How to write emails faster (without sounding like a robot)
You spend 2.6 hours a day on email. Templates, snippets, and inline prediction cut the boilerplate — the parts of every email that were never the hard part.
The average professional spends about 2.6 hours a day on email — 28% of the workweek, per McKinsey. Reading it is half the problem. Writing it is the other half, and most of what you're writing isn't the hard part. It's "Hope you're doing well," and "Let me know if you have questions," and "Best, Jeff."
To write emails faster, stop treating every email as new: reuse your best sentences with templates and snippets, and let inline prediction finish the boilerplate while you keep every judgment call. The hard part of an email was never "Best regards" — it was knowing what to ask. Automate the ritual, keep the thinking.
I have an unusual vantage point on this: my team fine-tuned a small language model specifically to learn how people write email. What it picked up first surprised us, and it explains exactly where the time in email actually goes.
What a model learns from email (and what that tells you)
When we fine-tuned a small model on everyday email, the first thing it got good at wasn't content. It was register — the little ritual phrases that make an email an email. "Thanks, Kim." "Best, Jeff." "Please let me know if you have any questions." "Happy to jump on a call."
That's what the model learned because that's what email is. The genuinely novel content — the ask, the decision, the explanation — is a minority of the words. The majority is connective tissue: greetings, context-setting, transitions, sign-offs. The same tissue, over and over, in every message you send.
Do the math on your own inbox. At 40 words per minute, a 150-word reply costs you about four minutes of typing. If you write 20 replies a day, that's 80 minutes of pure keystrokes — and most of those keystrokes are phrases you've typed hundreds of times before. The average office worker receives 121 business emails a day (Radicati Group), and the replies follow the same patterns.
So the honest strategy for faster email isn't "type faster" and it isn't "have AI write it." It's: stop re-typing the ritual, and stop outsourcing the thinking.
The three speeds of email writing
1. Snippets for exact-repeat text. Your address, your availability line, your "no problem, I'll get to it today." If you type it identically more than a few times a week, it belongs in a text expander. Set it up once, pay one keystroke forever. Free options exist for Mac (Espanso is the open-source standard); a dedicated tool is a one-time setup and nothing per use after that.
2. Templates for structure. The recurring email shapes — the follow-up, the intro, the decline, the scheduling dance. Keep five skeletons in a notes file and paste the right one. You still write the unique 30%; the skeleton saves you from re-deciding the structure every time.
3. Prediction for everything variable. This is the one most people haven't tried, and it's where the fine-tune lesson pays off. Inline autocomplete — ghost text that offers the next word or the rest of the line as you type — turns out to be exceptionally good at exactly the parts of email the model learned first: the greeting, the transition, the sign-off, the "does Thursday work for you?" TypeTab does this system-wide on a Mac, learning your phrasing on-device, so the suggestions are your own greatest hits rather than generic internet prose.
The three aren't competitors. The pros I know run all three: snippets for the fixed text, templates for the rare shapes, prediction for the daily river of half-repeated sentences.
What not to do: the chatbot rewrite trap
The tempting move is pasting the whole thing into a chatbot — "write a polite follow-up to this client." It works, once. Then you notice you're spending as long editing the output back into your voice as you'd have spent writing it. And the output has the averaged, slightly-anonymous quality that readers feel even when they can't name it. I wrote about that mechanism in how to write faster with AI without sounding like AI — when the model writes the whole sentence, you get the average of a million emails, and the average doesn't sound like you.
There's a real cost beyond time: the chatbot workflow trains you to stop noticing your own phrasing. The prediction workflow does the opposite — you're still in the driver's seat, every sentence starts with your words, and the tool only ever finishes what you began. Your voice compounds instead of eroding.
(A side note for anyone whose flagged-as-AI anxiety just spiked: finishing your own sentences is precisely the workflow that doesn't produce detector-bait, because the sentences are still yours. More on that in why AI detectors keep flagging human writing.)
The math, if you want the receipt
Say you spend 60 minutes a day typing email. Cutting a third of the keystrokes — conservative, if prediction handles just the ritual phrases — saves 20 minutes a day. That's roughly 80 hours a year. Nobody gets that from a typing course.
And unlike a typing course, this compounds. Snippets get set up once. Templates get refined once a quarter. Prediction gets better the longer you use it, because the model keeps learning your phrasing. The five minutes you save on Monday's email is the same five minutes; the model that learned your sign-off by Friday keeps saving you that time every day after.
FAQ
How do I write emails faster without sounding robotic? Keep authorship of the sentences. Use snippets and templates for text you repeat identically, and use inline prediction (ghost text you accept with Tab) for everything else — but write the unique content yourself. Avoid pasting whole emails into chatbots; editing their output back into your voice usually costs more time than writing it did.
How much time do people spend on email? About 2.6 hours a day, or 28% of the workweek, according to McKinsey Global Institute. The average office worker also receives around 121 business emails per day (Radicati Group). Most reply time goes to boilerplate — greetings, transitions, sign-offs — which is exactly what automation handles best.
Do email templates come across as lazy? Not if the unique content is genuinely written. Readers notice sincerity in the ask and the specifics, not in "Hope you're doing well." Templates handle structure so you can spend your attention on the parts the recipient actually reads.
Does inline prediction work in Gmail and Slack on a Mac? Yes — system-wide tools like TypeTab work in any text field, including browser-based Gmail, Slack, and Notion, because they watch the focused field via macOS Accessibility rather than waiting for apps to adopt an API. Apple's built-in predictive text, by contrast, only works in apps updated for Apple's text APIs.