Using AI to Coach Sales Reps on Discovery

By Robert Koehler, Scale GTM AdvisorAccess the full prompt hereWorking in the trenches with sales teams, I’ve found that most reps still struggle to conduct meaningful discovery and uncover underlying business context, personal stakes, and internal friction that drive real decisions. Discovery is where good sellers build the most leverage and value. We win or lose deals here. Yet despite the mountain of call recordings we now store in systems such as Gong, Chorus, and Attention, sales managers often spend little time reviewing them with their sellers. The result is that most sellers rarely receive coaching, and when they do, it’s delivered via one-off, unreinforced training sessions.This state of affairs calls for a new, AI-powered approach.

Building evaluation criteria based on best practices

AI (in this case, ChatGPT) can be an ideal assistant to increase the speed, quantity, quality, and consistency of discovery analysis, scoring, and coaching. When I first set out to create an AI sales coaching tool, my near-term goal was to use it to increase awareness of strengths/gaps and close specific discovery skill gaps for sellers, managers, and leaders. Longer-term, I hoped to improve stage-to-stage conversion, boost win rates, and shorten post-discovery time-to-close for our portfolio companies.Based on several years of experience reviewing and scoring discovery calls, I partnered with ChatGPT to critique and refine a call evaluation scorecard using documented best practices (e.g., Challenger, SPIN, MEDDPICC; experts like Dave Brock, Keenan, Kyle Norton).The core prompt included (access the full sales prompt here)

• Context: “You are an expert sales coach…
• Evaluation model: discovery broken into clear, observable steps aligned to best practices.
• Scorecard: 1 point per step; total possible equals 100%. (attached to the prompt)
• Output: a concise two-page report with seller/account/buyer roles/dates, a headline summary, Top 3 strengths, Top 3 areas to improve, and coaching recommendations.

And here is the 20-point scorecard I put together:CategorySub-categoryDescriptionPointsOpening & Framing(2 pts)Rapport & Icebreaker.Brief, natural open0.5Meeting Intent & AgendaClear purpose, agenda, outcome1.0Prospect’s Reason for InterestTrigger for the meeting0.5Problem Discovery & Context (7 pts)Business Goals & OutcomesTarget outcomes/metrics1.5Current ApproachProcess/tools/workflow today1.0Pain Points & ChallengesFriction, inefficiencies, risks1.5Business Impact / Quantification Time, cost, errors, risk1.5Alternatives / CompetitionWho/what else they’re considering0.5Evaluation CriteriaWhat matters, how they’ll decide0.5Priority & TimelineUrgency, sequencing, timing0.5Stakeholder & Decision Process (3 pts)Prospect’s RoleRole in the decision0.5Decision Process & StakeholdersWho/how/governance1.5Multithreading OpportunityPathways to execs/business buyers1.0Positioning & Value Alignment (4 pts)Tailored PositioningAfter discovery, aligned to outcomes1.5Value vs. FeaturesBusiness impact over feature dump1.0Relevant InsightsPatterns from similar companies/personas1.0Active Listening & Mid-Call Recap 0.5Closing & Next Steps (4 pts)Debrief QuestionHow does this compare to today1.0Recap & AgreementConfirm takeaways1.0Clear Next StepAligned to their buying motion1.0Schedule Next Step Live1.0Of course, you can adjust the scorecard weighting to your organizational preferences. My model gives ‘Problem Discovery and Context’ the heaviest weight because missing the pain, impact, and decision dynamics kills conversion. The “Stakeholders & Decision Process” elements might be more important for enterprise selling where multiple stakeholders are key.The prompt generates three report types (sample output report here):

1. Individual call discovery analysis
2. Individual seller analysis across multiple calls
3. Team-level discovery report for the CRO

Truly scalable coaching has never been easier

I implemented this approach at one Scale portfolio company and immediately saw massive changes:

• Quantitative changes (before → after):
  • Calls evaluated (scored + documented): 12 → 50
  • Cycle time (eval + report): 2–3 weeks → 2 days
  • Follow-on coaching sessions: 1 → 8
  • Time to enablement recommendations: 3.5 weeks → 2 days
• Qualitative changes:

• • Speed, structure, objectivity: A clear rubric and consistent scoring raised trust in the coaching.
  • Seller experience: Live, face-to-face coaching focused on 3 priorities for improvement max. Sellers felt a real investment and asked for more.
  • Behavioral shifts: Several sellers cut talk time 10–15%, asked more/better questions, and listened more.
  • Gamification: Sellers checked scores after calls and self-assessed.

We also walked away with a couple of learnings. First, humans in the loop remain essential. Managers should review summaries and recommendations to match company language and culture. Second, managers should focus on trends and scoring across multiple calls. In general, using relative movement and comparative performance to guide enablement is much more effective than focusing on the absolute figures.

Next steps to get started

1. Automate the workflow. Integrate the master prompt and LLM with your recording platform (Gong/Chorus/Attention) and your CRM so scoring/reporting is daily and automatic—shrinking the gap from call to coaching.
2. Deeper questioning analysis. Use the follow-up prompt that we developed to break down question types and quality (single open-ended, expansion prompts, reframes) and coach to ask better questions.
3. Evidence-based enablement. Build training from real call examples—what good looks like and what to avoid—aligned to the top three team gaps.

For help standing up AI sales coaching for your team, or access to the deeper questioning analysis, contact Robert Koehler (Scale GTM Advisor).