How it uses Jev
The call coach transcribes a live sales conversation in the browser and sends the recent transcript to Jev after each sentence. Jev returns a next-action Choice, a buying-stage Score, and seven Noul signals; local decision code applies confidence, smoothing, and hysteresis settings to display coaching guidance.
What Jev decides
Illustrative next-action questionExample answers · not a recorded Jev response · Source ↗
Question 1 · next best action
YOUR APP
INSTRUCTION
Choose the sales rep’s next action based on the customer’s most recent transcript statements.
STATE
The request state contains `sales_call_transcript`, populated with the recent transcript turns. No particular call transcript or observed answer is claimed.
JEV · CHOICE
- send_pricing
- book_demo
- loop_in_decision_maker
- handle_objection
- send_proposal
- ask_discovery_question
- nurture_follow_up
- no_action
App workflow
- Capture and transcribe speech
Use browser speech recognition to produce a shared transcript for the call.
- Request call judgments
Send the recent conversation turns to Jev for next action, buying stage, and signal assessments.
- Apply local coaching rules
Use confidence thresholds, smoothing, and hysteresis to stabilize the decision and choose matching playbook guidance.
- Show the rep what to do next
Display the action and confidence in the call-coach interface; the README limits the request to the latest 40 turns.
Why it is interesting
It uses a single structured request to produce several kinds of sales-call judgments, while a separate playbook maps decision keys to coaching tips. The user remains aware that browser speech recognition and conversation text are sent to external services.