AutomationFrog
Case Study Consultant (10+ calls/day)
🧠

Paste a URL. Walk away. 2.5 hours of admin, gone.

The Impact

"2.5 hours/day → zero."


A consultant running 10+ client calls per day was spending 15 minutes after each one manually processing transcripts. Copying into ChatGPT, running prompts, pasting into Google Docs. We built a system where they paste a Loom URL into a Google Sheet and walk away. Three polished documents land in Drive in under 5 minutes.

The Problem

15 minutes × 10 meetings = a stolen afternoon. Every day.

The client’s post-meeting ritual looked like this:

  1. Open the Loom recording
  2. Copy the transcript
  3. Paste it into ChatGPT
  4. Run 3 separate prompts (action items, topics, summary)
  5. Copy each output into a Google Doc
  6. Manually format the document
  7. Manually add Loom chapters for the client
  8. Repeat. All day. Every day.

That’s 2.5 hours of pure admin work. Time that should have been spent on the actual consulting. The work wasn’t hard. It was just relentless. And it was the kind of work that makes you question your career choices by Thursday.

What We Built

An AI analyst that never calls in sick.

The Trigger

Paste a Loom URL into a Google Sheet. That’s it. That’s the interface.

6-Pass LLM Analysis

Parallel processing on Modal: transcript cleaning + call type classification, action items, discussion topics, leadership summary, Loom chapters, executive synthesis. All six analyses run simultaneously.

Call Type Adaptation

The system classifies each call (Discovery, Strategy, Technical, Status, Training) and adjusts every analysis pass accordingly. A strategy call gets different prompts than a status check. Context-aware analysis, not one-size-fits-all.

Auto-Generated Visuals

AI-generated graphics (Gemini) and technical diagrams (D2) for each meeting. Uploaded to Drive and embedded directly in the documents. The deliverables look like someone spent an hour on them.

Document Delivery

3 formatted Google Docs land in a dedicated Drive folder: the full analysis, a leadership summary, and copy-ready Loom chapter markers. Plus full observability. Langfuse tracking, cost monitoring, and quality scoring for every run.

The Result

The client stopped noticing meetings had happened.

  • 2.5 hours/day reduced to zero minutes of manual processing
  • Paste a URL, get three polished, branded documents in under 5 minutes
  • Every call, every time. No skipped meetings, no “I’ll get to it later”
  • Client’s customers receive formatted deliverables with diagrams and have no idea AI made them

That’s the magical moment. The client’s end customers think there’s a dedicated analyst producing these documents. There isn’t. There’s a frog. 🐸

Ready to build something like this?

Stop doing work a machine should do. Let's find your margins.

Let's Talk