How To Think With AI

When people ask how to use AI effectively, what they're really asking is how to think differently. AI is a new type of tool. Like any significant tool, it doesn't just change what we can do. It changes how we think about doing it.

The difference with AI, compared to most previous tools, is that it's deeply cognitive. It doesn't simply speed things up or increase precision. AI augments thinking itself. This makes it both powerful and subtle. You can't get the most out of it by treating it like an ordinary piece of software. You have to learn to think alongside it.

But how exactly do you learn to think with AI?

‍

Understand How AI "Thinks"

First, recognize that AI doesn't think like humans do. It works by predicting likely next words, recognizing patterns, and referencing vast databases of information. People, on the other hand, reason by intuition, analogy, and experience. They feel their way toward answers, guided by years of subconscious knowledge and emotional insight.

This difference is fundamental. AI can seem intelligent in many areas, sometimes even uncannily human. But beneath the surface, its reasoning is statistical and probabilistic. It doesn't truly understand anything. It has no intuitive grasp of concepts or meaning. It knows only patterns.

If you internalize this difference, you'll stop expecting AI to think like you. Instead, you'll see AI as an alien intelligence that complements human thinking precisely because it does not replicate it. This perspective unlocks a powerful new mental model. It encourages you to combine your intuitive strengths with AI's analytical breadth.

‍

Ask Better Questions

If there’s a single skill crucial to effectively using AI, it’s learning how to ask better questions. Unlike traditional software, where commands and queries must be precise, AI rewards thoughtful ambiguity. The better your questions, the richer the AI’s responses become.

A useful technique here is to ask AI questions as if talking to a well-informed outsider. Provide context, clarify your intentions, and articulate uncertainty. Instead of simply asking, "How do I write this email?" ask, "I want to write an email that persuades someone skeptical about X. What would resonate most deeply?" This sort of question taps into AI’s strength: pattern matching across vast domains.

Great questions to AI often look like conversations, not commands. The better your conversation, the better your outcome.

‍

Iterate and Explore with AI

Thinking with AI is fundamentally iterative. Rarely does the first output of an AI represent the best possible result. Instead, view AI as a partner for repeated exploration and refinement.

Your initial prompt provides a foundation. Subsequent questions help you refine and explore. Each iteration gives you greater clarity. Each refinement builds better results, feeding your own curiosity and critical thinking.

But here’s where we reach a broader point, because iteration isn’t just valuable for human thinking. Increasingly, it defines the very nature of the software systems we are building.

‍

AI Within Systems

Today’s software systems themselves are beginning to leverage AI in exactly this iterative way. AI models are being embedded directly into products, platforms, and services. Rather than static pieces of software that perform fixed tasks, modern systems use AI to dynamically interpret and adapt to user input, evolving constantly in response to new contexts.

The shift toward systems that continuously adapt using AI is profound. Previously, software did exactly what programmers told it to do, no more and no less. Now, systems are becoming intelligent platforms that interpret intentions, not just execute commands.

Think about how this looks in practice. Systems that once required precise configuration now ask users what outcome they desire, and figure out configuration details on their own. Tools that previously delivered fixed results now offer suggestions, learn from user feedback, and refine themselves continuously.
This is probably why platforms like Replit and Manus where the interface is interactive are more palatable and successful.

‍

User Interfaces are Changing

This shift fundamentally changes how we design user interfaces (we detailed this topic hereAugmentation Imperative). Traditional UI design emphasized clarity and control: buttons did exactly one thing, labels explained precisely what would happen, users managed every step explicitly.

Modern AI-driven interfaces instead emphasize explanation, exploration, and iteration. Instead of just presenting results, they also display their reasoning, revealing the AI’s internal logic. Users interact with these systems through conversation and refinement. Interfaces become less about control and more about collaboration.

A great example is a new generation of creative tools. A traditional image editor expects explicit instructions—select this, crop here, adjust that. AI-powered editors instead offer prompts: "What mood are you trying to convey?" The user then iterates with the AI, refining and adjusting until the right outcome emerges.

This new type of interface is powerful because it acknowledges that users often don’t know exactly what they want at first. They discover it through iteration: exactly the process AI supports best.

‍

AI as a New Paradigm for Software

This evolution toward AI-integrated systems is not limited to consumer tools. Enterprise software and business applications increasingly rely on AI-driven adaptability and reasoning. Instead of static rules and rigid workflows, the systems we build now incorporate learning, reasoning, and continuous improvement. They proactively anticipate user needs and even autonomously identify opportunities or risks we might never have considered (e.g., Workday provides an example of an AI agent that helps identify and prepare employees for leadership roles).

But integrating AI into software also reveals an inherent tension. Most software, especially critical systems, depends heavily on deterministic outcomes. Financial transactions, healthcare devices, aviation software—all these domains require exact, predictable behavior. AI, by contrast, is inherently probabilistic. Its strength comes precisely from estimating likely outcomes rather than guaranteeing specific ones. This fundamental mismatch makes using AI in deterministic systems particularly challenging.

To address this, developers must clearly separate where AI’s probabilistic reasoning is acceptable and where strict determinism is non-negotiable. It often means using AI for exploratory, predictive, or advisory functions rather than final decision-making authority. AI can recommend paths, generate ideas, or identify risks, but critical decisions must still flow through deterministic logic or human oversight. Understanding clearly how AI "thinks", as a statistical predictor rather than a guaranteed calculator, is essential when embedding AI into deterministic systems. This thoughtful, nuanced approach is the key to safely and effectively blending AI’s strengths with the deterministic needs of critical systems.

‍

Preparing for the AI Era

Learning to think with AI isn't merely beneficial—it’s essential. Whether you’re a software engineer, product designer, or business leader, your effectiveness will increasingly depend on your ability to collaborate with intelligent systems.

The true skill of the future will be combining human judgment with AI-powered analytics and reasoning. It will mean being comfortable iterating rapidly, articulating clear intentions, and refining outcomes collaboratively.

Most importantly, it means understanding both the strengths and limitations of AI, treating it as a partner, not a replacement. As AI becomes embedded deeply into our daily tools and systems, the individuals and organizations that thrive will be those who master the art of thoughtful collaboration with AI.

‍Thinking with AI demands clarity about where human reasoning ends and machine reasoning begins. As powerful as AI tools are, they simply don’t replace the clarity of human judgment. They amplify it. Our goal should be to design thoughtful collaborations, using AI’s ability to explore broadly and probabilistically, while preserving human oversight and deterministic decision-making where precision matters most. Those who understand and embrace this partnership will help define the future of AI.