The average person types around 40 words per minute. The average person speaks around 130–150. That’s a 3–4× gap on the single activity most knowledge workers spend the most time on: getting words out of their head.
So why does everyone still type?
Because raw transcription was never the bottleneck — cleanup was. Speech is messy: we hedge, we restart sentences, we say “um” roughly every ten words. Old dictation tools typed all of that, and fixing it cost more time than typing would have. The speed advantage evaporated in the edit.
That’s the specific problem AI cleanup solves. When the tool removes filler and false starts, fixes punctuation, and leaves your meaning alone, the 3–4× gap stops being theoretical. You speak the Slack message; the Slack message that appears is the one you’d have typed — a minute earlier.
Where dictation actually fits
Honest answer: not everywhere. Dictation is at its best for prose you’d say out loud anyway — email replies, chat messages, doc drafts, commit messages, meeting follow-ups. It’s weaker for dense editing, tables, and open-plan offices. Most people land on a mix: speak the first draft, type the surgery.
Even that mix is a big number. If you write two hours of prose a day at 40 wpm, speaking the first drafts gives you most of those two hours back — every day.