AI Case Study

A Custom GPT Fed 20 Sales-Call Transcripts Fixed a Dead Sales Doc

← All AI case studies

Sales & CRMEmma
  • “As soon as I made those changes, bam, I put the document back out into the world and it started converting again.” — Emma

The problem

Emma tweaked the sales doc for her high-ticket pronunciation program to test a different copywriting technique, and it stopped converting — even though she knew the offer itself worked, because the previous version had sold. She couldn't see what was wrong with the document.

The AI solution

Emma had been taking notes on every sales call for her €997-2,000 program, writing down prospects' answers in their exact language — how pronunciation problems affect their daily life (getting tongue-tied in presentations, losing rhythm in speeches). She compiled around 20 of these documents and loaded them all into a custom GPT built around her avatar.

The custom GPT worked out patterns across the calls: it pinpointed exactly who the people booking sales calls are and surfaced the customer language they actually use. She then gave it the underperforming sales doc, and it returned specific feedback on what to change. She made the changes, put the document back out, and it started converting again.

She also uses ChatGPT operationally in the same program: after doing a student's assessment report herself, she feeds the reports and results to ChatGPT and has it create the student's syllabus, organizing everything into a logical order.

Tools & costs

Custom GPT (ChatGPT) sales-call notes in customers' exact wording

Results

The non-converting sales doc resumed converting after the GPT-suggested changes; context from the same funnel: 100% show-up rate, 88% conversion rate on calls, and about €20K from the program since launching.

How to copy it

  1. On every sales call, record the prospect's answers in their exact words, especially how the problem affects their daily life.
  2. Compile the call notes (Emma had ~20 documents) and upload them into a custom GPT for your avatar.
  3. Ask it to find patterns: who these people are and the language they use.
  4. Feed it any underperforming sales page or doc and ask what to change.
  5. Apply the changes and re-test the document in the market.

Source

High-Ticket Workshop with Emma (recording) · watch this segment on YouTube (15:13) →