Artificial Intelligence amplifies those who think – and replaces those who don’t

In 2023, New York attorney Steven Schwartz was handed a legal research task and decided to use ChatGPT to speed things up. The tool delivered exactly what he seemed to need: well-referenced cases, judges’ names, docket numbers, ruling dates. Everything fluent, everything convincing. Schwartz included six of those citations in his brief and filed it in federal court in a personal injury case against Colombian airline Avianca.

Avianca’s lawyers tried to locate the cases. They found none, because none of them existed. ChatGPT had invented every single one – with the same confidence it would have used to cite any real ruling.

When confronted, Schwartz admitted he simply had no idea AI could do that and claimed to have acted in good faith when making the mistake. The judge disagreed with the good faith claim – and handed down a $5,000 fine. The case made headlines around the world. And still, it wasn’t enough to change a lot of people’s behavior.

That’s the starting point for a conversation we need to have about artificial intelligence – not the one about whether the technology is good or bad, but the one that actually matters: what happens when we delegate our thinking to a tool that doesn’t think?

We’re living through a period of rapid AI adoption. These tools are everywhere: drafting emails, analyzing data, creating content, supporting strategic planning. And the promise is seductive – more speed, more capability, more output with less effort. But there’s a condition nobody puts on the label: for AI to amplify what you do, you need to know how to think.

What AI Really Is – and What It Isn’t

Before talking about how to use AI well, it helps to understand what it actually is.

A language model like ChatGPT, Claude, or Gemini doesn’t “know” things the way a person does. It doesn’t understand the world – it predicts. Based on enormous volumes of text, these systems learn statistical patterns: which words tend to appear together, which argument structures are common, which answers typically follow certain questions.

That means AI generates what is probable – not necessarily what is true or right for your context.

This is a fundamental distinction. When you ask a human expert something, they draw on real experience, situated reasoning, and awareness of their own limits. When you ask AI, it generates a plausible response – and plausible is not the same as correct.

This isn’t a bug to be fixed in the next version. It’s the nature of the technology. And understanding it completely changes how you should interact with it.

Why Mastering AI Is a Human Skill, Not a Technical One

There’s a common narrative that “knowing how to use AI” is a technical competency – that it’s enough to learn how to write good prompts, know the right tools, understand the available models. That helps, but it’s superficial.

The core competency for using AI well is something else: the ability to think clearly about what you want, critically evaluate what you receive, and know when to trust, and when not to. That requires three things AI cannot do for you.

Clarity of purpose. Before opening any tool, you need to know what you’re trying to solve. The vaguer your question, the more generic the answer. AI is extraordinarily good at filling in blanks – and that’s a problem when you don’t know exactly which blank you want filled.

A knowledge base to evaluate with. An AI response looks good when you have no foundation to question it. Someone with little background on a topic will accept anything that sounds coherent. Someone with depth will notice when something is wrong, incomplete, or out of context. The paradox is that AI is most useful to those who already know the most – precisely because they know what to ask and how to filter what they receive.

Awareness of your own context. AI doesn’t know who you are, what your company does, who your audience is, or what your real constraints are. It responds to a generic version of your problem. You’re the one who has to bring what’s specific – and that’s where the value lives.

Those who don’t think well receive mediocre answers and don’t realize it. Those who think well use AI as a powerful amplifier. The difference isn’t in the tool. It’s in the person using it.

The Risks of Using AI Without Critical Thinking

You don’t have to be an alarmist to take seriously the risks of unreflective AI use. They’re real, concrete, and already happening.

Language models invent data, citations, references, and facts with the same fluency they use to present accurate information. There’s no warning signal. A false response and a true one carry the same confident tone. Those who don’t verify, propagate error.

Research from Stanford’s RegLab and Institute for Human-Centered AI analyzed 200,000 legal queries submitted to language models and found hallucination rates between 69% and 88% – and when asked about a court’s core ruling, models fabricated information at least 75% of the time.

The Schwartz case was a warning – but it didn’t change much. In October 2025, a California attorney was fined $10,000 after 21 of 23 citations in his appellate brief turned out to be fabricated by ChatGPT – the largest such penalty in the state at the time. Hundreds of similar cases followed, and the pattern, in all cases, was the same: qualified professionals who confused fluency with the tool with reliability. These two things are not the same.

