Alchemist Untamed / Perspective

Give People Back Their Brainpower

Why AI fluency should become a basic life skill

I am very good at beginnings and endings.

Give me an idea, a problem, a business opportunity, something I want to create, and I can usually see where I want it to go. The part that has always been harder for me is the messy middle: the dozens of small decisions, administrative details, unfinished thoughts, and logistical steps that stand between an idea and actually making it happen.

I suspect I am not alone in this.

Most of us aren't walking around all day solving fascinating problems. We are remembering appointments, searching for information we've already seen, sorting email, figuring out where to begin, rewriting something that doesn't quite say what we mean, moving information from one place to another, and repeatedly asking ourselves some version of: What was I supposed to do next?

For years, I thought of that as simply the cost of having a full life.

Then I started working seriously with artificial intelligence.

I expected AI to make me more productive. It has. But that has turned out to be the least interesting part of the story.

What I didn't anticipate was how much lighter my life would feel.

AI has become part researcher, part organizer, part planning assistant, part editor, part information filter and, occasionally, the thing that simply helps me figure out what comes next. It helps me turn large goals into realistic steps. It helps me navigate calendars and logistics. It can filter information before I spend my attention on it. It helps me research subjects I want to understand more deeply and find the questions I didn't know I should be asking.

And sometimes, when I know exactly what I want to say but have spent far too long staring at an empty page trying to find the perfect first sentence, it gives me a place to begin.

Not the finished thought. The jump start.

That distinction matters. Because somewhere in our conversation about artificial intelligence, we have begun confusing assistance with surrender. And I think that is causing us to miss one of the more interesting possibilities of this technology.

What If Productivity Is the Wrong Measure?

Much of the conversation about generative AI is still organized around productivity: How many more emails can we send? How quickly can we produce a presentation? How many hours can a company eliminate from a workflow?

There is real evidence behind some of this enthusiasm. A randomized experiment published in Science found professionals using ChatGPT completed a set of writing tasks about 40% faster, while the quality of their work increased by 18%. A large field study involving more than 5,000 customer-support agents found an average productivity increase of roughly 14%, with especially large gains among less experienced workers. Other field experiments have found that access to generative AI can reduce time spent on email and accelerate portions of everyday knowledge work.

Those findings matter. But I think we may be asking the wrong follow-up question.

When technology gives us an hour back, why is our first instinct to ask how much more work we can fit into it?

What if some of that hour belongs to the human being?

This is where my interest in AI intersects with my much longer-standing interest in human performance and cognitive capacity. Our attention is finite. So is our time. And while not every mentally demanding task is meaningless, not every task deserves equal access to our best thinking.

Remembering that a meeting moved from Tuesday to Thursday requires mental effort. So does locating an old document, sorting an inbox, formatting information, organizing research, or recreating something we've already done before. But is that where we want to spend our most valuable cognitive resources?

Increasingly, I don't.

One of the questions I have begun asking in my own work is deceptively simple: What should I never have to start from zero again?

That question has changed the way I think about AI.

Cognitive Offloading Is Not Cognitive Surrender

There is an obvious counterargument here, and I take it seriously. If we continually hand cognitive tasks to machines, what happens to our ability to perform those tasks ourselves?

Humans have always used tools to extend cognition. We write lists because we don't want to remember everything. Calendars externalize time. Calculators perform arithmetic. GPS changed the way many of us navigate. Search engines changed the way we retrieve information.

Artificial intelligence represents a more complicated step because it can assist not only with remembering and retrieving but also with language, analysis, synthesis, and increasingly complex forms of reasoning. That deserves caution.

Research is beginning to show why. A 2025 study from researchers at Carnegie Mellon University and Microsoft Research examined hundreds of real-world examples of knowledge workers using generative AI. Greater confidence in AI was associated with less self-reported critical-thinking effort. At the same time, the researchers found that AI did not simply eliminate thinking; in many cases, it changed the nature of it, shifting effort toward activities such as verifying information, integrating responses, and overseeing the task.

More recent reviews of the research suggest a similar tension. Generative AI can support critical and creative thinking when people use it within structured, questioning, and reflective processes. Passive use and overreliance, however, can produce something very different.

So I don't think the serious question is whether AI will make us smarter or make us stupid. I think what AI does to our thinking may depend, in meaningful part, on how we learn to use it.

And that is why I have become increasingly convinced that AI fluency needs to become a basic life skill.

Fluency Is Not Knowing How to Write a Clever Prompt

There is a persistent fantasy about AI that you should be able to type your desired outcome into a box and receive something brilliant in return. When that doesn't happen, people conclude that AI isn't particularly useful. Or they accept the mediocre answer and assume that is what working with AI looks like.

Neither represents fluency.

The way I use AI today bears little resemblance to the way I first experimented with it. I have learned what kinds of context matter. I question answers. I redirect. I ask for evidence. I reject things that don't sound like me. I push back when something doesn't make sense. Sometimes I ask the system to challenge my own assumptions rather than reinforce them.

The human brain hasn't disappeared from that process. It has become more important.

My working definition of AI fluency is the ability to work deliberately with intelligent systems: to provide useful context, ask better questions, challenge outputs, verify what matters, protect what is sensitive, refine rather than simply accept, and recognize which decisions still belong to you.

