It’s usually not a good day when your accountant says something that sends a shiver up your spine. Yet this had nothing to do with my finances and everything to do with a decision big technology firms had made that I somehow hadn’t noticed.
“It’s incredibly annoying!” he exclaimed when asked about his GenAI use, something I’ve been doing with everyone lately. “Everything I do, even just writing an email, AI keeps bugging me and offering to help. I have to make a conscious effort to say ‘no.’”
As someone who’s made a career out of understanding how we think and behave, I began to picture what this meant for the brain of anyone who uses a computer or phone, which is basically everyone, and my brain started to explode.
It turns out that the people building our digital world made the quiet decision around 2024 to have generative artificial intelligence offer help on nearly every task, rather than let users seek input as needed. Whether you are writing an email, building a presentation, creating an image, or developing a critical business strategy, GenAI now pops in and encourages you to use it so you don’t have to think as hard. Put simply, this means humans today need to put in mental effort to say “no” to GenAI thinking for them.
The default effect
In behavioral science this relates to something called the “default effect.” This refers to our tendency, when given a choice, to passively stick with a preselected option rather than make an active switch to another option.
The default effect is a powerful phenomenon in decision-making science. Once you know people tend to go with whatever choice is made for them, you can build choice architecture that requires people to opt out, rather than opt in. (For this reason, I like to refer to the default effect as the “opt-out effect.”)
Want to maximize your subscription rate? Charge people’s credit cards automatically each month unless they do something to cancel. Want folks to donate their organs? Make it the default choice on their driver’s license form, instead of requiring them to sign up. Given that people prefer to do nothing rather than something, uptake of just about anything is radically higher with this opt-out strategy. In fact for organ donation, 90% of people donate in countries requiring people to opt out, compared to around 10% when people need to opt in.
If you are close to my generation, you probably remember the 2004 documentary “Super Size Me” by the late Morgan Spurlock. Starting around 1978, McDonald’s cashiers were required to ask most customers, “Would you like me to supersize that?” More often than not, when people were asked to supersize their order, they did.
The science of why this works so well is simple. Not having to think is one of the most attractive and delicious things we can offer to the human brain. Built over millennia when metabolic resources were scarce, strong reward networks light up anytime we detect an opportunity to achieve a goal with fewer physical or mental resources. A chain saw instead of an axe? Brilliant. A vacuum instead of a broom? Fantastic. An AI instead of your brain? Well, obviously.
The cognitive effort required to opt out is just one reason the default effect works. The second is that we also assume getting help from GenAI is the right decision. After all, your company has probably been telling you to use GenAI as much as possible. The third reason is the belief that using GenAI instead of thinking is the social norm, and adhering to social norms is one of the biggest drivers of human behavior.
Retailers of fast food learned to make extra billions by hacking the brain’s natural desire to conserve resources, and today’s tech firms are doing the same. Today, this very second, they are nudging anyone using technology anywhere to think less for themselves. That’s now the easy option. The harder option—the one that requires effort—is to think for yourself.
The trouble with this strategy, at least for the well-being of the general public, is that it works. In the fast-food world, people end up eating a lot more. In the digital world, people end up thinking a lot less.
AI can erode our skills
You might be wondering why this is an issue. Doesn’t using a calculator to save your brain from doing complex math help you solve higher-order problems? And can’t we just work out when we should and shouldn’t use these tools, and use them at just the right points?
Research shows it’s a little trickier than you might think. Multiple studies show that offloading a skill to AI results in losing that skill remarkably quickly. One study showed that after AI tools were added to physician’s workflows, their ability to detect specific health indicators dropped significantly. And in another, computer coding abilities were diminished once people received help from AI, even in people with years of experience.
