Claude Code prepares a reviewed dawn reply queue
Simonnet reports a weekday Claude Code agent that identifies inbox threads awaiting him and saves bilingual draft replies, which he edits and sends himself.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
At [weekday time], find threads truly awaiting my response. Exclude copied-only, already answered and teammate-owned mail. Draft replies in [my style] and the thread language, flag missing facts, and save drafts for my editing and sending.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Cyril Simonnet
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Simonnet reports a weekday Claude Code agent that identifies inbox threads awaiting him and saves bilingual draft replies, which he edits and sends himself.
- Human involvement
- Simonnet edits and sends all replies. The agent has no sending role in this workflow, and unavailable figures remain unresolved rather than invented.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Schedule inbox review for 5am on weekdays.
- Exclude copied-only, answered, informational and teammate-owned threads.
- Prepare concise replies in each conversation’s language.
Cyril Simonnet describes an agent that processes his inbox at 5am on weekdays, before he starts work. It looks for threads genuinely awaiting his response and excludes messages where he is only copied, has already replied, or a teammate has responsibility. Informational messages also stay outside the draft queue.
For selected threads, the agent prepares a reply in Simonnet's concise style, using French for French conversations and English for English ones. It knows to avoid taking an unsupported position: when requested financial figures are unavailable, it drafts a commitment to return with numbers rather than inventing them. Drafts land in the email Drafts folder, where Simonnet edits and sends them.
He reports that working from these drafts reduced his personal reply-preparation time from about ninety minutes to twenty, but the article supplies no independent timing study. The agent does not send anything itself. Simonnet's broader setup includes scheduled remote agents and local data pipelines, with retained context helping him enter a topic without starting from a blank page. The concrete result here is a filtered, bilingual morning reply queue. The useful copyable rule is to define what counts as waiting on the owner and distinguish preparation from sending, so the agent's work does not silently become an unreviewed business commitment.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Blog · Cyril Simonnet
Frequently asked questions
Which threads are excluded?
Copied-only messages, already answered threads, informational mail and issues owned by a teammate do not become drafts.
What happens when financial figures are missing?
The draft offers to return with figures rather than fabricating them; the human still reviews the wording and commitment.
Who sends the messages?
Simonnet explicitly says he edits and sends them himself. The agent only populates the Drafts folder.
More details
Remaining steps
- Avoid inventing missing figures or unsupported commitments.
- Save drafts for Simonnet to edit and send.
About this account
A scheduled Claude Code agent selects actionable threads, uses context and style instructions, and writes draft replies before the owner is awake.
Firsthand account by Cyril Simonnet; original publication date verified against Substack byline May 10, 2026. Outcomes are the author’s reports, not independently reproduced. Suggested prompt written by Agents In Use; not the author’s original prompt.
Credit: Cyril Simonnet
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Anthropic Routines · Email drafts
Inbox triage · Bilingual replies · Send approval
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.