# Flow 6 — Land · Interview Agents   ·  ~12 min  ·  Oli=Chrome · Oscar=Firefox
> Three AI agents prepare you to land the job: **Coach** (scored practice on your gaps), **SME** (deep-dive any topic), **Trainer** (structured training). Each has a distinct persona + voice; the questions must target *your* selected gaps.

**▶ Start here**
- **Links (three agents, three cards under "Land"):**
  - **Coach / Trainer:** `…/?startflow=interview`
  - **SME deep-dive:** `…/?startflow=sme-expert`
  - **Train on a topic:** `…/?startflow=sme-topic`
- **Lands on:** the **sign-up wall** (if logged out) → the **interview agent chat**.
- **Voice:** **the interviewer speaks.** **System sound ON**, allow the **mic** when asked. Answer by voice or text; the readiness **score climbs live** as you go.

## Preconditions
- [ ] **P1.** Logged in with a résumé loaded. For Coach: have run **Quick Match** and **selected some gaps** (Flow 5).
- [ ] **P2.** **Sound on**, allow **microphone** when asked.
- [ ] **P3.** DevTools → Console open.

## Stage 1 — Coach (after Quick Match) ⭐ (P0)
| # | Action | Expected result |
|---|--------|-----------------|
| 1 | From a Quick-Match result with gaps selected, start **Practice Interview** (Coach). On prod this is gated → see **Flow 7 (Payments)**. | The Coach agent appears with a **persona** (a name/face, not a brain emoji) and greets you. |
| 2 ⭐ | Answer questions. | Questions target **your exact selected gaps** — 5 per gap (3 gaps → 15). They are not generic. |
| 3 | Listen + speak. | The agent **speaks** (voice + a speaking glow on the avatar); your **mic** captures your answer. |
| 4 | Finish the session. | A **score / readiness** + a **cheat-sheet** you can download. |
| | | **RED FLAG** | questions ignore your gaps; voice/mic dead; no score or cheat-sheet at the end. |

## Stage 2 — SME (deep-dive any topic) ⭐ (P0)
| # | Action | Expected result |
|---|--------|-----------------|
| 5 | From **Land → SME**, give it a topic (e.g. "Basel III for a risk director"). | A short intake, then the **SME persona** (the "expert") deep-dives the topic — explains, then probes. |
| 6 ⭐ | Let it run a few turns. | It does **not** trip mid-explanation or cut itself off ("take your time" while it's still talking). It finishes its thought. |
| | | **RED FLAG** | the SME stops mid-sentence, repeats, or the voice cuts on background noise. |

## Stage 3 — Trainer (structured training)
| # | Action | Expected result |
|---|--------|-----------------|
| 7 | From **Land → Trainer**, pick an industry/role/topic via the wizard. | Structured training begins; it teaches the topic and checks understanding. |
| 8 | Reach the end. | A learning score; option to continue to the next subtopic; a cheat-sheet. |

## Stage 4 — Personas are distinct
| # | Action | Expected result |
|---|--------|-----------------|
| 9 | Compare the three agents. | **SME · Coach · Trainer** feel different (greeting, face, voice). Voice **gender matches** the face. |
| | | **RED FLAG** | all three feel identical, or a voice/face mismatch. |

## Stage 5 — Language (optional)
| # | Action | Expected result |
|---|--------|-----------------|
| 10 | Set the interview language to a second language (e.g. Spanish) and run a few questions. | Questions, voice, and the agent's prose are in that language; the app chrome stays English by design. |

## Report
🐞 **Found a bug?** any screen — or **Submit a run**: step # · browser · expected · actual · screenshot. ⭐ = P0. Voice/mic issues: note your OS + browser.

## 🎥 Video (what to record · ≤3 min)
Record: from a Quick-Match with gaps → start Coach → show questions hitting your exact gaps + the agent speaking + your mic → finish with score + cheat-sheet. Then a quick SME deep-dive on a topic. Narrate "it preps you on exactly what you're weak on."