Key Takeaways
- AI excels at speed and scale, but humans bring purpose, perspective, and critical judgment.
- “Thinking with AI” means prompting iteratively, reflecting deeply, and treating AI as a thinking partner—not just a shortcut.
- A structured mindset, like the FEVER loop, transforms AI use into a collaborative, creative process across education, research, and work.
Since 2023, I’ve been publicly advocating a shift in how we relate to generative AI—not just using it for output, but thinking with it. In my book AI Literacy for the Age of Large Language Models, I explained how iterative prompting is essential for critical analysis and high-quality results from LLMs. That philosophy shapes how I now research, write, and teach. Human writing has changed fundamentally. Increasingly, it will be the human who thinks—with the machine doing the writing, through guided iteration. This blog shares how to make that shift deliberate, mindful, and empowering.
Are you using AI as a tool—or are you thinking with it?
When a new idea strikes, it’s often cloudy—vague, half-formed, and hard to pin down. Human thinking starts messy. But according to the American Federation of Teachers, deep knowledge sharpens our ability to reason, solve problems, and think critically. Knowledge provides the structure that supports sharper insights.
Large Language Models (LLMs) are like hyper-knowledgeable colleagues who have read everything but still need guidance. They don’t know what matters to you until you ask. They respond only as well as you frame the problem. They don’t reason like humans—but when prompted clearly, they help surface deeper thinking. LLMs don’t replace thought; they reflect and expand it.
That’s why in 2025, the highest performers in education, research, and technical professions aren’t just using AI—they’re building thinking systems with it. They’re not outsourcing judgement or creativity but working with AI to enhance both—asking better questions, testing ideas, and sharpening outcomes through collaborative iteration.
This blog introduces a practical framework for “Thinking with AI”—a mindset, a process, and a set of adaptable habits that help you think more clearly, learn more deeply, and create more powerfully in collaboration with machines.
What Does It Mean to “Work with AI” vs. “Use AI as a Tool”?
Using AI as a tool often means asking for quick answers, summaries, or shortcuts. It’s transactional—you input a prompt, get a result, and move on.
Working with AI, on the other hand, is iterative and relational. It means:
- Treating AI like a thought partner, not a vending machine
- Engaging in dialogue: refining ideas, asking follow-ups, and testing boundaries
- Building workflows, prompt libraries, and habits that evolve over time
- Integrating AI into your thinking—not outsourcing thinking to it
💡 The difference is like asking a calculator for a result vs. asking a colleague to brainstorm strategy with you.
Working with AI is slower at first—but it leads to faster thinking, deeper insight, and higher quality outcomes in the long run.
Why This Framework Now?
In 2025, generative AI is no longer a novelty. It’s integrated into search, writing, coding, and learning platforms. But as LLMs become commonplace, the way we engage with them becomes our true edge.
- Without structure, AI just speeds up confusion.
- Without critical dialogue, AI reinforces surface thinking.
- Without reflection, AI risks deskilling rather than empowering.
That’s why we need a new framework—not just to use AI, but to grow with it.
💡 What are the benefits?
- Save time by reducing trial-and-error with AI
- Gain deeper insight by working with AI, not just through it
- Strengthen critical thinking and digital fluency
- Stay relevant in a world where AI fluency is a core skill
The Thinking with AI Framework: Core Components
This framework consists of three layered components:
🔁 1. The Mindset
Your starting point is not the tool, but your attitude. Thinking with AI begins by adopting a mindset of:
- Partnership – You and the AI contribute, test, and build on ideas together
- Clarity over control – You don’t need to master the tool, just steer it clearly
- Growth and curiosity – You treat AI as a space to test, play, and evolve your thinking
💡 Key shift: From prompting for answers to prompting for clarity.
🧩 2. The Process: FEVER – A Thinking Loop for AI Collaboration
To think with AI, we need more than a prompt—we need a process. This is where FEVER comes in: a simple, memorable loop to help you collaborate with AI across writing, learning, decision-making, or coding.
Let’s walk through the five steps using one running example:
Scenario: You’re a teacher preparing a short guide for students on how to avoid common exam mistakes in geography.
🔍 Step 1: Frame the Problem
Clarify what you want: the goal, the audience, and the kind of help you need from the AI.
Prompt: “I want to create a 300-word guide for Year 10 students about common exam mistakes in geography and how to avoid them. Use student-friendly language.”
💬 Step 2: Engage in Dialogue
Don’t settle for the first answer. Ask the AI questions, seek alternatives, and refine together.
Prompt: “That’s a good start. Can you rewrite the third point to make it more practical?”
Follow-up: “What’s one misconception students often have about climate graphs?”
🗣️ Step 3: Verbalise and Reuse
Capture effective phrasing, prompts, or formats. Turn these into reusable tools.
Example: Save the prompt: “Explain [concept] to [audience] using [format] and include a common mistake to avoid.”
Verbalisation tip: Articulate why an output works. “This intro works because it uses a relatable student scenario.”
🧪 Step 4: Evaluate the Outcome
Judge not the polish, but the clarity, relevance, and usefulness of the result.
Prompt: “Does this guide clearly explain the problem and give students something actionable?”
Follow-up: “Rewrite if needed to focus more on cause-effect relationships.”
