Nobody Decided You'd Stop Writing Your Own Emails

The biggest decisions about AI in your working life are the ones nobody made, and your emails are a good example.

There was no meeting, and nobody sat you down and said AI writes your emails from now on. One day there was a button offering a draft. You pressed it because you were busy, the draft was fine, and the next day you pressed it again. Six months on, you can't remember the last work email you wrote from a blank screen.

You never decided that. Whoever put the button there decided it, and so did the plain fact that pressing the button was a shade easier than not.

Ethan Mollick, who writes One Useful Thing, has a name for this. He calls it cognitive surrender, which means quietly handing over your thinking. A tool that takes no effort invites you to stop doing a thing, without you ever choosing to stop.

Every AI feature arrives with a default, the setting it starts on, like "suggest a reply" or "summarise this thread". A product team sets that default, and their job is to get the feature used. The surest way to get anything used is to make it the easiest path.

There's nothing sinister in that, but it does mean someone else is deciding how much thinking you keep.

You could fairly say that every tool sets defaults, and nobody checks the spellchecker or the calculator. And life would be unbearable if every convenience needed a conscious decision.

I agree, and most of these defaults are fine. The auto-summary is a mercy next to forty messages about a car park. What's different is how far these defaults reach.

A spellchecker decides whether a word is spelt right, while these features decide whether you still put your own thoughts into writing. That difference is worth one sitting.

Checking the defaults

Mollick's answer is to keep using the tools, and to decide on purpose what you keep. That means finding the defaults you've already accepted, which is harder than it sounds because the whole trouble is that they've gone invisible. Let AI walk you round them.

I want to list every place AI now does something for me by default, without me asking each time. Go through the tools a typical office worker uses, such as email, calendar, documents, chat, phone, search and browser. For each one, ask me whether any AI feature is switched on that acts automatically. One tool at a time. Don't tell me whether to keep them. Just help me see the list.

Then, with the list in front of you, work through these steps.

1. Strike out anything you ask for, and what's left is what happens without asking. For example, the suggested reply, the auto-summary, the sentence that finishes itself, and the meeting notes that turn up in your inbox.
2. Next to each one, write whether you chose it. Saying "it was there and I started using it" isn't choosing, and most of the list will say exactly that. Nobody is marking this. You're only finding out which decisions were yours.
3. Take one back, and keep one on purpose. Taking one back means switching it off, or doing that job by hand for a month to see whether you miss the feature. Keeping one on purpose means saying why, out loud. For example, "the auto-summary of long threads saves me an hour a week and I don't miss reading them." Once you've said it aloud, it has become a choice.
4. Write both down, one line each. "I write my own replies to people I know. I let it summarise threads I'm cc'd on." That's your policy, and nobody else wrote it.

This takes twenty minutes, and you only do it once. Then take a quick look every few months, because new features will keep arriving and each new default brings the same question back. Would I have chosen this?

When the check is done, take the same question to the rest of your life. Make a list of what you refuse to automate, meaning the things you won't hand over to a machine.

Credit to Ethan Mollick, whose phrase "cognitive surrender" and idea of choosing what you keep both come from "Choosing to Stay Human" on One Useful Thing (May 2026). Read it for the full argument.