Itzam.ai
All articles

Case study

Case study: AI that turns WhatsApp voice notes into review-ready analysis

By Carlos Adrián Bueso · CTO & Co-founder · · 2 min read

A practice in Mexico received WhatsApp voice notes up to fifteen minutes long. We built a system that transcribes them, analyzes them with the practice's own methodology and makes them searchable with citations.

Client
Practice in Mexico with its own methodology (anonymous)
Channel
WhatsApp
Related services
Customer Support Engine · Business Brain Lab
Status
In production

The challenge

The people the practice works with share their progress through WhatsApp voice notes, sometimes fifteen minutes long. Listening, transcribing and analyzing them with the specialist's methodology was manual work, and each case history was scattered across chats and documents.

What we built

  • A WhatsApp bot that receives the voice note and processes it as soon as it arrives.
  • Transcription in Mexican Spanish and two independent analyses: one of the content, using the practice's methodology, and one of the voice and language (prosody).
  • PDF reports saved automatically to Google Drive, with each case's status tracked in Google Sheets.
  • A private dashboard with semantic search (RAG): the specialist asks in plain language and gets answers with citations that link to the source document.

~100

case records

270+

documents indexed

~4,900

passages searchable with citations

23

versioned methodology prompts

Decisions that matter

  • Human in the loop. Reports are never sent automatically: they go to Drive and the specialist reviews them first.
  • The methodology leads. The 23 prompts live as versioned documents with declared dependencies; no step re-derives what another already validated.
  • The simplest stack that works. At this scale, exact in-memory search instead of a vector database: faster, cheaper and no extra infrastructure.
  • Sensitive data, restricted access. The dashboard requires login and the material is never published.
  • Visible costs and safe deploys. The system estimates the cost before running a full analysis, and every update passes a 6-point health check that rolls it back automatically if anything fails.

What it means for a sales team

The same pattern — WhatsApp as the channel, AI analysis and cited search over your company's knowledge — is the foundation of our Customer Support Engine and Business Brain Lab. If your team lives on WhatsApp, start with the free AI Assessment.