Why I switched, and what I stopped missing
Four mailboxes, three ecosystems, and a default mail app that had quietly stopped being adequate. Notes from moving, including the parts that were annoying.
3 min read

I did not switch mail clients because I was excited about mail clients. I switched because I had reached four mailboxes across three ecosystems and the arrangement I had was quietly costing me an hour a week.
The arrangement
Work on Microsoft 365. Personal on Gmail, because it has been there since university and moving it is a project I keep not starting. A domain I own, on a small IMAP host, which is where anything I actually care about goes. And a shared mailbox for a side project on Google Workspace.
Four mailboxes. The built-in client on my phone handled three of them acceptably and the fourth badly. So I had a second app for that one. Then I had a third app on the desktop because the desktop client from the same vendor was worse than the mobile one.
Nobody decides to run three mail clients. You arrive there one reasonable decision at a time.
What was actually wrong
Not features. Every one of those clients could send and receive mail.
The problem was context switching, and specifically the small tax on every single message: which app is this in, which account am I replying from, have I already dealt with this on the other device.
The reply-from-the-wrong-address mistake is the one that finally did it. I sent a client a proposal from my personal Gmail. Nothing catastrophic. Just unmistakably unprofessional in a way I could not un-send.
What changed
The unified inbox is the obvious answer and it is only half the reason it worked.
The half nobody mentions is per-account notification rules. Work mail can interrupt me. The side project cannot. Newsletters do not even get a badge. Before, my choice was notifications on for everything or off for everything, and I had chosen off, which meant I was checking manually all day, which is worse than either.
Being interrupted only by things that warrant it turned out to be the entire difference.
Clusters took a week
Grouping by sender rather than folder felt wrong for about five days. I kept looking for folders.
Then I stopped, because the thing I was looking for was already at the top. Six months of notifications from one service collapse into one row. The thread with a person is one row containing that whole relationship rather than forty rows scattered across a year.
I would not have believed this from a feature list. It needed a week.
The AI part, honestly
I was ready to be annoyed by this. Most AI in most products in 2026 is a feature nobody asked for, bolted to a button you cannot remove.
What I use, in practice, is summarizing a long thread I have been added to late. Twelve messages of a decision I was not part of, and I need the shape of it before replying. That is a genuinely good use, because it is a task I would otherwise do badly by skimming.
What I do not use is having it write my replies. Not on principle. It just does not sound like me, and the editing costs more than the writing saved.
Take from that what you will. The summarizing is worth having. The rest I leave alone.
The annoying parts
In fairness:
Migration is not zero effort. Mail comes across on its own, because it lives on the server. Signatures do not. Neither do the local rules I had accumulated. That was an evening.
I had to learn new swipes. Two days of archiving things I meant to snooze.
One account needed reconfiguring. I had originally set my work account up as IMAP years ago, which is why I had never had a working calendar on that phone. That was my error, inherited, but it surfaced during the move.
Would I go back
No, and the reason is smaller than the pitch would suggest.
It is not that the software is remarkable. It is that I now have one place where all of my correspondence is, it tells me about the things that matter and stays quiet about the rest, and it is the same on the phone, the laptop and the machine at the office.
That should be an unremarkable baseline. It took four mailboxes and one badly addressed proposal to work out that it was not.
