RepCompass
An AI sales coach that listens to every call and builds the playbook as you close.
Record calls, score reps, surface winning patterns, generate objection libraries and coach each rep weekly — without a dedicated ops team.
WHY IT WORKS
RepCompass plugs into the calling stack and captures every conversation. Transcripts flow through a scoring agent that grades calls against a simple, editable rubric — discovery, qualification, objections handled, next step committed. Over a week, the scores turn into a coaching plan per rep: which call to listen to, which objection to rehearse, which competitor question to prepare for.
The platform also extracts patterns. The deals that close, closed because reps said certain things in certain sequences — RepCompass turns those moments into a living playbook that updates itself every week. New reps onboard from day one against the playbook the best reps have already written without noticing.
It's not another giant CRM: the data model is thin — transcripts, scores, playbooks, deal momentum — and everything else connects to an existing CRM via the same AIM CRM adapter the AI agents already use.
ARCHITECTURE · LAYERED BLUEPRINT
How the microservices wire up.
Read top to bottom. Channels are where people arrive, orchestration is how the AI thinks, domain is the startup-specific logic, data & identity keeps the source of truth, and platform is the invisible glue.
Scroll horizontally to see the full architecture.
INVENTORY
Every block in RepCompass.
Each block is a microservice with a clear role — grouped by layer so you can see what belongs to channels, orchestration, domain, data and platform at a glance.
HOW TO SHIP IT
Quick-start checklist.
- 01 Bridge the dialer / meeting tool to ChatRecorder.
- 02 Upload the scoring rubric; PlaybookStudio seeds from the first 200 calls.
- 03 Wire DealMomentum to the CRM pipeline.
- 04 Publish the weekly coach digest; every rep gets a Monday plan.
WHAT'S NEXT
Compose RepCompass in a visual workspace.
Log into AIM Engine 2.0, drop the blueprint on a canvas, fork the domain microservices you need, and deploy to your own domain. Everything shown here is already a real component.