AI Sales Rep Evaluation

Sales managers can listen to maybe 5% of their team’s calls. The other 95% go unreviewed. This system fixes that by running every call through GPT and scoring it against 13 specific competencies — from product knowledge to objection handling to meeting setup.

How it works

Upload a call recording or transcript. The system transcribes audio with Whisper, evaluates the conversation against the company’s actual sales script and criteria, and returns a detailed scorecard with ratings, specific feedback, and improvement recommendations.

The 13 evaluation criteria span three categories:

  • Client knowledge — Can the rep articulate what the client does, demonstrate product benefits, speak the prospect’s language?
  • Messaging execution — Does the rep follow the script’s discovery flow, uncover pain points, build value, handle objections?
  • Call closing — Does the rep suggest sending information, attempt to schedule a meeting, offer specific times?

Each criterion gets a 1-5 rating with written justification. If a call ends early and the rep never had the chance to demonstrate a skill, the system marks it N/A and adjusts the score accordingly.

Analytics

The real value shows up over time. The analytics dashboard aggregates scores across calls and date ranges, color-coding performance from dark green (4.5+) to red (below 2). A coverage chart reveals which competencies reps are consistently demonstrating and which they’re skipping entirely. GPT generates a summary of the top strengths and improvement areas across any date range.

Call prep

The system also works forward, not just backward. Before a call, reps enter the prospect’s title, industry, and the product being discussed. The system generates the three most likely pain points with how the company’s services address each one — giving reps a concrete game plan instead of winging it.

Tech stack

Streamlit, GPT, Whisper API, PostgreSQL, Plotly. The evaluation criteria, sales script, and company information are all configurable — making it adaptable to different teams and products without code changes.