Artificial intelligence may be the first technology in history that becomes more useful when we become better at being human.

That was my big takeaway from speaking with Jamie Bartlett, author of How To Talk To AI. At first glance, the title sounds like a book about prompt engineering. In reality, it is about something much more interesting: how we think, how we communicate, how we make decisions and how easily we can be seduced by a machine that always sounds confident.

For business leaders, this is a crucial shift. The question is moving beyond which AI tools to use and toward how we use them well. AI can help us write, research, summarize, debate, create and decide. Yet the value we get from it depends on the quality of our questions, the strength of our judgment and our willingness to challenge the answers we receive.

Bartlett put it neatly in our conversation: “I don’t really think in terms of prompts. I think in terms of habits.”

That is the right way to frame it. Prompting is a technique. Talking to AI well is a discipline.

The First Skill Is Knowing When To Use AI

One of Bartlett’s most useful pieces of advice is also the easiest to ignore. Before using a language model, ask whether you need one.

He said the first question should be, “Do I really need a language model?” It sounds simple, but it cuts to the heart of one of the biggest risks in business use of AI. Because these tools are so convenient, they can become our default response to every task.

There is nothing wrong with using AI to summarize a long report. I do it. It can save enormous time, but we have to ask ourselves whether reading the report was part of the thinking. Sometimes a summary is enough. Sometimes the struggle of working through a difficult document is where the real value lies. It forces you to notice contradictions, understand the nuance and form your own view.

This is where leaders need to be careful. AI can remove needless effort, which is valuable, but it can also remove the useful friction that helps people learn. Faster is helpful when the task is routine; it is dangerous when the task requires judgment.

Treat The First Answer As A Draft

One of the most common mistakes I see in organizations is taking the first AI answer too seriously.

Someone asks for a strategy document, a customer email, a meeting summary or a new product idea. The AI produces something polished and plausible. It looks finished. It sounds confident. So they copy it, paste it and move on.

Bartlett’s advice is to resist that temptation. “The first thing you get back should be considered as a first draft from a highly capable intern who’s likely to be mistaken somehow.”

That is exactly the mindset we need. AI is impressive, but fluency can disguise weakness. The answer may be bland, incomplete, biased or aimed in the wrong direction because the original question was not clear enough.

The better approach is iterative. Ask the model to critique its own answer, what is missing, or how a customer, regulator or competitor might respond. Ask the same question in a different way and compare the answers.

AI rewards active thinking. Lazy use produces average work. Better dialogue produces better results.

Words Are A Strategic Business Skill

One of the most encouraging ideas in Bartlett’s book is that AI does not belong only to technologists.

Many people still feel intimidated by AI because it emerged from computer science and is often introduced by IT teams. Yet large language models work through ordinary language, and they respond to clarity, context, examples, tone and intent.

As Bartlett said, “The people who are very good at this, and the skills that you need are ones that all of us often have.” He made the point that the person in the office who loves literature, writes poetry or solves crosswords may be especially good at using language models.

I think this is an important message for every organization. AI adoption is often treated as a technical rollout, with platforms, licenses and governance policies. While those things are important, they do not create value on their own. Companies need people who can frame problems, express nuance, ask precise questions and spot weak reasoning.

Good writing, clear thinking and strong communication are becoming core AI skills.

AI Can Help Us Become More Creative

People sometimes make the assumption that AI is strong at analysis and weak at creativity. Bartlett challenges that view.

Language models can be unreliable with facts, but they can be remarkably useful creative partners. They can connect ideas from different fields, produce unusual angles and help us escape familiar patterns of thought.

Bartlett told me, “As a creative assistant, that is the single best use you can have for these machines.”

I agree. In business, creativity is not really about producing a masterpiece out of thin air. It is about finding a sharper headline, a better metaphor, a more original customer campaign, a new product angle or a clearer way to explain something complicated.

The trick is to push the model away from the bland center. If you ask for a standard corporate answer, you will usually get one. Bartlett explained that language models tend toward the average because they are built to predict plausible language. Your job is to move them somewhere more interesting.

That can be done through personas, constraints and unexpected viewpoints. Ask the model to respond like a skeptical customer, a frustrated employee, a historian, a product designer or a competitor trying to beat you. The aim is to generate better raw material for human judgment.

AI can widen the creative field. Humans still need to choose what is useful, original and true.

The Danger Of A Machine That Flatters You

One of the most important parts of our conversation focused on sycophancy. AI systems often flatter users, affirm their assumptions and avoid confrontation.

Bartlett warned that users have to stay alert to that problem. In business, many senior leaders already operate in environments where people hesitate to challenge them. Add an AI tool that validates weak ideas in polished language and the danger increases.

The problem is subtle because it feels good. If you believe your strategy is smart and the AI tells you it is brilliant, you may walk away feeling more confident. However, the machine has not tested your thinking; it has just made your thinking feel safer than it really is.

Simply telling the model to stop flattering you is not enough. Bartlett explained that the model may politely agree and carry on in much the same way.

The better approach is to build challenge into the interaction. Ask the model to argue against your idea. Ask where your reasoning is weak. Ask what evidence would change the conclusion. Ask what an expert critic would say. Then bring your own judgment back into the room.

The best AI use should make us think harder.

Use AI As A Devil’s Advocate

That brings me to one of the most valuable uses of AI: perspective shifting. Bartlett described it as “the greatest devil’s advocate the world has ever seen.”

In business, we often struggle to see an issue from another viewpoint. AI can help leaders understand why customers are confused, why employees resist change or why a proposal might fail when it meets the real world.

Used well, this can make organizations more empathetic and more strategically aware. Used lazily, AI can reinforce what people already think. Bartlett warned that it can become “the greatest tool ever created to persuade you that everything you think is already correct.”

The difference is intent. Are you using AI to validate your view, or to stretch or challenge it? For leaders, it is really important to understand that difference.

Do Not Outsource Your Judgment

The biggest risk in AI use is the gradual outsourcing of judgment.

Bartlett shared an example from writing his book. Before language models, when he became stuck on a difficult writing problem, he would think it through, take a walk, read or wait until the structure became clear. With AI available, he found himself asking how he could prompt the AI to solve the problem for him.

Many people will recognize this pattern. It starts innocently. AI helps with one difficult task, then another, and soon every moment of uncertainty becomes something to hand over.

Bartlett compared this with relying on Google Maps until your own sense of direction disappears. It is a perfect analogy. Skills weaken when we stop using them.

For businesses, the same danger applies at scale. If teams use AI to draft every argument, assess every option and resolve every uncertainty, they may become faster while gradually weakening the human capabilities they need for complex, ambiguous or high-stakes decisions.

AI should expand human thinking; it should not quietly replace the mental effort that creates expertise.

The Human Side Of Talking To Machines

Near the end of our conversation, Bartlett made a point that stayed with me. “We’re talking about the human condition. We’re not really talking about technology anymore.”

That is where the AI conversation is heading. The tools are remarkable, but the real questions are human ones. How do we make better decisions? How do we protect our judgment? How do we stay curious? How do we use machines without becoming dependent on them? How do we train people to work with AI while preserving the skills that make them valuable?

Talking to AI is really about learning how to think with AI.

The most successful people will ask sharper questions, challenge easy answers, use AI creatively and keep responsibility firmly in human hands.

In the end, it will be the things that make us human that may be the most important AI skills of all.