Emerging institutional definitions are moving in a similar direction. The OECD and European Commission's AI literacy work, for example, goes well beyond knowing how to operate an AI tool. It includes understanding AI, critically evaluating its outputs, using it responsibly, and making informed decisions about its risks and opportunities.

That last piece is particularly important to me. Because AI fluency doesn't require believing that AI is wonderful. It requires understanding enough about it to decide for yourself.

You Cannot Meaningfully Accept or Reject Something You Don't Understand

I understand why people are afraid. Some of the concerns are entirely legitimate. AI systems can produce false information with extraordinary confidence. Privacy matters. Bias matters. Security matters. Overreliance matters. Increasingly autonomous systems introduce questions far more consequential than whether a chatbot writes a good email.

Even NIST's framework for generative AI risk recognizes both sides of the problem: humans can place too much trust in AI systems, but we can also become so averse to them that we fail to use beneficial applications.

Fear, then, is not necessarily irrational. But fear is a poor substitute for fluency.

Throughout periods of significant technological change, the ability to understand and navigate new systems has affected who benefits from them and who does not. Early evidence suggests AI may be no exception. Adoption already differs across education, income, geography, and levels of digital literacy.

That makes AI literacy more than a workplace issue. It may become an equity issue.

Imagine one group of people using intelligent systems to help them navigate complicated information, learn unfamiliar subjects, prepare for important conversations, organize their lives, communicate more effectively, and understand the options available to them. Then imagine another group encountering those same systems primarily as something mysterious happening around them—or to them.

I don't think the answer is to insist that everyone embrace AI. I think the answer is to make sure people understand it well enough to exercise genuine choice.

You can learn how AI works and decide that there are places in your life where you don't want it. That's agency. You can understand its capabilities and decide that certain information is too private to share, certain decisions too important to delegate, or certain skills too valuable to stop practicing. That's agency too.

But you cannot exercise meaningful discernment about something you have never been given the opportunity to understand.

The Work Worth Protecting

There are things I don't want AI to take from me. I don't want it to determine what I believe. I don't want it to replace curiosity with instant answers or turn research into the passive acceptance of a beautifully formatted summary. I don't want it to eliminate the productive discomfort of wrestling with an idea until I understand what I actually think. And I certainly don't want it making the moral decisions that belong to human beings.

What I do want is help clearing some of the debris around those things.

I want the research organized so I can spend more time interrogating it. I want a first draft I can disagree with. I want the logistical pieces of a complicated week arranged so that I can actually be present when I arrive somewhere. I want to spend less time remembering what I was supposed to do and more time deciding what is worth doing.

This has become surprisingly personal for me. As AI has taken some of the logistical and administrative weight out of my life, I haven't experienced myself becoming less engaged with it. I've experienced the opposite.

I have more room for movement and health. More ability to be present as a parent. More capacity to learn. More opportunity to develop ideas that previously might have remained unfinished because there wasn't enough time to chase them. I can research subjects that matter to my work and clients more deeply. I can turn an idea into an actionable plan before it disappears beneath everything else demanding attention.

AI hasn't removed the human component from my life. Used well, it has helped me make more room for it.

That is an important distinction, because the productivity narrative gives us a remarkably unimaginative vision of what we might do with this technology. We create an extraordinary tool capable of saving human time, and our grand ambition is to answer more email? Surely we can think bigger than that.

Don't Spend the Entire Dividend on More Work

If intelligent technology really does allow us to reclaim portions of our time and attention, then perhaps we need to start talking about what I think of as the human dividend.

What happens to the capacity we get back?

Businesses will understandably convert some of it into productivity. Individuals will too. But we don't have to spend all of it there.

Maybe some of that capacity goes toward thinking deeply enough to change our minds. Maybe it goes toward reading something longer than a screenful, learning an unfamiliar subject, walking outside, making dinner, exercising, creating something original, sitting across from someone we love without mentally sorting tomorrow's obligations.

Maybe leaders use it not to squeeze another meeting into the afternoon but to actually think about the decisions only they can make. Maybe parents use it to carry less logistical information in their heads so they can be more present with their children. Maybe a person who has always found certain administrative tasks overwhelming gets a kind of scaffolding that allows their actual abilities to become more visible.

And perhaps this is where the conversation about AI and human performance needs to go next. Not simply: How much more can we do? But: What deserves to be done by us at all?

Give People Back Their Brainpower

Artificial intelligence may become one of the most consequential cognitive tools humans have created. That is precisely why I don't think fluency should belong only to technologists, executives, early adopters, or people already inclined to experiment with it.

People need enough understanding to decide what role they want AI to play in their own lives. They need to understand what it can do, what it cannot reliably do, what should be questioned, what should be protected, and where their own judgment becomes more—not less—important.

And then perhaps we need one more kind of fluency. We need to become fluent in deciding what to do with the capacity technology gives back.

Because the question shouldn't simply be how many reports, emails, presentations, and tasks we can now fit into a human life. Maybe the better questions are these:

What should I never have to start from zero again?

And once I don't:

What will I do with the brainpower I get back?

Shine bright and step into the light,
Chelsea

Take small steps to big wins

---

Sources & Further Reading

Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187–192.

Brynjolfsson, E., Li, D., & Raymond, L. R. Generative AI at Work.

Lee, H.-P., et al. (2025). The Impact of Generative AI on Critical Thinking. Microsoft Research / Carnegie Mellon University.

OECD & European Commission (2026). Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education.

National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (NIST AI 600-1).