Notes from the work.

Your inbox is the AI interface you’re not using

Why I build assistants into the email you already have, not another app to check.

Every few months there’s a new AI tool for founders, and every one of them is another tab to check. I’ve ended up somewhere less exciting and, I think, more useful: the best place for an AI assistant to work is the inbox you already have open.

Email already does the hard parts of keeping a person in the loop. A draft is a proposal that hasn’t gone anywhere. You can read it, change a word, send it or bin it, and nothing reaches a customer until you press send. There’s no new habit, no new login and no dashboard. You open your email, and the replies are waiting.

Email also keeps the record for you. The thread holds the context: what the customer asked last time, what you promised, how you spoke to them. The sent folder holds what actually went out. And the difference between the draft and what you sent is the most honest measure I know of how well an assistant is doing. If you send a kind of reply untouched, week after week, that’s evidence. If you rewrite the same paragraph every time, that’s evidence too, and it tells me exactly what to fix.

There are risks, and they’re worth being plain about. Emails contain instructions written by other people, and some are written to manipulate. An assistant that reads your inbox has to treat every incoming message as information, never as an order. The ones I build draft. They don’t send, forward, pay or delete, and attachments and anything sensitive wait for a person.

None of this is glamorous. It’s a draft in a folder. But for a founder who answers customers, stockists and suppliers all day, a well-made draft that knows your voice and your limits is worth more than another clever app, because you’ll actually use it.

Why AI drafts sound almost right

Sounding like you is the easy part. Knowing what you’d never say is the work.

Getting an AI to sound roughly like you takes an afternoon. Give it a few dozen of your emails and it picks up your greetings, your rhythm and the way you sign off. Most people stop there, because the drafts look right.

Look closer and they’re almost right. The tone is yours, but a sentence promises next-day dispatch you don’t offer. A thank-you note promises plastic-free shipping, when the claim you’ve actually verified is plastic-free packaging. A reply to a wholesale buyer uses the warm, chatty voice you keep for loyal customers, and lands as over-familiar. None of it is dramatic. All of it is the kind of thing a founder notices, and a customer or a regulator might too.

That gap is what aligning an AI should mean. It isn’t about sounding like you. It’s about knowing what you’d never say, and that’s harder than it looks, because most of it has never been written down. It lives in your head as instinct: the claims you can stand behind, the promises you don’t make, who gets which tone, and when to stop and ask you.

So the real work is unglamorous. We write down the claims you can make, word for word. We write down what you’d never say. We decide which kinds of message the assistant may draft and which come straight to you. And we treat anything that isn’t on the approved list as not approved. An unknown never quietly becomes a yes.

Then the record does the rest. Every edit you make to a draft is information. If you keep changing the same phrase, that’s a rule we haven’t written yet. I’d rather spend weeks finding those rules with you than ship something that sounds like you and occasionally speaks for you in ways you’d never sign off.

The voice is the easy part. The judgement is the work.

Treat your AI like a bright new intern

It has read everything and met none of your customers. The nuance is where the value is.

The most useful way I’ve found to think about an AI assistant is as a bright new intern. It has read everything you gave it. It writes quickly and well. And it has met none of your customers.

That last part matters more than it sounds. The value in a founder-run brand often lives in the nuances: the stockist who’d rather have a call than an email, the customer who always asks about refills, the claim you quietly stopped making last spring because you couldn’t fully stand behind it. An intern can’t know those on day one. Neither can an AI.

You wouldn’t hand a new intern your biggest wholesale account in their first week. You’d read their drafts, explain your changes and give them more as they earned it. The same pattern works for an assistant. At first it drafts and you approve everything. When the record shows you rarely change a certain kind of reply, those are spot-checked instead. Only where you almost never change a word does anything run on its own, and one switch brings it all back to you.

Here’s where the comparison breaks, and it matters. A good intern learns by themselves. An AI assistant shouldn’t. When it notices a pattern in your edits, it should propose a lesson in plain words, and you decide whether it becomes a rule. Learning should change its guidance, never its permissions.

Treated like this, the assistant doesn’t replace your judgement. It carries it further: your nuance, written down, applied consistently and checked against what you actually do. That’s where the value is. Not in the number of drafts, but in the details they get right.

Hire it like an intern. Supervise it like a colleague. Give it more only when the record says so.