# P13 — Task List

**Updated:** 30 August 2026 · All tasks ≤ 1 day. `F-` = founder/business, `C-` = configuration delivery. Estimates are labelled estimates; log actuals in `PROGRESS.md` (hours-per-client tracking feeds the kill criterion).

## Phase P1 — Foundation

| ID | Task | Est. | Depends on | Done when |
|---|---|---|---|---|
| F-01 | Open Retell account (primary, D1); load ≤ $50 credit (O4); buy 1 US number; create Vapi account as dormant fallback | 2h | O4 approved | Test number rings and an out-of-box assistant answers |
| F-02 | Build throwaway test assistant; make 10 varied test calls; record **real measured $/min** from the usage dashboard; log vs the $0.10 planning number | 3h | F-01 | Measured $/min + latency figures in `PROGRESS.md` |
| F-03 | Shortlist 3 pilot clients from the network (O1); run the missed-call audit on each (after-hours + lunch call, save recordings) | 3h | — | 3 names + 3 audit recordings filed |
| F-04 | Produce the before/after demo asset: audit recording vs test-assistant recording of the same request | 2h | F-02, F-03 | 90-second A/B audio ready to play on a phone |
| F-05 | Draft the 1-page pilot agreement (price, minute cap, month-to-month, recording disclosure, rollback promise); billing entity per market (O2) | 3h | O2 | Sendable PDF |

## Phase P2 — Restaurant template v1 (the library's first entry, D4/D5)

| ID | Task | Est. | Depends on | Done when |
|---|---|---|---|---|
| C-01 | Write restaurant prompt pack from `SPEC.md` §3.1: identity/tone, knowledge schema (hours, menu, policies), forbidden topics | 5h | F-02 | Prompt pack v1 in the template repo folder |
| C-02 | Configure call flows: reservation (≤8 turns), FAQ, message capture, escalation triggers (D8), after-hours behaviour | 5h | C-01 | All flows demonstrable on a test call |
| C-03 | Booking integration: Cal.com (or Google Calendar) availability-check + create-booking tools per `API_CONTRACT.md` §2; SMS/WhatsApp confirmation | 4h | C-02, O5 | Test call books a real slot; confirmation message arrives |
| C-04 | Results delivery: call-ended webhook → n8n/Make → daily summary + instant lead alert to a WhatsApp number, per `API_CONTRACT.md` §3–4 | 4h | C-03 | Summary and alert arrive for test calls |
| C-05 | Run the 20-call QA script on the template (`TEST_PLAN.md`); fix failures; repeat until pass | 4h | C-04 | QA sheet all-pass, filed |
| C-06 | Bangla + Arabic feasibility: 5 scripted test calls each; log verdicts in `COMPATIBILITY.md` (D10 — honest, no marketing spin) | 3h | C-02 | Verdict rows updated with date + evidence |

## Phases P3–P6 — Pilot delivery (repeat per pilot; hours logged separately per client)

| ID | Task | Est. | Depends on | Done when |
|---|---|---|---|---|
| C-10 | Pitch meeting with audit recording + demo asset; agreement signed; intake form completed (business facts, calendar, escalation number, tone choices) | 3h | F-04, F-05 | Signed agreement + completed intake |
| C-11 | Clone vertical template → client instance; load business knowledge; set voice/language | 4h | C-05 | Client assistant answers a test call correctly |
| C-12 | Client integrations: calendar hookup, CRM/Sheet lead webhook, owner's WhatsApp for alerts | 3h | C-11 | Booking + lead land in the client's systems |
| C-13 | Client-specific 20-call QA (`TEST_PLAN.md`) incl. owner listening to 3 sample calls and approving voice/tone | 4h | C-12 | QA sheet pass + owner sign-off recorded |
| C-14 | Go-live per `DEPLOYMENT.md`: forwarding rules set (or WhatsApp call button), live smoke calls, rollback rehearsed with the owner | 2h | C-13, `RELEASE_CHECKLIST.md` | Real customer calls flowing; rollback demonstrated |
| C-15 | Handover doc per client (1 page): what it does, how to reach a human, how to read the daily summary, how to pause it, support contact | 2h | C-14 | Doc delivered; owner confirms understanding |
| C-16 | Week-1 watch: daily transcript review + fix list (protocol in `TESTING.md` §2) | 30m/day ×5 | C-14 | Week-1 review notes filed; fixes applied |
| C-17 | New-vertical clone (clinic/salon/real-estate) when the next pilot's vertical differs: adapt prompt pack + flows + QA additions from `SPEC.md` §3 | 6h | C-05 | New template passes 20-call QA on a fictional business |

## Steady state (recurring — these hours are the kill-criterion meter)

| ID | Task | Cadence | Budget |
|---|---|---|---|
| C-20 | Weekly sampled-call QA batch, all clients (`TESTING.md` §2) | Weekly | ≤ 2h total for 3 clients |
| C-21 | Monthly per-client report: minutes used, infra cost vs 40% floor, containment, bookings/leads captured; send client the wins summary | Monthly | 30m/client |
| C-22 | Prompt/knowledge updates on client request (menu change, new staff, holiday hours) | Ad hoc | ≤ 30m each; if a client generates >4/mo, that's a template/self-serve-form gap — fix the template, not just the instance |
| F-10 | P8 gate memo: retention outcomes, hours-per-client actuals, margin table, continue/stop recommendation (`ROADMAP.md` P8) | Once, P8 | 2h |

**Blockers legend:** O-items in `SPEC.md` §6 block where noted; nothing else external blocks P1 — F-01 can start the day this plan is approved.