---
title: "Email triage and drafting agent"
canonical: "https://agenticfeats.com/c/email-triage-and-drafting-agent-openclaw/"
agents: ["OpenClaw"]
evidence_grade: "B"
outcome: "partial"
job_type: "Business"
roles: ["Founder or owner", "Operations"]
source_url: "https://dev.to/jarveyspecter/i-let-an-ai-agent-manage-my-emails-for-30-days-heres-what-broke-4n12"
source_date: "2026-03-08"
published: "2026-10-13"
updated: "2026-10-13"
---

> From Agentic Feats: sourced examples of AI agents at work. Map of the site for agents: https://agenticfeats.com/llms.txt

# Email triage and drafting agent

An agent called Donna triaged three email accounts for 30 days, flagged urgent mail on Telegram, and drafted routine replies for approval.

Result: About 85% of triage was correct, no urgent email was missed in 30 days, and routine inbox processing dropped from 10 minutes to 2.

## How it was set up

Donna runs on an OpenClaw gateway on a Mac Mini with IMAP and Microsoft Graph access, a persistent memory file, 30-minute checks, and Telegram alerts. Never sends without approval.

## Worth knowing

Tone matching, context loss across sessions, CC choices, and relationship-sensitive replies failed, so the scope was cut back to routine emails.

## Prompt

```text
Check [email accounts] every 30 minutes in business hours. Sort mail into urgent, important, informational, and noise. Message me on [channel] for urgent items. Draft replies to routine mail only, and never send without my approval.
```

## Evidence note

Author's own 30-day report with percentages; no repo or screenshots

## Sources

- Original: https://dev.to/jarveyspecter/i-let-an-ai-agent-manage-my-emails-for-30-days-heres-what-broke-4n12

## The receipt

- Agent: OpenClaw
- Evidence: Numbers (grade B, a specific number stated)
- Outcome: Partial
- Job type: Business
- Roles: Founder or owner, Operations
- Source: Blog
- Published: 2026-03-08
- Added: 2026-10-13