I Let an AI Agent Named Donna Run My Inboxes for 30 Days: ~85% Triage Accuracy and Three Failure Points
An author running five businesses and receiving 80–150 emails a day handed three inboxes to Donna — an agent on a Mac Mini OpenClaw gateway — for 30 days: checks every 30 minutes, four-way triage, Telegram urgent flags, and drafting routine replies, with a non-negotiable rule of never sending without approval. Triage accuracy was about 85%, roughly 70% of drafts were near send-ready, and about 40% of incoming mail was handled as informational — but tone matching, cross-session context, and CC handling all failed in specific ways. Figures are the author's own 30-day measurements; not independently re-tested by this site. Task: Posted on Mar 8 Thirty days ago I handed my email inbox to an AI agent called Donna and told her to manage it. Not assist with it. Manage it. Triage, flag, draft replies, decide what needed my attention and what did not. I run five businesses and receive somewhere between 80 and 150 emails a day. I was drowning. The experiment felt necessary. Here is an honest account of what worked, what broke, and what I would do differently. The Setup Donna is one of 23 agents in my fleet, running on a dedicated OpenClaw gateway on a Mac Mini. She has access to three email accounts via IMAP and Microsoft Graph API, can read and send emails, and has a memory system that persists context between sessions. The configuration was simple: Check emails every 30 minutes during business hours Triage into: urgent (needs my attention today), important (can wait 24h), informational (FYI only), and noise (unsubscribe candidates) Flag urgent emails via Telegram message Draft replies for anything that looked routine Never send without my approval That last rule was non-negotiable from the start. Donna drafts; I approve. No autonomous sends. What Worked Surprisingly Well Triage accuracy was about 85% right out of the gate. This surprised me. I expected the agent to struggle with context — not knowing which suppliers were critical, which clients were sensitive, which government entities needed careful handling. But the memory system helped. Donna had access to MEMORY.md with relationship notes, and she learned quickly from my feedback when she miscategorized. Urgent flagging was excellent. When a critical email came in — a government demand letter, a supplier holding up a project, a client escalation — Donna pinged me on Telegram within minutes. I never missed an urgent email in 30 days. That alone was worth the experiment. Draft quality for routine emails was genuinely useful. Acknowledgements, meeting confirmations, simple information requests — Donna drafted these at about 70% ready-to-send quality. I would edit a sentence or two and approve. What used to take me 10 minutes of inbox processing took 2 minutes of review. Volume reduction was dramatic. Donna quietly handled about 40% of incoming emails as informational — newsletters, automated notifications, supplier updates I was cc'd on for awareness. She logged them and moved on. I stopped seeing noise. What Broke Tone matching was the biggest failure. I have different registers for different relationships. There are long-term suppliers I am informal with. There are government officials where the language needs to be precise and formal. There are clients who are also friends. Donna defaulted to a professional-but-generic tone that was fine for strangers and wrong for everyone else. I found myself rewriting drafts not because the content was wrong but because the voice was off. The email to my operations manager sounded like it was written by HR. The reply to a contractor I have worked with for three years felt like a form letter. I tried adding tone guidance to MEMORY.md — "with X, be informal; with Y, match their energy" — but this helped only partially. Tone is subtle and relationship-specific in ways that are hard to encode in a config file. Context loss between sessions was a real problem. Donna's sessions are stateless. She reads her memory files at the start of each run, but an email thread that spans three days involves context from multiple sessions — and stitching that together from memory files is imperfect. The clearest failure: a supplier was negotiating a discount on a large order. The negotiation went back and forth over four emails across two days. In session three, Donna had the thread history but missed a nuance from session two — a verbal agreement I had referenced in passing. Her draft for the fourth email ignored it. The draft was not wrong, but it missed something important. CC handling was a mess. Who to include on a reply, when to loop in a colleague, when a one-on-one thread should stay private — Donna was inconsistent here. She erred on the side of inclusion, which meant several drafts had unnecessary CCs that would have been awkward to send. I started adding explicit CC rules to her config, but it felt like playing whack-a-mole.; tools: OpenClaw. Human role: - Setup: initial rules and the four-way triage; Oversight: correcting classification and tone, maintaining MEMORY.md; Review: approving every draft before it is sent; Exception handling: taking over escalated urgent items. - "Donna drafts, I approve, nothing is sent autonomously" is the author's non-negotiable rule. (This field is rebuilt from the full text of the original post; the earlier CMS record's human field was an incomplete sentence.); limitations: - Tone matching was the biggest failure: long-term suppliers, government officials, and client friends each need different tones, and the default professional tone is "right for strangers, wrong for everyone else." - Context is lost between sessions: stateless sessions are stitched together from memory files; in a discount negotiation spanning four emails over two days, a draft missed a verbal agreement the author mentioned in the second email. - CC handling is inconsistent: Donna tends to add people, and several drafts carried unnecessary CCs. - The original post is truncated at "legal or compliance matters" in the record; anything after that point cannot be confirmed by this site. - This site verified only that the source link is reachable; the outcome was not independently re-tested..
Tools
Published on Agent Work Atlas:
Overview
An author running five businesses and receiving 80–150 emails a day handed three inboxes to Donna — an agent on a Mac Mini OpenClaw gateway — for 30 days: checks every 30 minutes, four-way triage, Telegram urgent flags, and drafting routine replies, with a non-negotiable rule of never sending without approval. Triage accuracy was about 85%, roughly 70% of drafts were near send-ready, and about 40% of incoming mail was handled as informational — but tone matching, cross-session context, and CC handling all failed in specific ways. Figures are the author's own 30-day measurements; not independently re-tested by this site.
