By Dorsey Standish, MS

My wife jokes that Claude is my new girlfriend because I spend so much time building agentic systems for my business. I am a heavy AI user, and I am paying attention to what happens when getting an answer becomes easier than thinking through a question.

Cognitive patience is a useful way to describe the willingness to stay with a hard question long enough to do some thinking of your own. It might mean writing a first pass before asking AI, opening a source before trusting a summary, or taking time to learn a tool you don't understand yet.

There is research worth paying attention to here. In a 2025 study of 319 knowledge workers, greater confidence in generative AI was associated with less self-reported critical thinking. The study also found that critical thinking shifted toward checking information and deciding how to use the output. It was a survey about people's experience of their work, not evidence that AI had damaged their brains.

Here are three practices I use to stay involved in my own thinking.

1. Use AI on purpose

When I was growing up I loved going to visit my grandparents because I always got to eat Stouffer's mac and cheese. The only downside was that once they got into their 70s, they pretty much stopped being willing to learn how to use new technology. Even though it was the early 2000s, they still drove an Oldsmobile from the 80s and never learned how to use a microwave. The Stouffer's mac and cheese took an hour to cook in the toaster oven.

As an adult I can appreciate that it honestly probably tasted way better that way. But as a kid I can remember rolling my eyes and thinking, “Why can't they just learn how to use a microwave? You literally just hit a few buttons.”

As an adult, especially watching older people try to adapt to smartphones and now AI, I understand this a lot better. It is scary to try to learn something new, especially when you do not fully understand it. People are right that there is real value to doing things the old-fashioned way. Writing an essay by hand is akin to making Stouffer's mac and cheese in the toaster oven. It takes longer, but you might feel a greater sense of satisfaction in making and enjoying it.

AI can be a useful tool for creativity and exploration. Learning it takes trial and error, and some willingness to explore and not know.

Choose a task you understand well enough to judge. Decide what a useful result would look like before you ask. When the answer comes back, read it and decide what you would keep, change or question. Give yourself permission to prefer your own version.

If you learn how to use AI but still want to write everything by hand and manually open every folder, that is your prerogative.

2. Still Google things

I have Google Chrome as one of my top three buttons along with texting and phone calls on my iPhone. I try to use the phone for texting, calling, and looking stuff up online, staying in that investigative, engineer, figure-out-the-problem, problem-solver mode that scanning Google requires you to do.

Even when an AI summary comes up at the top, you can keep scrolling, scan other headlines and find other sources. The useful part is opening those sources and checking what they actually say.

One night we were watching Lucas the Spider with our son. We were wondering how many eyes spiders have, and we were thinking that spiders had eight eyes, but we could only see four eyes on Lucas. So Mari asked ChatGPT how many eyes Lucas has. ChatGPT said eight.

We looked again. We still only saw four. So we said, really? Where are they? We only see four. And ChatGPT gave very specific instructions that there were two big ones, two smaller ones, and four even smaller ones around these eyes.

This was simply not the case. There were no invisible eyes or smaller eyes whatsoever. When we looked it up, Lucas had four eyes.

What stayed with me was how competent ChatGPT could seem while giving us a wrong answer. The detail made it sound more convincing, even while we were looking at the contradiction on the screen.

I have had the same thing happen using Claude to do my weekly brain health research review. It gave me a summary with citations, and when I clicked the DOI link, it was a completely different topic. The study it described did not exist.

For a research claim, I now want to see the study itself. For a recommendation, I want to understand what it is based on. A search result can be wrong too, so the source and its evidence matter more than which tool found it.

3. Think first, then reach for AI

Before you ask AI, ask yourself.

Pick one thing this week you will work out for yourself before you touch AI. One memo. One strategic question. One conversation you do not prep with Claude. Try a three-minute timer with no tabs open, and write your own first pass.

Three minutes is a starting experiment. You can take longer if the question needs it. The point is to arrive at the tool with something of your own to compare against its answer.

You might find that AI notices a gap you missed. You might decide your original judgment was stronger. Either way, you have something to contribute to the exchange.

The next time you open a chat window, write down what you think first.

Dorsey Standish, MS leads Mastermind, a neuroscience-based brain health and performance company in Dallas. She holds a master's degree in cognitive neuroscience from UT Dallas and a mechanical engineering degree from Penn. Explore Mastermind's work with corporate teams.

Related reading: You Can't Outsource Your Humanity.