There are two phenomena happening here. When you are freed up from lower-level work — say, email writing — doing exclusively high-level work, such as analyzing complex data sets, is very taxing on the brain. Second, simply being given a tool that helps us do so much more creates a double-edged sword where people are more engaged and excited about their work, but also work longer hours because the volume has increased overall.

In their research, Aruna Ranganathan and Xingqi Maggie Ye found in an eight-month ethnographic study of 200 employees that AI usage intensified work rather than making it easier. Other research from BCG has found a “brain fry” effect: Using AI well, on top of performing our other tasks, is making work doubly or triply effortful, leading to more errors and poorer outcomes.

At NLI, we investigate what’s going on in the brain and devise solutions for making work more manageable within the brain’s natural limits. Leaders who want their teams to benefit from AI’s time-saving capabilities, without fighting burnout in the process, can follow a few insights.

Watching AI parse entire documents and generate complex analyses in seconds makes it easy to forget our own cognitive limits. The brain isn’t an infinite computational machine. In fact, it gets quite tired very quickly; just ask your brain to remember a list of 10 items, and you’ll immediately notice how fast it just kind of falters.

The brain fatigues for a few reasons. One is the limitation posed by working memory; we’re just not all that good at holding lots of information in mind. For years, experts thought the brain was capable of keeping around seven items in mind at once; however, recent studies have shown the true number may be closer to three to five.

More recently, neuroscientists have discovered a layer of “intermediate term memory” as well, or the brain’s capacity to hold information in mind over a period of hours. Intermediate term memory is also highly constrained, often more than we like to believe. We may think of ourselves as expert multi-taskers, but science shows that just isn’t the case. Our brains are poor jugglers of information.

Another factor is the cost of task switching. Even if it may feel like we are effortlessly switching between jobs like tabs on our Internet, browser, research has shown it can take more than 20 minutes to fully recover our focus when we go between disparate tasks—such as toggling between AI prompting and applying its outputs. Add in a meeting or two interrupting these tasks, and we have little hope of getting much of anything done over a full workday.

And yet, employees have to remain productive, so what happens? To make room for new processing, other inputs have to go, which is how over-worked, overworked employees—even all-star team members—end up dropping the ball on minor details.

Using AI within an already busy workday taxes our brain. The technology eats up more space in our overall cognitive processing, especially as we fill open time slots with additional prompting, so our brains never feel fully rested. We started using AI to become more productive and focused; instead we’re losing our edge and fatiguing along the way.

A key finding from the research into creativity and innovation is that “Eureka!” moments don’t happen in noisy brains. They happen when the brain is quiet, perhaps when the person is taking a shower or out walking the dog. During these periods of quiet, the unconscious mind has a chance to make connections that the noisy conscious mind can’t perceive, because it’s too busy attending to important work matters.

Organizations should make it a priority to carve out quiet time free from meetings or AI use, during which employees can work heads-down or free-associate about their work. Crucially, this time must be systematized and held sacred so that it’s not lost to last-minute meeting requests or seemingly urgent to-dos that all too easily wipe away seemingly disposable “free” time.

If someone is managing eight AI agents all day, every day, their work is almost sure to be too taxing under normal conditions. Going forward, work may need to be measured more by outcomes rather than inputs and hours worked. At the same time, individuals must develop the self-awareness to know when to take breaks, to take a nap or go for a walk or other facets of the Healthy Mind Platter, so their brains can recharge.

Organizations should also educate people on using AI so that it’s enhancing their work, not strictly multiplying it. Our research shows a small percentage of users work with AI as a partner; the majority offload their work, which creates an extra layer of work to manage the AI. The key skill in this case is metacognition: thinking about one’s own thinking to ask AI to improve current hypotheses, solutions, and conclusions.

When teams can pair these two strategies—dedicated quiet time for insight generation and using AI to enhance one’s thinking—they should see brain drain reverse into a worthy investment of time and energy that treats AI as a sophisticated tool. The brain didn’t evolve for infinite prompting. But with the right guardrails, it doesn’t have to be.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.