The Paradoxical Art of Doing and Not Doing

I came across a post on Reddit linking to a Tom’s Hardware article about a study published in the Harvard Business Review. Researchers from UC Berkeley spent eight months embedded in a 200-person tech company, watching what happened when employees genuinely embraced AI tools. The finding wasn’t that AI failed. It’s that AI succeeded — and that success made everything worse.

Workers took on more tasks. They expanded their roles. They filled lunch breaks and coffee waits with prompts. They multitasked across AI-assisted workflows. And they burned out.

Nobody told them to do this. The pressure came from within.

That finding reminded me of an Aldous Huxley quote I’ve been sitting with:

The harder we try with the conscious will to do something, the less we shall succeed. Proficiency and results come only to those who have learned the paradoxical art of doing and not doing, or combining relaxation with activity.

It turns out there’s real science behind this — and it connects directly to what’s happening with AI in the workplace right now.

Your brain can still learn. That’s the good news.

Before getting into what goes wrong with performance under pressure, it’s worth starting with something encouraging: your brain is still capable of learning new things. At any age.

Neuroplasticity — the brain’s ability to form new neural pathways — doesn’t shut off when you turn 30, or 40, or 60. Research has consistently shown that older adults can pick up new languages, learn instruments, and develop complex cognitive skills. The brain keeps rewiring. The acquisition phase is real, and deliberate effort is exactly what drives it.

You have to try hard to learn something new. That’s not a weakness — that’s how the system works.

So if you’re feeling behind on AI, or overwhelmed by a new tool, or unsure whether you can pick up a skill that seems to belong to someone younger or more technical: you can. Your brain is built for it. The learning will take conscious effort, and that effort is legitimate and necessary.

The key is that learning and performing are two different things — and they follow different rules.

Once you’ve learned something, forcing it breaks it

Here’s where Huxley’s paradox kicks in, and where the psychology gets interesting.

Yerkes-Dodson Law. Performance on complex tasks peaks at moderate effort and degrades when you push harder. There’s an optimum, and overshooting it doesn’t get you more — it gets you less. Simple tasks tolerate brute force. Skilled work doesn’t.

Flow states. Csikszentmihalyi’s research on peak performance found that the best work happens when conscious self-monitoring drops away. You’re fully engaged but not trying in the effortful, self-conscious sense. You’re in it, not watching yourself be in it.

Choking under pressure. Sports psychology has documented this extensively. Expert performers fail when they shift from automatic processing back to explicit, step-by-step conscious control. A skilled free-throw shooter misses when they start thinking about their mechanics. The conscious override disrupts compiled skill.

Ironic process theory. Wegner’s research showed that trying to suppress or force a mental state tends to produce the opposite. The harder you try not to think about something, the more you think about it. The harder you will yourself to perform, the more tension you create.

The pattern is consistent: for already-learned skills, conscious will is the wrong tool. Relaxed execution is the right one.

AI removed the governor

This is where the HBR study lands squarely in Huxley’s territory.

Before AI, there were natural speed limits on how much a knowledge worker could produce in a day. The friction of research, writing, coordination, and context-switching imposed a ceiling. That ceiling was sometimes frustrating, but it also functioned as a governor — a built-in regulator that kept effort within a sustainable range.

AI removed the governor.

The study found that workers who embraced AI didn’t use the freed-up time to rest or focus. They used it to take on more. Product managers started writing code. Researchers took on engineering tasks. People attempted work they would have previously outsourced, deferred, or avoided. The scope of their roles expanded — not because management demanded it, but because the tools made “more” feel possible.

And here’s where it connects to the psychology: every one of those expansions pushed workers further past their Yerkes-Dodson optimum. More tasks, more context-switching, more cognitive load, more fragmented attention. The conditions for flow disappeared. The conditions for choking, fatigue, and error increased.

Workers who felt like they were getting a lot more done were often getting only a little more done — and burning out in the process.

The pressure that may not exist

One of the most striking aspects of the study is that the company didn’t mandate AI use. There were no new targets, no top-down pressure to adopt the tools. The intensification was self-imposed.

I think a lot of people are experiencing this right now. There’s a pervasive ambient pressure — from headlines, from LinkedIn, from conference talks, from the general cultural narrative — that says you need to be using AI for everything, right now, or you’ll be left behind. Leadership may or may not actually expect this. But the perception of that expectation is enough to drive the behavior.

And once you start driving harder, the ironic process kicks in: the effort to keep up creates the very stress that degrades your performance. You’re trying harder with the conscious will, and succeeding less.

A Hacker News commenter captured this perfectly in response to the study: expectations tripled, stress tripled, actual productivity went up maybe 10 percent. The felt pressure to justify the AI investment — whether that pressure is real or imagined — creates a vicious cycle of diminishing returns.

Learn deliberately. Perform relaxedly.

So what do you actually do with this?

Give yourself permission to learn incrementally. Your brain can handle new skills. Neuroplasticity is on your side. But learning takes time, and the acquisition phase requires the kind of conscious, deliberate effort that can’t be rushed. Pick one tool, one workflow, one skill. Get comfortable with it before adding the next thing. The compound effect of incremental learning is enormous — but only if you don’t burn yourself out trying to learn everything at once.

Recognize when you’ve shifted from learning to performing — and change modes. Once a skill is internalized, stop micromanaging yourself. Trust the compiled skill. This is Huxley’s “doing and not doing” — you’re engaged and active, but you’re not white-knuckling it. The relaxation isn’t laziness; it’s the operating condition where your best work happens.

Protect your downtime. The study found that AI’s conversational interface made it easy to blur work into rest time — a quick prompt at lunch, another while waiting for coffee. That erosion of recovery time is directly correlated with burnout. Rest isn’t wasted time. It’s where your brain consolidates learning, restores cognitive resources, and prepares for the next round of focused work. Filling it with prompts isn’t productivity. It’s debt.

Question the pressure. Ask yourself: is my leadership actually expecting me to use AI for everything, or am I responding to a narrative? Is the expectation real, or am I manufacturing urgency? A lot of the stress people feel around AI adoption is anticipatory — fear of falling behind, fear of being replaced, fear of looking like they’re not keeping up. Those fears are understandable, but they’re also the kind of conscious striving that Huxley warned about.

Batch your AI work. The HBR researchers suggested sending prompts in batches and actioning responses within specific windows, rather than maintaining a constant conversational back-and-forth. This is practical advice that maps directly onto the attention science: single-tasking with focused blocks produces better results than continuous partial attention across multiple AI-assisted workflows.

The synthesis

Huxley wasn’t being mystical. He was describing a neurological reality that modern research has confirmed from multiple angles: the systems that drive learning and the systems that drive peak performance are different, and the conscious will that serves one actively sabotages the other.

AI has made it easier to do more. It has not made it easier to know when to stop. And the inability to stop — to rest, to focus, to trust your own competence without constant augmentation — is what’s burning people out.

The paradoxical art isn’t doing nothing. It’s doing the right thing, at the right intensity, and then having the discipline to let it be enough.

Learn at your own pace. Perform with relaxed focus. And protect the spaces where you’re not doing anything at all. That’s where the real work happens.