One way to understand this is as a motivation problem. Imagine it usually takes you three hours to build a presentation, and then you find a way to do it in 10 seconds. After a few weeks of doing this in 10 seconds, it feels excruciating to go back to the three hours. Your brain rebels. It sends an alert that something is wrong, that there’s a way you could be conserving resources. The outcome is similar in the brain to experiencing actual pain, in that it’s incredibly hard to focus. As well as being so painful, over those few weeks your brain started the process of recycling the pathways you had built for your multihour process. It was so sure this skill would not be needed in the future that it started to delete these pathways from your networks.
This is happening in a matter of weeks or months, even on a skill built over decades. Being highly proficient in a task isn’t being as protective as we imagine.
Here’s where it actually gets scary. A recent study by Boston Consulting Group explored the kinds of skills we are most likely to lose when we start to use GenAI ubiquitously. The two skills most at risk? “Judgment and decision-making” and “problem understanding and framing.” These are the exact skills necessary to decide whether to use GenAI or not, and whether its output is right. The result is a rapidly increasing use of these tools for anything and everything. It’s a fast downward spiral.
This means that using GenAI a lot makes us worse at discerning when and how to use GenAI. When you combine this with help from GenAI being offered automatically in nearly every digital task, we end up on a slippery slope to using it for everything. Why does that matter? Well, consider a second study in 2026 of 3,000 people. In summary, people answered questions with or without help from AI. People using AI to answer questions said “I don’t know” only 3% of the time, compared with “I don’t know” 44% of the time when not using AI. In other words, people felt like they had answers to their questions almost all the time. Meanwhile, people’s confidence in what they shared when using AI rose from 30% to 76%. But here’s the kicker: When using AI, being actually correct dropped from 27% to 9%.
Think about that for a moment. . . . Take a lack of curiosity, mix in some false confidence, and then add in being wrong, a lot. Now multiply this by around 6 billion people, who are increasingly using this tool for everything. Maybe you see why my brain started melting down talking to my accountant.
Using GenAI for tasks here and there isn’t a big issue. The challenge is the consistent reliance on it. Or as the study above said, “As AI suggestions grow ubiquitous and unsolicited, they may not simply affect answer accuracy; they may even alter the metacognitive threshold at which people decide if they know enough to answer.”
And remember, one of the skills we’re losing fastest is judgment, which includes knowing how to best use this technology itself. In short, technology firms are offering to supersize every mouthful of every meal, to every person, all day, every day. And people are gobbling it down.
We need healthy AI consumption to be the default
I am not suggesting that we remove GenAI from our diet. There are healthy ways to consume it. Yet we can’t leave it to the big technology companies to do the right thing and nudge us toward healthy use. They never did with social media, which negatively impacted the mental health of an entire generation, and they probably won’t with this technology now.
I propose that it is up to the overseers of the digital ecosystems in our schools and our organizations—the principals and administrators, the CEOs, the COOs, the CTOs, and the CHROs—to do the right thing.
The first thing is we need to stop encouraging the widest possible use of these technologies, and instead promote more thoughtful use, with humans still doing the thinking. This is “human in the lead,” where artificial intelligence is a tool to help us think better. As part of this, we need to show people how to use these tools well, which means doing harder thinking. The mission for HR and talent functions then becomes making hard thinking easier, instead of not thinking being the default.
Second, we need to turn off the automatic question—“Would you like to super size that?”—at a systems level inside the digital ecosystems in our companies and in our schools, so that people are required to make the conscious decision to get help from AI. If people have to choose to use GenAI for that email, document or presentation, that moment of choice gives them space to make the right decision. The default should be thinking well, not offloading.
GenAI is like when cars replaced horses. We’re all going faster now. Letting kids drive a car is not a great idea, nor is letting cars go as fast as possible. We still have time to put in place some road rules and speed limits. McDonald’s stopped offering the “supersize me” option when the documentary came out, roughly 20 years after starting the practice. Someone, somewhere, in that organization made the decision that despite it being excellent for profits, it was the wrong thing to do. I’m not sure we want to wait two decades for today’s tech firms to decide the same.