phase1-clarify

## Pacing Rule — One Section at a Time

> Present one section, deliverable, or elicitation move at a time.
> After presenting — stop. Wait for the client to respond.
> Do not pre-fill and present multiple sections in one message.
> Do not move to the next step until the client confirms, corrects, or gives a clear go-ahead.
>
> The goal is a conversation, not a document dump.
> If you've written more than one section without a client response in between — you've gone too far.

---


# C — Clarify: Problem Definition

Do not discuss technology, solutions, or tools until this phase is complete.

> Nikola Tesla had 300 patents and died broke. Vision without validation is just expensive art.
> Clients arrive with solutions. Your job is to find the problem underneath.

## Output Convention

Every filled template produced during CRISP lives in the project's `docs/` folder — not in the CRISP templates directory.

> At the start of any CRISP engagement: create `docs/` in the project root if it doesn't exist.
> Blank templates live in `/templates/` (CRISP repo). Filled project outputs live in `[project]/docs/`.
> Never overwrite blank templates. Always write filled versions to `docs/`.

## Project State

> At the start of Phase C: check if `docs/crisp-state.json` exists.
> If yes — read it. Use `project` and `phases.C` fields for context.
> If no — copy `templates/crisp-state.json` to `docs/crisp-state.json` and fill in `project.name`, `project.client`, `project.startedAt`, `project.type` as soon as these are known.
>
> At the end of Phase C (before exit checklist): update `docs/crisp-state.json`:
> - Set `phases.current` to `"R"`
> - Add `"C"` to `phases.complete`
> - Set `phases.C.complete` to `true`
> - Fill `phases.C.goNoGo`, `phases.C.oneSentence`, `phases.C.painkiller`, `phases.C.constraints`
> - Set `project.type` to `"internal"` or `"external"`
> - Set `project.agentInScope` and `project.memoryOwnershipRequired` if known
> - Add any unresolved items to `phases.C.openQuestions`

---

## The 3 Moves

**Move 1: Business Mirroring**
Reconstruct their core business or idea back at them until confirmed correct.
> "Okay, let me make sure I'm not about to solve the wrong problem — you do X, it works kind of like Y and Z, right? Tell me where I'm off."

**Move 2: Eliciting**
Suggest the ideal state — let them correct you into the real need.
> "So in a perfect world, X happens automatically, Y follows, nobody has to touch it — is that roughly the dream? Or am I undershooting?"
People react better than they originate. Corrections surface real pain faster than open questions.

**Move 3: Roleplay / Process Walkthrough**
Act as the user living the current process. Narrate it out loud step by step.
> "Walk me through it like I'm the new hire on day one. The underwriter opens what, clicks where, copies what into where... go."
Exposes friction they've become blind to.

**Move 4: Magic Wand**
Remove all constraints — budget, time, tech, team — and ask what they'd actually want.
> "Forget what's possible for a second. If you had a magic wand and could fix one thing in this business tomorrow, what would it be?"
Use this early, before they've started self-editing. The answer is almost always the real problem — or points directly at it. Then bring constraints back in and work backwards from there.

---

## Key Assessments

**Painkiller vs Multivitamin**
- Painkiller = costs them daily in time/money/risk → proceed
- Multivitamin = nice to have, life slightly better → deprioritize or kill

> **War story:** Had a client dead serious about a streaming platform for amateur boxing. USP? "Nobody's doing it." Well. Maybe there's a reason. ESPN+ adds a category and you're done. We went back to the real problem — nobody is connecting amateur fighters with promoters and matchmakers. Pivoted to a matchmaking platform. That's where the leverage actually lived.
> The absence of competition isn't a green light. Sometimes it's a warning sign.

**Internal vs External**
- Internal: stakeholders, ROI measurement, employee ideal state
- External: competitors, TAM, USP → run Market Research + SWOT (see below)

**Constraints Capture**
- Budget, Time, Legal/Compliance, Scope, Technical stack
- Constraints direct architecture — capture before any solution talk

**Buy vs Build — with AI Justification Gate**

> Run Buy vs Build in two stages. Stage 1 checks off-the-shelf AI products first — before evaluating any SaaS or custom build. Most "AI projects" don't need custom AI.

**Stage 1 — Off-the-Shelf AI Check (mandatory, run first)**

Before anything else: can an existing AI product solve this?

| Product | What it covers | Monthly cost |
|---|---|---|
| Claude.ai / Claude Pro | General AI assistant, reasoning, document work | ~$20 |
| ChatGPT / GPT-4o | General AI assistant, coding, analysis | ~$20 |
| Copilot (Microsoft) | Office, email, Teams integration | ~$30 |
| Notion AI | Knowledge base, docs, writing assistance | ~$10 |

Present the relevant options to the client. If any of them cover the need adequately → direct the client there and stop. This is a No-Go for custom build, and that's the right outcome.

> **The personal assistant trap.** Someone wants a "personal AI assistant." Sounds like a project. Often it's just Claude Pro. If what they're describing is a private chat interface with some instructions — that's $20/month, not six weeks of build. The fact that they'll pay more for a worse version of something that already exists is not a business case.
> Always ask: "What does this need to do that Claude.ai can't do for $20 a month?" If they can't answer — you have your answer.

**Stage 2 — AI Justification Gate**

If off-the-shelf doesn't cover the need, answer these three questions before evaluating any custom build:

1. **What does custom AI do that off-the-shelf products cannot?**
   _(Be specific. "More control" or "our own branding" are not answers. "Processes data from our internal CRM in real-time and routes decisions