The short answer: inbox zero treats the symptom. The problem was never unread count... it's that you are the only person who can process your email. In 2026 that's a choice. An AI executive assistant can triage every message, label what needs you, draft replies for the routine stuff, and hand you a short briefing instead of a wall of unread. Mine does exactly that, every day.
Back in 2020 I wrote a post called "Forget Inbox Zero". The argument: chasing an empty inbox is a treadmill, because email arrives faster than you can ritually process it. The fix back then was filters, delegation and lowering your standards on purpose.
The argument still holds. The fix got a massive upgrade.
Why inbox zero never worked
Inbox zero is a scoreboard for a game you can't win. Every email you answer teaches the world you're the person who answers. The reward for processing your inbox faster is more inbox. You weren't drowning because you were disorganised. You were drowning because you were the only processor.
What does an AI executive assistant actually do with email?
Here's what mine does with my real inbox, every day:
- Triage. Every incoming thread gets read and sorted: needs Carl, EA can draft, waiting on someone else, done. I see a labelled inbox, not a pile.
- Drafting. Scheduling, logistics, acknowledgements, routine replies... drafted and sitting ready for my approval. Anything involving opinions, money, relationships or commitments stays with me. That boundary is deliberate and written down.
- The briefing. Instead of scrolling, I get a short daily rundown: what needs me, what's stuck, what's coming. One idea per line. I act on a summary, not a feed.
- Follow-up memory. The threads I owe people, surfaced before they get awkward. This alone is worth the setup.
Isn't it risky letting AI touch your email?
It is if you hand over sending. So don't. The pattern that works is draft-first, earn autonomy: the AI drafts everything and sends nothing until it has a track record. You review its drafts like you'd review a new hire's. Over weeks, categories it never gets wrong graduate to more autonomy... exactly the way you'd extend trust to a human EA. Nothing goes out silently, ever.
How do you set one up?
The tools change monthly, the sequence doesn't:
- Write the rules first. What can it draft? What must always come to you? What must it never do? Ten minutes with a document beats ten weeks of vibes.
- Give it context. An AI with no knowledge of your business writes like a polite stranger. Feed it your voice, your offers, your key relationships. (This is what I call the AI Business Brain, and it's the difference between generic and yours.)
- Connect it to your inbox through your AI platform's email connector.
- Run draft-only for two weeks. Correct it like you'd correct a new assistant. It compounds.
What changes when it works?
The best description I have is a moment from one of my own events: personalised pre-event emails went out to every attendee... while I was on the floor playing trains with my daughter. Nothing about that afternoon looked like productivity. The machine was working. I wasn't. That's the point of all of it.
Inbox zero asked you to become a faster processor. The AI EA asks a better question: why are you the processor at all?