01 the business
Residential solar installer
40 staff, ten salespeople
The sales manager means to listen to calls but rarely finds the time. New reps learn by guessing, and the same objection about financing gets fumbled for months. Coaching depends on the one call the manager happened to overhear.
02 before
The problem
Sales managers mean to listen back to calls but rarely find the hour. New reps learn by guessing, and the same objection gets fumbled for months. When coaching does happen, it rests on the one call the manager happened to hear.
- The manager reviews perhaps one call per rep per month.
- New reps take months to handle common objections well.
- Agreed next steps from calls often never reach the CRM.
03 the build
What the AI does
- Step 1: Pulls new call recordings from Zoom, Microsoft Teams or your phone system once each call ends.
- Step 2: Transcribes the call and marks where the rep asked about budget, timing and decision makers.
- Step 3: Compares the conversation with your written playbook and notes which questions were skipped.
- Step 4: Writes a short coaching note for the rep with quotes from the call, not a score out of ten.
- Step 5: Logs the next step the customer agreed to against the deal in HubSpot or Salesforce.
- Step 6: Groups the objections heard across the team each week so the manager can plan one focused session.
paused: waiting for a person to approve
How it was put together
- Pulled recordings from the phone system and video calls as soon as each call ended.
- Compared each call with the written playbook: budget, timing, decision makers and financing.
- Wrote a short coaching note with quotes from the call instead of a score.
- Grouped objections across the team each week for one focused training session.
04 human in the loop
Where a person stays in charge
The sales manager reads each coaching note before the rep sees it and can edit or drop it. Nothing is used for pay or performance decisions unless the manager chooses to.
- Callers are told calls are recorded, as the law requires where you operate.
- The manager reads each note before the rep sees it and can edit or drop it.
- Notes are not used for pay or performance decisions unless the manager chooses to.
05 rollout
From kickoff to live in 4–5 weeks
- Week 1Turn the playbook into a checklist the system can test against.
- Weeks 2–3Generate notes for the manager only and calibrate the tone.
- Weeks 4–5Share notes with reps and start the weekly objection session.
06 after
What changes
Every rep gets feedback on real calls, not just the ones a manager overheard. Managers spend their coaching time on patterns across the team instead of hunting for recordings.
Assume ten reps making 40 recorded calls a week each, and a manager who can review two calls an hour.
10 × 40 = 400 calls a week, which would take the manager 200 hours to review. With notes, every call gets feedback and the manager reads summaries in about 2 hours.
What to measure
- Share of calls with a coaching note.
- Playbook questions covered per call over time.
- Ramp time for new reps.
- Close rate by rep, before and after.
07 evidence
What the research says
Across 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and lower-skilled agents.
NBER, Generative AI at Work (Brynjolfsson, Li and Raymond) (opens in a new tab) · 2023
AI used to recruit or select people, or to make decisions about promotion, termination, task allocation or performance, is classed as high-risk.
EU AI Act, Annex III: high-risk AI systems (opens in a new tab) · 2024
08 lessons
Mistakes to avoid
- Scoring people. A number invites arguments; quotes invite learning.
- Skipping consent. Recording rules differ by state and country.
- Coaching on everything at once. Pick one or two skills a month.
09 questions
Common questions
- Does it replace the manager?
- No. It finds the moments worth discussing so the manager's time goes on coaching, not searching.
- Can we use it in the EU?
- Yes, with care. Using AI to make decisions about performance or promotion is classed as high-risk under the EU AI Act, which is why notes stay advisory.