Champions
Find the people who make Workstate a habit for everyone else.
Who makes a good champion
A champion is a member of the team who helps colleagues use Workstate well. Look for someone who:
- already uses an AI agent every day;
- is trusted by the team, and already gets asked for help;
- writes things down without being asked;
- is curious about what works, and honest about what doesn't.
Plan on one champion for every 8 to 15 people, and at least one in each team. Champions don't need the Admin role. They need time.
What champions do
- Help people connect. Sit with anyone whose agent can't reach Workstate. Most problems are a key the agent isn't receiving, a missing namespace header or a source that hasn't synced yet.
- Run office hours. 30 minutes a week, where people bring a real question or a prompt that isn't working.
- Collect good prompts. When a prompt works, add it to the team page.
- Keep the ledger useful. Review the week's topics. Look for one decision or finding per topic, a clear summary, citations in the current state, and changed decisions that supersede the old topic instead of contradicting it.
- Watch the sources. Check that each source synced on schedule, and tell an admin when one fails.
- Report back. Send the rollout owner a short update each week.
A weekly rhythm
| Day | Task | Time |
|---|---|---|
| Monday | Review last week's ledger topics in your team's projects | 20 minutes |
| Tuesday | Run office hours | 30 minutes |
| Wednesday | Share one example that worked, with its prompt | 10 minutes |
| Thursday | Check the Source health panel on the Overview page, in each of your team's namespaces | 5 minutes |
| Friday | Send the weekly update | 10 minutes |
That's about 1 to 2 hours a week, including time helping people as it comes up.
Review the week's topics
The console's Ledger page lists every topic in the selected namespace, the most recently updated first, so last week's topics are at the top. Use the repository filter to narrow it to your team's projects.
To review them quickly, give their ids to your agent. The prompt is illustrative:
text
Call ledger_get on each of these topics: [topic ids]. For each one, list the type, status, who opened it and the last entry. Flag topics that cover more than one decision, have no citations, or contradict an older topic without superseding it.The weekly update
text
Week [number], [team]
Connected: [number] of [number] people
Topics recorded this week: [number], for example [a notable one]
Worked well: [one example, with its prompt]
Stuck: [problems, and who can fix them]
Sources: [all synced, or which ones failed]Problems champions can fix
| Symptom | Likely cause | Fix |
|---|---|---|
| Every call fails with 401 | The agent isn't receiving RAG_API_KEY | Follow the client's guide. On macOS, apps started from the Dock don't read shell profiles. |
| Search misses things it should find | The wrong namespace, or a source that hasn't synced | Check the X-RAG-Namespace header, and the source's last sync on Sources |
| Topics duplicate each other | The agent records without searching first | Check that the agent instructions are in place |
| The console says No namespace yet | Nobody has given the person a namespace | Ask an owner or admin to give them one in the Namespaces column on Team |
Recognise the work
Name your champions in the pilot's updates, and make room for the work in their plans. The rollout depends on them.