| Platform foundation | Own private repo with the market research, the build spec, and tests on every change. 17 changes merged. |
|---|
| Call scoring engine | Every recorded sales call graded 1–5 on the nine agreed rules — no quote, no score — plus hard counts: talk share, question timing, jargon, booked next step. First clean run done. |
|---|
| Call list, results first | The Calls tab is now one row per call — date, type, score, and where it left the deal — grouped one block per company; grades fold out on click. The missing 6 Aug KAGE / Urban Phenix intro was found in Slack, recorded, and scored, and the full summary of the May Eviction Wizard call is finally in hand. |
|---|
| Pitch practice kit | The timed two-hander (Sean 30s, Fede 90s), Sean's half assembled only from sentences he has actually said on record, plus the objection list. |
|---|
| Camellia case study | The paper-to-AI story with production numbers (23-second replies, 73% of renewals machine-started). Draft — needs Fede and JP-Co sign-off before any client sees it. |
|---|
| Proposal generator | Meeting transcript in, client-specific proposal out. The AI writes only the summary and the played-back section; it cannot touch a number. |
|---|
| Proposal tracking wall | The email gate with view tracking, reusing the pitch-deck mechanism. Built and tested in the product codebase — held for Fede's go before it goes live. |
|---|
| Demo console | This page: team-gated, one stable link, rebuilt from repo data on every change. Made concise on request — facts or a blank, never hedging. |
|---|
| Five verbal targets, verified | Every building counted from county records. Three of five verbal unit counts were wrong; two identity surprises found (the Wiseman contact is the lawyer, not the owner; the sponsor target has an in-house manager). |
|---|
| Master funnel ingested | Sean's 546-company spreadsheet is now platform data. Confirmed the Baron identity, named the Archway contact, killed one bogus row. |
|---|
| Whole list enriched | All 530 remaining rows worked by research agents: 390 companies identified, 92 property-software fingerprints, dead and mislabeled rows flagged instead of guessed. |
|---|
| Ranking engine | Weights and timing windows live as sourced config; every score comes with a written reason per term and honest gates for claimed-not-counted numbers. |
|---|
| Whole funnel tiered | The empty Tier column is filled: 9 companies rankable on counted numbers, 440 scored but held apart as unverified, 95 out of band. |
|---|
| Every active conversation profiled | All 9 contacted-or-interested companies deep-verified — 14 ranked cards on the Prospects tab, each with units, software, people and evidence. |
|---|
| Rejection analytics | Every "no" recorded as structured data with a fixed reason category, so losses can be counted, not just remembered. First entry: the Arizona managers' no. |
|---|
| Prospect pipeline board | Prospects redrawn as a pipeline: eight stage columns grounded in the documented CRM defaults, the not-yet-contacted queue stack-ranked, simple cards that open to full detail. The furthest-along deal now has its own card, alone in Proposal. |
|---|
| App shape decided | The Board (option C) chosen from three researched prototypes: cards move right as deals progress. The console adopts it; the future app builds on it. |
|---|
| Live-call capture research | Zoom auto-record plus transcript webhook recommended; live in-call coaching researched and parked (prompts hurt experienced operators). |
|---|