4 min read

Make Claude write emails that actually sound like you

AI emails that don't sound like you are a settings problem, not a Claude problem. How to capture your writing style so every draft sounds like you wrote it.

AI-generated emails have a tell. Too formal, too polished, a bit of a lecture — and anyone who actually knows you can spot it in the first line. That is the number one reason people try Claude for email, feel embarrassed by the output, and quietly go back to typing everything themselves.

Here is the thing: that is not Claude being bad at writing. It is Claude having no idea how you write. Fix that once and every draft afterwards gets dramatically better. There are two parts — teaching it your voice, and a few drafting habits most people miss.

The writing-style setting in action.

Why the drafts sound generic

Left to its own devices, a model writes for the average reader in the average register. It does not know that you open with “Hi” not “Dear,” that you keep things short, that you sign off with your first initial. So it defaults to a corporate voice that belongs to no one. The answer is not to re-explain your style in every prompt. It is to set it once.

Teach it your voice first

Inside Claude for Outlook there are a couple of settings most people never open, and they are the whole game:

  • Capture your writing style. Point Claude at your sent emails and let it learn your actual patterns — tone, structure, greetings, sign-offs, how long you tend to run. From then on that becomes the baseline for every draft, instead of the internet average.
  • Write custom instructions. In plain words, tell it who you are and how you communicate: “I’m direct, I keep emails to a few lines, I don’t use exclamation marks, I always propose a next step.” This is where you encode the things your sent folder does not obviously show.

Treat it as iterative, not one-and-done. The first capture gets you most of the way; when a draft still is not quite right, tweak the instructions rather than rewriting the email. Over a week or two it converges on genuinely sounding like you.

The rule underneath all of it: don’t just let Claude draft — teach Claude how you write first. Five minutes of setup beats a hundred re-edits.

The drafting habits that make it smoother

Once your voice is set, a few small habits separate a smooth workflow from a fiddly one:

  • Give it the context before it writes. Point it at the actual thread, not a summary. It is far better at responding to real content than inventing generic content.
  • Keep the thread intact. There is usually a single option that lets Claude work on the live reply rather than making you copy-paste in and out — use it. It sounds trivial; over a day of email it is the difference between “helpful” and “not worth it.”
  • Start from your rough notes. Dictate or bullet what you want to say and let Claude shape it. You supply the intent and the facts; it handles the phrasing. That keeps the substance yours and the polish fast.
  • Write straight into the compose window so the draft lands where you actually send from.
  • Iterate out loud. “Warmer.” “Cut the last line.” “You invented that date — remove it.” Correcting a draft twice is faster than engineering the perfect prompt once.

The part you keep

None of this removes your judgement, and it should not. Claude drafts; you own the send. That is especially true for anything with a number, a date, or a commitment in it — read before you hit send, every time. Get the voice right and the habits in, though, and email stops being the thing that eats your morning.

This is one skill in a bigger set — see everything Claude can do in Outlook for the full picture, and the hidden skills most people never find once your drafts are sorted. And if you want your team writing in their own voice with AI rather than sending robotic replies, that is the kind of thing we build in our workshops.

Want this done inside your team, not just read about?

The AI Execution Sprint runs six weeks with the people doing the work. No pitch decks — just a conversation about whether it's worth doing at all.