Bring a decision, not a question
Prism works on a choice between options, with something at stake. Open questions get an answer about what Prism can do, not a council.
Two or three options are enough. If your options aren't clear yet, Prism asks one question to get them.
Give it your context
- Nothing set up yet: list the 2 to 4 facts the decision depends on in your message, or Prism runs a "low context" pass and says so in the verdict.
- To stop answering the same questions: keep a context document, any format, with your rates, clients, capacity and current numbers. Attach or paste it before asking.
- No clients yet: the context template has a "Starting out" section for this. Expect "Not yet" verdicts more often; that's the honest answer at this stage.
Keep dated facts current
Facts older than three months get a re-check prompt before the council runs with them.
Read the verdict
Every run ends with one of four outcomes.
Three lines matter more than the verdict
- Main risk: what could make this call wrong.
- What would change this: the fact that would flip the verdict if it turned out different.
- Do first: the one action to take before anything else.
For a client-facing decision, the verdict also includes an objection to prepare for: the pushback you're most likely to hear.
Prism: {your decision}
Verdict: {the call}
Why: {the reasoning in one or two lines}
Main risk: {what could make this wrong}
Objection to prepare for: {the likely pushback}
What would change this: {the fact that would flip it}
Do first: {the one next action}
Look behind the verdict
Ask "show the council" to see:
- each of the five advisors' answers, by name
- both reviews: the ranking of their arguments, and the gaps found in them
- the checker's result, and the first draft if the verdict was revised because of it
- comparable offers found online, if a market check ran
This is where disagreement between advisors shows up, and it's often more useful than the verdict alone.
After the verdict
- Do the "Do first" action within 48 hours, before the decision loses momentum.
- Add the result to your context document, dated, once you know it.
- Keep a short log of what you decided and what happened. Prism can write the entry for you.
Store outcomes, never verdicts
Don't paste Prism's past verdicts back into your context document. Past verdicts bias the next council; outcomes don't.
When it goes wrong
The verdict feels generic
It likely ran with low context. Add the 2 to 4 facts the decision actually depends on and ask again.
It didn't run the council
Your platform may not support the independent-advisor pipeline. Use Claude Code for the full council.
The result says "Reduced Prism"
The council ran in sequence instead of independently, because the platform couldn't parallelize it. The verdict still stands, with less diversity between advisors.
It's slow, or uses a lot of my limit
Five advisors plus two reviews and a check is a real workload. Save it for decisions worth that cost, not everyday questions.
Is it advice I can rely on?
It's a structured second opinion, not a professional one. Legal, tax, and similarly regulated questions get escalated rather than answered.
Ready for your first decision?
Writing, not deciding? Plume writes in your voice.
