← Back to Sample Cases
Sample Case 01

A Sales Team Wants AI to Summarize Discovery Calls

A fictional customer asks for “AI call summaries.” The FDE task is to avoid stopping at summarization and instead design a workflow that improves follow-up quality, CRM hygiene, and sales execution speed.

Customer request

“Can AI turn our customer calls into notes, next steps, and CRM updates?”

What they are really asking for

This is not just a summarization problem.

The real customer pain is that sales conversations create too much unstructured information. Important details get lost, follow-up quality varies by rep, and CRM data becomes stale or incomplete.

The solution should not blindly write AI-generated notes into the CRM. It should create a reliable review workflow where the account owner can quickly approve, edit, or reject structured outputs.

Questions I would ask first

  • Where do call transcripts come from today: Zoom, Google Meet, Gong, plain notes, or CRM activity logs?
  • Who reviews the summary before it becomes customer-facing or CRM-visible?
  • What does a good follow-up look like for the sales team?
  • Which fields matter most: pain points, stakeholders, timeline, budget, risks, next steps, or objections?
  • What failure is more costly: missing a key action item, or adding an incorrect one?

Proposed MVP

Transcript in, reviewed follow-up out.

The smallest useful version should process one call transcript at a time and produce a structured draft that a sales rep can review in under two minutes.

Transcript Structured extraction Human review CRM-ready note

Technical architecture

Input

Call transcript, meeting metadata, account name, opportunity stage, and optional CRM notes.

Extraction

LLM extracts structured fields: customer goals, pain points, objections, risks, stakeholders, and next actions.

Validation

Rules check required fields, confidence, missing owner/date, and risky statements that need human review.

Review

Account owner reviews the generated summary before CRM writeback.

Output

CRM-ready note, Slack follow-up, and internal handoff summary.

Customer-facing explanation

“We will start with a safe assistant that drafts call notes instead of automatically updating your CRM. Your sales reps stay in control, but the system removes the repetitive first draft work and makes every follow-up more consistent.”