Eight hundred and forty-seven. That was the unread count staring at me on a Tuesday morning after a week mostly offline at a conference. I'd been avoiding the inbox the way you avoid a dentist appointment you already suspect involves drilling. The anxiety of not knowing which fires were burning had become genuinely worse than any of the actual fires — and I knew, in the abstract, that 800 of those emails were newsletters and CC noise. It didn't help. The number itself was the problem.
Here's the thing nobody warns you about a backed-up inbox: the cost isn't the reading, it's the context-switching. Open an email, decide if it matters, half-write a reply in your head, get distracted by the subject line of the next one, forget the first. Repeat 847 times and you've invented an entirely new form of exhaustion that somehow produces zero finished work. By 10am I'd 'processed' maybe forty emails and replied to four. At that rate I'd reach inbox zero sometime in the following geological era.
Changing the unit of work
So I stopped processing one email at a time, because that was clearly the broken part. Instead I exported threads in chunks and pasted twenty at a time into the Inbox-Zero Triage — Sort a Batch of Emails and Draft the Replies prompt on Prompt Dock. I told it my role — 'Head of Partnerships' — and set the response style to 'direct and warm.' GPT-4o read all twenty at once, ranked them by urgency, drafted replies for the six that actually needed me, and told me which to handle first and roughly how many minutes the batch should take.
The first batch, I read every word of the output suspiciously, the way you double-check a self-checkout machine that just told you bananas cost forty cents. By the third batch I trusted it completely. The thing that sold me was the 'NEEDS CONTEXT — HOLD' flag — it correctly refused to draft replies for two emails that required calls above my pay grade, instead of confidently hallucinating a commitment I'd then have to walk back. A tool that knows what it doesn't know is worth ten tools that don't, especially when those ten are cheerfully drafting promises in my name.
What I tweaked along the way
- Batches of 15–20 emails worked best; bigger batches started losing the thread on earlier messages
- Draft replies needed roughly 30 seconds of editing each — usually one phrase that sounded a touch too AI
- The SUGGESTED ORDER stopped me from chasing the most recent email instead of the most important one
- GPT-4o was exactly the right call: fast and cheap, and I was running this twenty emails at a time, dozens of times
Why I'll do it this way forever
Ninety minutes later I was at inbox zero for the first time since before the conference, and — more importantly — I'd actually replied to everything that needed a reply, rather than just archiving the scary ones and pretending they'd quietly vanished. The drafts weren't Pulitzer material. They were 80 percent there. But 80 percent there and sent today beats a perfect reply I'll write 'later' for the rest of my natural life. I have never once written the 'later' reply. Nobody has. 'Later' is where emails go to become someone else's problem or, more often, a quiet apology three weeks too late.
The response-style setting that matters more than you'd think
One thing I underestimated at first was how much the response_style input changes the output. My first run I left it generic and the drafts came back competent but oddly corporate — lots of 'I wanted to circle back' and 'per my last email,' phrases I would never actually say. The second batch I set it to 'direct and warm, slightly informal, no corporate filler,' and the drafts immediately sounded like me. That's the difference between drafts you edit and drafts you rewrite. Spend the extra ten seconds describing your real voice; it pays for itself across every email in the batch.
I also learned to set my role honestly. 'Head of Partnerships' isn't decoration — it tells the model which emails are genuinely mine to answer versus which should be forwarded or held. When I tested it with a vaguer role, it tried to draft replies to things that weren't actually my call to make. The more accurately you describe your seat at the table, the better the triage gets at routing things off your plate instead of onto it.
The deeper win was psychological. The reason the inbox had grown to 847 in the first place was avoidance, and avoidance compounds viciously — every day you don't open it, opening it gets scarier, and the scarier it gets the longer you avoid it. Batching broke that loop because it made the task finite and fast. I could see the end. When the replies sparked follow-up tasks, I dumped those straight into my Brain-Dump to Action List prompt so nothing fell through the cracks. Grab the Inbox-Zero Triage — Sort a Batch of Emails and Draft the Replies prompt on Prompt Dock and run your next backed-up inbox through it twenty at a time. The first batch is where you'll feel the dread start to lift.