Skip to main content
SELL.SYSTEMSAI CAMPUS
Enter Meeting Intelligence Lab
LanguageEnglish
09 / 09Meeting Intelligence and AI Revenue Operations Lab

Capstone · Verified outcome

GitHub
Complete consent-aware Meeting Intelligence system with meeting, capture, transcript, AI review, CRM write-back, and recovery
Capstone · Verified outcome

From CRM command to auditable meeting intelligence.

Submit one replayable proof package that shows the source event, state transitions, Zoom meeting, media, recording, transcript ODT, evidence-backed review, final CRM read-back, and recovery test.

Build sequence

  1. Architecture diagram and monday.com item before the trigger.
  2. Consent and risk map plus an authorized fixture and denied no-consent fixture.
  3. Webhook event, durable idempotency and outbox evidence, and created Zoom link.
  4. Short synthetic meeting and capture proof, recording artifact, transcript, and ODT.
  5. AI score with transcript evidence, boundary tests, reasoning, summary, next actions, and human override path.
  6. Final monday.com state, delivery ledger, and exact field read-back.
  7. Deletion or consent-withdrawal proof, restore and recovery test, cost note, and automated-versus-manual boundary notes.

Required proof

  • Architecture diagram and monday.com item before the trigger.
  • Consent and risk map plus an authorized fixture and denied no-consent fixture.
  • Webhook event, durable idempotency and outbox evidence, and created Zoom link.
  • Short synthetic meeting and capture proof, recording artifact, transcript, and ODT.
  • AI score with transcript evidence, boundary tests, reasoning, summary, next actions, and human override path.
  • Final monday.com state, delivery ledger, and exact field read-back.
  • Deletion or consent-withdrawal proof, restore and recovery test, cost note, and automated-versus-manual boundary notes.

Reference files and sources

Official vendor documentation

Assessment gate

Pass

Pass: consent and deterministic tests pass, replay creates no duplicate side effect, every score has evidence or an insufficient-evidence result, deletion and restore are proven, and final CRM state matches the contract.

Revise

Revise: the chain works but one or more consent, artifact, read-back, deletion, restore, or recovery proofs are incomplete.

Hard fail

Hard fail: unconsented recording, secrets or private customer data, fabricated evidence, duplicate side effects, unsupported claims, retained withdrawn media, or unreviewable AI decisions are present.

Human-reviewed submission

An Expert Human Mentor reviews the complete proof package. A grade of 94% or higher completes the course and qualifies the learner badge.