AI-powered Scrum Master Starter Kit
The AI-powered Scrum Master Starter Kit is one text file your AI assistant keeps current, a toil log, plus a handful of things you say to it. It takes about twenty minutes to set up. The assistant does the typing, and you keep the judgement. An AI-powered Scrum Master automates the toil by partnering with AI to improve the system the team works in.
Before you start
You need somewhere your AI assistant can read and write a text file: a synced folder, a repository, a notes workspace or a project in the assistant you already have. Copilot or Claude both work.
You don't need to be technical. You need to be willing to talk to it for two minutes at the end of the day.
Setup
Paste the block below into your AI assistant. Answer its questions. It will create the log.
I'm a Scrum Master and I want to keep a toil log that you and I maintain together. Before creating anything, ask me these questions one at a time and wait for each answer: 1. What team or teams do I support, and how long are our sprints? 2. Which events and recurring tasks take most of my time each week? 3. Where does the team's work live: which tracker or board, which chat? 4. When in the day can I spend two minutes on a spoken reflection? Then create one file, toil-log.md, with: - a short header holding my answers; - a table: ID, date, what took my time, toil or judgement, SPOT (S/P/O/T passed), outcome I want, status. Rules for how you maintain it: - Every row gets a stable ID that never changes: T-001, T-002 and so on. - Status is always one of: logged | passes SPOT | building | automated | dropped. - Toil is work that needs doing but doesn't need me. Judgement is anything about people, priorities or trade-offs. If you're not sure which, ask me. - SPOT means Simple (clear steps, no debate), Predictable (same inputs, same outcome), Observable (you can see when it's done right), Tedious (it repeats and drains me). Only mark an item "passes SPOT" when I agree it passes all four. - Never invent a date, an owner or a time saving. Write TBD and tell me. Show me the empty log before you write it.
The toil log template
If you'd rather start the file yourself, copy this. The two rows are examples: one item that passes all four SPOT checks, and one that is judgement and stays with you.
# toil-log.md <!-- The two rows below are examples. Delete them when you add your own. --> | ID | Date | What took my time | Toil or judgement | SPOT | Outcome I want | Status | |---|---|---|---|---|---|---| | T-001 | 2026-11-30 | Chasing board updates before standup | toil | S P O T | Before standup, a list of items that haven't moved in three days, with who touched them last | passes SPOT | | T-002 | 2026-11-30 | Deciding whether to raise a tension in the team | judgement | S - - - | (stays with me) | dropped |
Check the empty log before it writes anything. If a column doesn't make sense to you, change it now. A log you understand is one you keep using.
The moves
These are the things I say to my AI. Say them in your own words. The assistant does the writing.
Daily
"Log my day."
End of every day. Talk for two minutes, then: "Sort what I just said into toil and judgement. Add the toil to my toil log." Use your voice. Talking keeps the detail that typing filters out.
"SPOT this."
When something feels like toil: "Check this task against Simple, Predictable, Observable and Tedious. Tell me which fail and why."
When something passes SPOT
"Describe the outcome."
"Help me write what done looks like for this. The result, not the steps."
"Build it with me."
"Help me turn this into a saved prompt, a scheduled flow or a small script."
Partnership
There are four levels of questioning: telling, leading, powerful questions, and triggers and hooks. These two moves work at the fourth.
"What do you notice?"
Before an event: "Look at this board export. List patterns worth noticing. Observations only, no conclusions." This is the trigger.
"Give me a hook."
"Turn this into three open questions for the team. No telling, no leading." The hook stays human. I pick the one that fits these people.
Weekly
"What is stale?"
Friday: "List the toil items older than a week. For each one, ask me: build it, keep it or drop it." Answer all of them.
What my AI never does
Write this down before you automate anything.
The assistant drafts, sorts, notices and builds.
The assistant does not:
- Decide what counts as judgement. I do.
- Speak to the team for me. I post, I ask, I coach.
- Invent an owner, a date or a decision to fill a gap.
- Read team conversations without the team's consent, and without saying what it reads and why.
- Own the team's effectiveness. That stays with me.
When it goes quiet
The failure is the same as with any log: you stop talking to it, items pile up, and nobody builds anything. It doesn't break loudly. It goes quiet.
Two fixes. Keep the log to one file. Then put "What is stale?" on your own calendar as a standing fifteen minutes on Friday.
From the session
What is an AI-powered Scrum Master?
The session kit from Give Thanks for Scrum, November 24th, 2026: the handout, the slides and steps to try this week.