Reading Time: 5 minutes

Hey AI Community,

I want to tell you about a conversation I had with someone who was very proud of how much time AI was saving them.

They told me they used to spend two hours writing their weekly report. Now they spend thirty minutes. That's ninety minutes saved every week. They were genuinely excited about it.

Then I asked them to walk me through what those thirty minutes looked like.

They opened Claude.

Typed a rough prompt.

Got something back that was too formal. Reprompted.

Got something back that missed the key points. Added more context.

Got something closer but the tone was wrong. Tweaked the prompt.

Got a better draft. Spent fifteen minutes editing it into something that actually sounded like them.

I asked how long the prompting and editing took.

They thought about it. "Maybe forty-five minutes."

So the report that used to take two hours now takes an hour and fifteen minutes. Which is genuinely better. But it is not the ninety minutes saved they thought it was.

And the thirty minutes they quoted me was only the part where Claude was doing something. They had stopped counting the rest.

That is the productivity illusion. And almost everyone using AI has it.

Before I get into why this happens and what to do about it, here’s something worth your attention today.

Last month, 48% of visitors to documentation sites across Mintlify were AI agents, not humans.

Claude Code, Cursor, and other coding agents are becoming the actual customers reading your docs. And they read everything.

This changes what good documentation means. Humans skim and forgive gaps. Agents methodically check every endpoint, read every guide, and compare you against alternatives with zero fatigue.

Your docs aren't just helping users anymore. They're your product's first interview with the machines deciding whether to recommend you.

That means: clear schema markup so agents can parse your content, real benchmarks instead of marketing fluff, open endpoints agents can actually test, and honest comparisons that emphasize strengths without hype.

Mintlify powers documentation for over 20,000 companies, reaching 100M+ people every year. We just raised a $45M Series B led by @a16z and @SalesforceVC to build the knowledge layer for the agent era.

Make Your Docs Agent-Ready

Here is what happens when most people measure how much time AI saves them.

They measure the output time. The thirty minutes Claude spent generating the draft. The ten minutes the image took to produce. The five minutes the code took to write.

They do not measure the input time. The fifteen minutes figuring out why the first prompt gave them something unusable.

The ten minutes rewriting the brief.

The twenty minutes editing the output back into something that actually works. The five minutes they spent staring at a response wondering if it was right before deciding to just check it themselves anyway.

Output time is visible. Input time feels like thinking. And we don't count thinking as work even when it is the most expensive part of the process.

After watching a lot of people use these tools in real workflows, I think there are three categories.

The first group genuinely saves time.

These are people who knew exactly what they wanted before they opened Claude. They had a clear brief, a specific format in mind, and enough experience with the tool to get there in one or two prompts. For them the productivity gains are real because they removed the generation time from a process where the thinking time was already done.

The second group breaks even. They save time on generation but spend it on prompting and editing. The total time is roughly the same as doing it manually. The output might be slightly better or slightly worse depending on the day. These people feel productive because they associate the tool with efficiency even when the numbers don't support it.

The third group loses time. These are the people spending forty-five minutes getting AI to write something they could have written in twenty. They are prompting a tool to do something they could do faster themselves, adding friction to a process that didn't need it, and then counting the AI's output time as the only time that mattered.

Most people think they're in the first group. A lot of them are in the second or third.

The productivity illusion has a specific cause.

When AI generates something it feels like you did less work. Even when you spent significant time getting it there, the experience of watching text appear on a screen feels passive in a way that typing the same text yourself doesn't. So you discount the time you spent prompting and editing because it didn't feel like the same kind of effort.

This is the same reason people think outsourcing saves more time than it does. The time you spend briefing, managing, and reviewing someone else's work feels different from the time you spend doing the work yourself. So you don't count it the same way.

AI has the same accounting problem. The prompting and editing are work. They just don't feel like it.

The people genuinely saving significant time with AI share one specific thing.

They front-load the thinking.

They know exactly what they want before they open the tool. They have a clear format in mind. They have examples of what good looks like. They have already made the decisions the AI would otherwise force them to make through trial and error.

When you do the thinking before you open Claude, the session becomes execution rather than exploration. Execution is fast. Exploration is expensive and most people are counting it wrong.

The other thing that consistently produces real gains is skill files and saved context. When the AI already knows your voice, your audience, your format preferences, and your standards before you type a word — the prompting time collapses.

This is exactly why the Skills Pack works for the people it works for. Not because Claude is smarter but because the setup time moved outside the session.

Get The Claude Skills Pack

Here is how to find out which group you're actually in.

Next time you use AI for a task, start a timer when you open the tool and stop it when the output is ready to use. Not when Claude finishes generating. When it is actually ready to send, publish, or hand over.

Compare that number to how long the same task would take you to do manually at a reasonable pace.

Do this for a week. The results will probably surprise you.

If you're genuinely saving time — great. Keep doing what you're doing.

If you're breaking even or losing time — the problem is almost never the tool. It's the absence of a clear brief before you start. Fix that first and the productivity gains become real.

Reply and tell me honestly — have you actually measured the time? Or have you been counting the output and ignoring the rest?

I want to know what this community finds when they actually look at the numbers.

— Prompt Guy

This newsletter goes out to over 95,000 people every week. If you have a product or AI tool you want introduced to this audience, reply with "sponsorship" and I'll send you the details.