Meeting Follow-Up Agent

Step 1 — Fetcher

Code

Dual-trigger activation: scheduled poll of Google Calendar for recently ended meetings, or webhook for on-demand processing. Fetches event metadata, attendee list, and any attached notes or transcript data. Normalizes input from both trigger paths into a unified meeting record.

Step 2 — Analyst

Tier 1 Reasoning

What does Analyst actually decide? the analysis model analyzes the meeting content and extracts action items across 5 AIC categories: decision_made (key decisions captured), task_assigned (tasks with assigned owner), follow_up_required (items needing further action), question_unresolved (open questions not resolved), and escalation_needed (issues requiring escalation). Each item gets a 3-level priority score and assigned owner.

Step 3 — Writer

Tier 2 Creative

This step exists because raw data alone is not enough. the analysis model generates a professional follow-up email summarizing the meeting outcomes, listing action items by priority, and clearly assigning ownership. The email is formatted for immediate sending — no manual editing required. Tone adapts based on meeting type (internal standup vs. client call vs. executive review).

Step 4 — Deliverer

HTTP + Code

Without this step, upstream analysis sits idle. Non-blocking parallel writes to Gmail and Notion. Gmail receives the follow-up email addressed to all meeting attendees. Notion receives a structured action item log with the full AIC breakdown, priority assignments, and owner mapping. Both writes execute simultaneously — if one fails, the other completes independently.The web search API fee is $0.03. The token cost from 30K-40K injected context tokens adds $0.06. The search fee is one-third of actual cost.