# Meeting Intelligence Pipeline

> Audio Transcription to Structured Tasks in 60 Seconds — Recording → AI Analysis → Tasks → Team Notified

**Tags:** Claude API, Zapier, Google Drive, Notion API, Discord Webhook, Bash

## Situation

Meeting action items were getting lost. Participants would discuss tasks, but follow-up depended on someone manually reviewing notes and creating tickets. By the time tasks were logged, context was lost and deadlines were missed. Even getting audio off the recording device was a manual step someone had to remember.

## Task

Build an end-to-end pipeline from physical recording device to structured tasks — starting with automatic file transfer, through AI analysis, to task creation and team notification — with zero manual steps.

## Actions

### 1. Device-to-Cloud Transfer

Zapier bridges the recording device and the processing pipeline — no manual upload needed.

- Plaud AI recorder syncs transcription files automatically
- Zapier detects new files and transfers them to a Google Drive folder
- Orchestrator script picks up new files for processing
- Processed files archived to prevent double-processing

### 2. AI-Powered Analysis

Claude AI extracts structured data from raw transcription text.

- Structured JSON output: meeting name, 3-point summary, action items
- Only genuine action items become tasks — informational items captured in summary
- JSON validation with error handling and graceful skip on failure

### 3. Multi-Service Dispatch

Extracted data simultaneously pushed to task management and team notification channels.

- Notion: task pages with meeting context, status set to "To Do"
- Discord: webhook notification with meeting summary and task count
- Each task carries full meeting context for traceability

## Results

- Action items appear as structured tasks within 60 seconds of transcription upload
- No manual task creation needed — every action item is captured automatically
- Full meeting context preserved in each task
- Connected to Discord Operations Hub for ongoing task management

---

Next case study: [AI Log Triage](https://kcode.dev/en/portfolio/log-triage.md)

Part of the [K Code portfolio](https://kcode.dev/en/portfolio.md). Web version: https://kcode.dev/en/portfolio/meeting-intelligence · Contact: kenishida@kcode.dev