AI was trained on human-generated text – and human text carries prejudices, dominant perspectives, and historical blind spots. When you use AI without questioning it, you’re often reproducing and legitimizing those biases without realizing it. Stanford HAI’s 2025 annual AI Index Report documents persistent bias concerns across multiple domains and model families – including systematic differences in feedback quality and professional recommendations depending on demographic cues embedded in the text.

When many people use the same tools to think through the same problems, the answers start to look alike. Business strategies that sound identical. Texts with the same rhythm. Solutions that follow the same patterns. AI, used without discernment, doesn’t expand the space of thinking – it compresses it.

Every skill you completely outsource, you stop developing. If you never draft an argument from scratch again, your ability to articulate ideas weakens. If you never analyze a problem without automated support, your analytical capacity degrades. Short-term comfort can be costly in the medium term. The cost of not verifying is also concrete: 2025 data shows knowledge workers spend an average of 4.3 hours per week fact-checking AI outputs – and still, 47% of enterprise users admitted to making at least one major business decision based on incorrect AI-generated information.

These risks don’t mean you should avoid AI. They mean you should use it consciously – and that starts with knowing the risks exist.

How to Use AI to Expand – Not Replace – Your Capacity

The good news is that there’s a clear way to integrate AI into your work and your creative process so it serves your thinking rather than substitutes for it. Here are some practical principles.

Use AI to expand, not to outsource. Before asking AI to do something for you, try sketching your own perspective first. Then use AI to challenge, complement, or deepen it. The order matters: when you start from your own point of view, the tool works in service of your ideas. When you start from the AI’s response, your ideas become hostage to it.

Treat responses as drafts, not deliverables. AI is excellent at generating raw material – a list of possibilities, a structural outline, a synthesis of information. But the work of refining, contextualizing, and making something yours is irreplaceable. If you hand over AI output without going through that process, you’re handing over something generic with your name on it.

Actively question what you receive. Develop the habit of asking: Is this correct? Does this apply to my context? Is there a perspective that wasn’t considered here? AI won’t raise those questions for you – it will answer what you asked, not what you should have asked.

Verify facts and sources. Never use a data point, citation, or reference generated by AI without confirming it at the original source. This isn’t paranoid distrust, it’s basic intellectual rigor. The cost of verifying is small. The cost of propagating an error can be significant.

Use AI to think, not just to produce. Some of the most powerful applications of AI aren’t the most visible ones. Ask it to argue against your hypothesis. Ask it to identify blind spots in a strategy. Ask it to simulate how different audiences might react to an idea. This doesn’t replace your judgment, it exercises it.

The Human Edge AI Doesn’t Replicate

There’s a recurring conversation about what’s left for humans in a world with AI. The answer isn’t sentimental – it’s functional.

AI has no lived experience. It doesn’t know what it’s like to make a difficult decision with incomplete information, to negotiate in a real high-stakes situation, to create something knowing it will be judged, or to learn from a failure that actually cost something. It can describe these things. It cannot inhabit them.

AI has no situated ethical judgment. It can process abstract ethical principles, but it can’t navigate the nuances of a concrete situation with all its human, relational, and contextual variables. People who make decisions that matter need judgment, and judgment is cultivated over time, through attention, reflection, and practice.

AI has no perspective of its own. It has patterns derived from other people’s perspectives. Originality – the capacity to see something from an angle no one had seen before, or to connect ideas in unexpected ways – comes from a singular history, a unique way of being in the world. That’s not replicable.

And there’s something subtler: AI has no intention. It responds. You decide. And it’s precisely there, in the capacity to decide with intention, to choose what to do with what the technology offers, that the human edge no tool can replace.

The Question Worth Sitting With

Artificial intelligence is, genuinely, one of the most transformative technologies ever to emerge. It can expand what you produce, accelerate what you learn, and push the boundaries of what you’re able to create.

But it only does that for people who are willing to think.

The tension won’t go away. Between trusting and questioning, between delegating and owning, between using the tool and being used by it – that’s a tension each person will have to navigate on their own, every day, in every context.

The question worth leaving isn’t “are you using AI?” It’s a different one: when you use AI, who’s in charge, you or it?