Workflow
Posted on Mar 8 Thirty days ago I handed my email inbox to an AI agent called Donna and told her to manage it. Not assist with it. Manage it. Triage, flag, draft replies, decide what needed my attention and what did not. I run five businesses and receive somewhere between 80 and 150 emails a day. I was drowning. The experiment felt necessary. Here is an honest account of what worked, what broke, and what I would do differently. The Setup Donna is one of 23 agents in my fleet, running on a dedicated OpenClaw gateway on a Mac Mini. She has access to three email accounts via IMAP and Microsoft Graph API, can read and send emails, and has a memory system that persists context between sessions. The configuration was simple: Check emails every 30 minutes during business hours Triage into: urgent (needs my attention today), important (can wait 24h), informational (FYI only), and noise (unsubscribe candidates) Flag urgent emails via Telegram message Draft replies for anything that looked routine Never send without my approval That last rule was non-negotiable from the start. Donna drafts; I approve. No autonomous sends. What Worked Surprisingly Well Triage accuracy was about 85% right out of the gate. This surprised me. I expected the agent to struggle with context — not knowing which suppliers were critical, which clients were sensitive, which government entities needed careful handling. But the memory system helped. Donna had access to MEMORY.md with relationship notes, and she learned quickly from my feedback when she miscategorized. Urgent flagging was excellent. When a critical email came in — a government demand letter, a supplier holding up a project, a client escalation — Donna pinged me on Telegram within minutes. I never missed an urgent email in 30 days. That alone was worth the experiment. Draft quality for routine emails was genuinely useful. Acknowledgements, meeting confirmations, simple information requests — Donna drafted these at about 70% ready-to-send quality. I would edit a sentence or two and approve. What used to take me 10 minutes of inbox processing took 2 minutes of review. Volume reduction was dramatic. Donna quietly handled about 40% of incoming emails as informational — newsletters, automated notifications, supplier updates I was cc'd on for awareness. She logged them and moved on. I stopped seeing noise. What Broke Tone matching was the biggest failure. I have different registers for different relationships. There are long-term suppliers I am informal with. There are government officials where the language needs to be precise and formal. There are clients who are also friends. Donna defaulted to a professional-but-generic tone that was fine for strangers and wrong for everyone else. I found myself rewriting drafts not because the content was wrong but because the voice was off. The email to my operations manager sounded like it was written by HR. The reply to a contractor I have worked with for three years felt like a form letter. I tried adding tone guidance to MEMORY.md — "with X, be informal; with Y, match their energy" — but this helped only partially. Tone is subtle and relationship-specific in ways that are hard to encode in a config file. Context loss between sessions was a real problem. Donna's sessions are stateless. She reads her memory files at the start of each run, but an email thread that spans three days involves context from multiple sessions — and stitching that together from memory files is imperfect. The clearest failure: a supplier was negotiating a discount on a large order. The negotiation went back and forth over four emails across two days. In session three, Donna had the thread history but missed a nuance from session two — a verbal agreement I had referenced in passing. Her draft for the fourth email ignored it. The draft was not wrong, but it missed something important. CC handling was a mess. Who to include on a reply, when to loop in a colleague, when a one-on-one thread should stay private — Donna was inconsistent here. She erred on the side of inclusion, which meant several drafts had unnecessary CCs that would have been awkward to send. I started adding explicit CC rules to her config, but it felt like playing whack-a-mole.
Reported result
- Triage accuracy started at about 85%; roughly 70% of routine drafts reached send-ready quality; about 40% of incoming mail was treated as informational. - No urgent email was missed in 30 days; the original 10-minute inbox routine shrank to a 2-minute review. - Overall failure-rate statistics: Not disclosed by the source; the three failure points are qualitative records. (All are the author's own 30-day measurements; not independently re-tested by this site.)
Human role
- Setup: initial rules and the four-way triage; Oversight: correcting classification and tone, maintaining MEMORY.md; Review: approving every draft before it is sent; Exception handling: taking over escalated urgent items. - "Donna drafts, I approve, nothing is sent autonomously" is the author's non-negotiable rule. (This field is rebuilt from the full text of the original post; the earlier CMS record's human field was an incomplete sentence.)
Limitations
- Tone matching was the biggest failure: long-term suppliers, government officials, and client friends each need different tones, and the default professional tone is "right for strangers, wrong for everyone else." - Context is lost between sessions: stateless sessions are stitched together from memory files; in a discount negotiation spanning four emails over two days, a draft missed a verbal agreement the author mentioned in the second email. - CC handling is inconsistent: Donna tends to add people, and several drafts carried unnecessary CCs. - The original post is truncated at "legal or compliance matters" in the record; anything after that point cannot be confirmed by this site. - This site verified only that the source link is reachable; the outcome was not independently re-tested.
Try it
I want to set up an email triage agent using OpenClaw on a Mac Mini. I have three email accounts accessible via IMAP and Microsoft Graph API. Please configure the agent to check emails every 30 minutes during business hours. Triage incoming emails into four categories: urgent (needs attention today), important (can wait 24h), informational (FYI only), and noise (unsubscribe candidates). Flag urgent emails via Telegram. Draft replies for routine emails like acknowledgements and meeting confirmations, but never send without my explicit approval. Create a MEMORY.md file to store relationship notes and tone guidelines. Start with a dry run on 10 sample emails, showing me the triage category, draft, and suggested CCs for each. Wait for my feedback before adjusting the configuration. Do not handle legal, compliance, or negotiation emails autonomously. Measure success by triage accuracy and draft edit rate.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.