When AI speeds you up, and when it slows you down
The tasks where it saves an afternoon, and the ones where it costs you one.
There's a quiet assumption baked into most AI advice: that using AI is always faster. Sometimes it saves you an hour, but other times it can cost you one. Knowing the difference is the actual skill. This is a short, honest map of where AI earns its keep and where it doesn't.
Where AI genuinely speeds you up
The pattern is the same every time: the work already exists in some form and needs transforming, not inventing. A blank page becomes draft zero you can edit. Ten product descriptions become variations on the one you've already approved. Rough notes become a client email in the register you asked for. A long document becomes the two paragraphs you actually needed to read.
It is also fast when you never intend to use the output at all. Twenty headline options, ten ways to frame a pitch, five angles on a problem. You aren't shipping any of it. You're using it to unstick yourself, and an option you reject still moves you forward.
Speeds you up
Blank-page work: first drafts, outlines, structuring messy notes
Repetitive work with a clear pattern, once you've done the first one
Translation and tone-shifting
Research scaffolding, not the research itself
Brainstorming when you're stuck and need volume
Slows you down
Anything needing judgement you can't verify quickly
Short tasks where the round trip costs more than just typing it
Output that has to be exactly right in a domain you can't check
Creative work where your voice is the point
Work you don't actually understand yet
AI speeds up the start of things and slows down the end of them.
Where AI quietly slows you down
The hidden cost is always verification. If you can't tell at a glance whether the output is right, you have to check it, and checking a confident paragraph you didn't write takes longer than writing it would have. That's why AI is at its worst in areas you don't know well. It hasn't sped you up; it has handed you something that looks finished.
Two smaller versions of the same problem. Short tasks, where the round trip costs more than typing the thing. And work where your voice is the point: AI can draft it, but if the finished piece has to sound unmistakably like you, you'll spend longer rewriting its version than writing your own.
A simple test before you reach for it
One question, before you open the tool. Can I judge this output quickly, or will I have to verify every claim? If you can judge it at a glance, AI will probably save you time. If you can't, you are better off starting it yourself.
Three things to do this week
01
Ask whether you can judge it at a glance
If you'd have to fact-check every sentence, you'd have been faster writing it yourself.
02
Skip it when the round trip costs more than the task
A two-sentence Slack message doesn't need a prompt. By the time you've written one, you could have typed it.
03
Keep it away from anything that has to be exactly right
Legal language, financial figures, final client facts. Draft with it, don't finish with it.
Conclusion
01
AI is a force multiplier for work you already know how to do, and a liability for work you don't.
02
It's brilliant for drafts and bad for finals.
03
The teams getting the most out of it aren't using it for everything. They know which mode they're in.
Go to next