Grading your calls with AI coaches: the setup

Pull call recordings from SmrtPhone, transcribe them, and score each conversation against the cold call, lead manager and closer rubrics with quotes.

• The three coach skills, one per seat :
• Cold Call Coach: grades first-touch calls on opener, motivation probing (4 pillars), objection handling, tonality, close quality.
• Lead Manager Coach: grades follow-up and qualification calls on qualifying questions (condition, timeline, motivation, price), 4 pillars depth, objection handling, next-action discipline.
• Closer Coach: grades offer calls on discovery deepening, the money conversation, offer presentation, objection handling, commitment locking.
• The pipeline (skills: cold-call-coach): pull the call log and recordings straight from SmrtPhone (no API exists for this; the engine drives the web session), filter to real conversations (60+ seconds
• How grading works (skills: cold-call-coach): each rubric criterion is scored against explicit 0/3/5 anchors, never overall impression; every score requires a verbatim supporting quote (verified by exa
• The outputs: one coaching report per call plus a per-caller scorecard (average score, trend, the ONE recurring gap, the prescribed playbook drill). An Excel scorecard posts to Slack .
• The rubric provenance: trained on roughly 300 real calls plus the synthetic-audience personas; each call scored with quotes; a new rubric version ships every 2-3 months .
• Transcription cost roughly $0.002 a minute in the July build .
• Gradeability floor: 60+ second calls with recordings (skills: cold-call-coach).
• A live coach run: pull_calls, transcribe, then one full CALL GRADE REPORT with its quotes.
• A per-caller scorecard and the Slack post.
• The rubric file itself, showing one criterion's 0/3/5 anchors.
• The coaching loop: record, transcribe, grade, one fix, re-record, drawn as a closed circle.
• The three-seat rubric map: cold caller / lead manager / closer, each with its 5 graded categories.
• The scorecard trend line: one caller's weekly average climbing with annotations at each "one fix."