A conversation with the Action Plan team
How do I automate my email with AI — safely?
I spend two hours a day in my inbox and most of it doesn't need me. But I can't have an AI sending things in my name unsupervised — one bad reply to the wrong customer costs more than all the time saved.
Safe email automation has one rule: the AI does the reading, you keep the send button — until it has earned otherwise. Triage everything, draft replies for approval, surface only what needs you, and escalate by explicit rules. Expect a training week; then your inbox becomes a ten-minute digest.
The two hours aren't really "answering email" — they're deciding, over and over: does this need me, now, and what's the move? That decision fatigue is the actual cost. Most email tools fail you in one of two directions: dumb filters that file things and still leave every decision to you, or over-eager AI that acts in your name without judgment. The middle path is an assistant with judgment and boundaries.
How I'd structure it, in any tool:
- Triage into decisions, not folders. Four buckets: needs you now · reply drafted, approve it · scheduled/filed with a reason · noise.
- Draft, don't send. Every reply prepared in your voice, waiting for your click.
- Escalate by rule. Named senders, money topics, upset customers, anything uncertain → straight to you.
- Digest, don't drip. One morning briefing beats forty notifications.
- Earn autonomy per category. After weeks of perfect meeting-confirmation drafts, maybe those auto-send. Your call, one narrow category at a time.
Here's what that looks like as an Action Plan morning digest — Valentina runs it, the rules are yours, and every decision she made overnight is on the record:
- YOU 3 need you: contract question, upset customer (escalation rule), intro request
- APPR 9 drafts ready in your voice — approve, edit, or reject each
- FILED 19 handled: scheduled, filed, or archived — each with a one-line reason
Setting expectations, because I'd rather under-promise: week one, you'll correct my triage and edit my drafts — that's me learning your categories and your voice, and it's effort. By week two the drafts start sounding like you. The send button stays yours for as long as you want it. Some of my best clients never give it up, and that's a perfectly good end state.
Safe email automation, step by step
- Triage into decisions: needs you · draft to approve · filed with reason · noise.
- Draft, don't send — replies prepared in your voice, awaiting your click.
- Escalate by explicit rule — named senders, money, sentiment, uncertainty.
- One daily digest instead of constant notifications.
- Autonomy is earned per category, with your explicit opt-in — never the default.
Will the AI send email in my name?
Not by default, and never without your opt-in. Draft-don't-send is the safe baseline; auto-send is something a category earns after consistent perfect drafts.
How long until it's useful?
A training week of corrections, then most inboxes compress to a short digest plus drafts to approve.
Action Plan is in private preview. When your invite opens, onboarding starts from exactly this conversation.
Got it — you're on the early-access list, and I've saved your goal: . When your invite opens, this is the first thing we'll work on together. Watch your inbox.