- Choose Create a Project Plan, pick the target project and a target completion date.
- Upload the brief — PDF, DOCX, Markdown, or a screenshot (PNG/JPG) — and plan it with Claude.
- Review the proposed plan, then import the finished plan into the project.
Prefer spreadsheets? Download the import template, fill it in (XLSX, XLS or CSV), and import that instead — the same door without the drafting step.
It runs routines for you
- Weekly Report — a written summary of the week: progress, risks (it says so when no risks were flagged), and what needs attention. Configure weekly report delivery.
- Scheduled runs — routines run automatically on a schedule you set, with repeat rules.
- Thresholds — you decide when a ticket counts as stale (no activity for N days) and when to flag overdue, so its nagging matches your team's rhythm.
It learns your preferences
What KAMI PM Has Learned shows the lessons drawn from your decisions — accept or reject its proposals and future ones bend toward how you actually run things. Everything it does is listed in the Activity Log, and its settings decide whether changes are applied automatically or proposed for approval.
You can give it standing custom instructions, and rename the agent if "KAMI PM" is not what your team would call it.
Tokens and cost
KAMI PM runs on a prepaid token balance, shown on the page with usage. Top up with a preset or custom amount. When the balance is empty the AI features wait until it is not — your tickets and boards are untouched.
What Happens Next
- An imported plan creates real tickets in the target project, which then behave like any others.
- Scheduled runs produce entries in the Activity Log whether or not they found anything.
- With auto-apply on, KAMI PM changes tickets directly and records each change in the log; with it off, it proposes and waits.
- Accepting or rejecting a proposal feeds what it learns, so the first month shapes the rest.
Tips
- Review plans before importing. The AI drafts fast and well, but it has read a brief, not your codebase or your calendar. The import step is the human checkpoint — use it.
- Start with proposals, not auto-apply. Turn on automatic changes once you have watched it make the right calls for a few weeks. It is far easier to grant that trust than to unpick a fortnight of unreviewed edits.
- Set thresholds to your real rhythm. A five-day stale threshold on a team that works in fortnightly cycles produces constant false alarms.
- Read the Activity Log for the first month. It is how you learn what the agent is actually doing, and it is the record if something goes wrong.
- Write custom instructions in specifics. "Flag anything blocking the release" beats "be helpful".
Troubleshooting / FAQ
Q: KAMI PM has stopped responding. Check the token balance. When it is empty the AI features pause; nothing else is affected.
Q: It proposed something clearly wrong. Reject it — that is the feedback loop. Rejections shape later proposals, which is why reviewing early matters more than reviewing later.
Q: The weekly report says no risks, but I know there are some. It reports from ticket state. Risks that live in people's heads rather than in tickets are invisible to it — which is itself a useful thing to notice.