Don’t get bot hurt. Get bot even.

Last week we talked about restraint and the idea that not every task needs a bot hovering over your shoulder.

Sometimes the fastest move is still the human one: write the note, sketch the idea, send the draft.

But restraint isn’t the end of the story. It’s the starting line.

When you do bring AI into the ring, the goal isn’t to hand it the job. The goal is to make the work sharper.

Think sparring partner, not substitute.

Pressure-test the idea

Good fighters don’t train by shadowboxing alone. They need someone pushing back, someone willing to test the stance and expose weak spots.

AI can do this surprisingly well.

Paste a draft of your argument and tell the bot to poke holes in it. Ask it to take the opposite side. Ask where the logic gets thin or the evidence starts to wobble.

Now the tool isn’t writing for you. It’s pressure-testing the structure.

The same move works in everyday work. Before a meeting, ask the bot what objections the room might raise. Before sending a proposal, ask what assumptions you’re making without noticing.

Used that way, AI feels less like hiring a ghostwriter and more like having a skeptical colleague on call.

Sharpen the question

Another useful move: flip the relationship.

Instead of asking the bot for answers, ask it for better questions.

Say you’re researching a new topic. Most people start with: “Write me a summary.” A better opening move is: “What questions should someone ask about this before making a decision?”

You’ll usually get five or six angles you hadn’t considered — the kinds of questions that widen your thinking before they speed it up.

That’s where the real value shows up. Not speed — perspective.

Final Bot Thought

The smartest people using AI right now aren’t outsourcing their thinking They’re stress-testing it. The bot throws the punches. You learn to read them.

Bot Talk: The bot isn’t stealing the job.
It’s hovering around the to-do list.

Everyone keeps asking the same question: Which jobs will AI replace?

This week, Anthropic released research examining how its Claude chatbot is actually used at work.

By combining real usage data with occupational task data, researchers found the highest exposure in roles such as computer programming, customer service and data entry.

In other words: the bot is hovering around tasks, not marching in to replace the whole job.

^{Source: Anthropic}

But the gap between what AI could do and what people are actually using it for remains large.

According to the report, AI coverage across many occupations remains a fraction of its theoretical capability.

So far, researchers also find no meaningful increase in unemployment among workers in highly exposed occupations since ChatGPT’s release.

There are early hints, however, that hiring may be slowing in some exposed roles, particularly for younger workers.

Longtime Bot Hurt readers may remember an early issue titled Not all bots want your job or your soul. 

The new research suggests that idea still holds — for now.

🚀 Coming up next week …

Even bots need to step away from the ring sometimes.

Bot Hurt will be on spring break for two weeks.

When we’re back, we’ll keep exploring the human side of working with AI — where the tools sharpen your thinking, where they quietly dull it, and how to tell the difference.

See you soon!