🔁 Step 5: Reflect and Adapt
Think about what worked and what didn’t. Adjust your future approach.
Prompt: “What did I learn from this session? What would I try differently next time?”
Action: Add the refined prompt to your personal or team prompt library.
🧑🏫 I use this FEVER loop regularly with my PhD students while co-developing manuscripts. We often iterate several times, refining both the thinking and the output. Sometimes it takes longer—but the end result is far clearer, sharper, and more collaborative than either of us working alone. We also save our best prompts for reuse—creating a growing toolkit of insights over time. In fact, for some recurring writing tasks—like structuring a methods section or drafting a literature gap—we even build custom GPTs tailored to those goals. This helps ensure consistency, saves time, and allows for collaborative refinement across multiple projects.
📌 FEVER is iterative, flexible, and designed for building thinking habits over time—not just generating outputs.
🧠 3. The Practices
These are habits or rituals that embed the framework into real life. They’ve been adapted from research, expert behaviour, and high-performing AI users. You don’t need all nine—start with one and build.
🧱 1. Start with a Structure
Before prompting, define your goal, the audience, and the type of response you’re aiming for. For high-stakes or complex tasks, I often spend considerable time—sometimes hours or even days—thinking about how to frame the problem. This includes reviewing notes, gathering examples, and reflecting on what I really need from AI. Thoughtful setup reduces wasted cycles.
Prompt: “Help me plan a persuasive email to a busy colleague.”
🔁 2. Think Out Loud with AI
Once you’ve framed the task, treat your AI interaction like a conversation. Ask follow-ups, test assumptions, and refine gradually. You might start with one idea and pivot based on the AI’s feedback. Don’t be afraid to question the output—or your own assumptions.
Prompt: “What are some counterpoints I might be missing?”
📦 3. Save What Works
When something clicks—whether it’s a prompt, a phrase, or an output—capture it. I often keep a running document of prompts that worked particularly well, especially after long iterations. Over time, this becomes a personalised toolkit. In some cases, we even convert these into custom GPTs for recurring tasks.
Tip: Save one good prompt or output per week.
🧭 4. Learn Your Tools
Understand how AI works: hallucinations, limits, context windows. But also be aware that some of the limitations are nuanced. AI models tend to agree by default—so if you want challenge, you often need to invite it. Prompting is not just about wording; it’s about aligning the AI’s response with your intent. That means experimenting, adjusting, and gaining experience over time. Knowing how to phrase prompts to elicit disagreement, critique, or alternative views becomes a key skill.
Prompt: “Explain what context windows mean with examples.”
🎯 5. Optimise for What Matters
Don’t chase perfection—chase clarity, confidence, and usefulness.
Prompt: “Make this clearer for a first-year student.”
🧪 6. Run Weekly Experiments
Test one new AI workflow each week—small, specific, and measurable. Given how fast AI models and platforms evolve, it’s almost impossible to keep up with everything. But regular experimentation—paired with a curious mindset—helps you stay relevant and discover practical, personalised uses. Follow a trusted newsletter, YouTube channel, or blog (like this one!) and test what resonates.
Example: “Try summarising one research paper per week with AI.”
🔄 7. Build Flexible Systems
AI tools change rapidly—but the fundamentals of effective prompting and reasoning stay relevant. Focus on building workflows that aren’t locked to one version or product. Understand why a prompt works, so you can adapt across tools. For example, prompts developed on GPT-4 can often be repurposed in Claude or Gemini with slight tweaks. That’s also why I wrote my book AI Literacy for the Age of Large Language Models (2023) to focus on core principles that outlast tool updates. Despite being published “early” in AI terms, its foundation still holds.
Tip: Save prompts by category (e.g., writing, learning, decisions), and revisit older prompts to see if new models handle them better.
⚠️ 8. Challenge and Verify
Don’t blindly trust AI. Ask it to critique itself or find counterevidence.
Prompt: “What’s wrong with this answer? Give me opposing views.”
🤝 9. Share and Scale (for Teams)
Use AI collaboratively—share prompts, critique ideas, document wins.
Tip: Create a shared team doc with your top 5 prompts or workflows.
🎓 How This Framework Aligns with Various Levels of Thinking
Learning and thinking happen in layers. Bloom’s taxonomy is a widely used framework that categorises cognitive skills—from basic recall to complex creation.
What makes “Thinking with AI” powerful is how naturally it maps onto these layers. Whether you’re studying, writing, problem-solving, or teaching, this approach can help deepen your learning at each level.
| Bloom’s Level | AI Thinking Practice Example |
|---|---|
| Remember | Use AI to summarise content or define key terms |
| Understand | Ask AI to explain a concept in simple terms or use analogies |
| Apply | Prompt AI to create examples or draft applications of a theory |
| Analyse | Compare outputs, find gaps, or challenge reasoning with AI |
| Evaluate | Use reflection prompts to judge the clarity, logic, or usefulness of output |
| Create | Collaboratively generate new structures, ideas, or strategies with AI |
💡 This alignment makes the framework especially useful in educational, training, and research contexts.
Conclusion: Thinking with AI is a Process, Not a One-Time Skill
Like learning to research, write, or teach well—thinking with AI takes time and intention. You don’t master it in a day. But you do get better by practicing with purpose.
Start small. Try framing just one task differently this week. Ask AI to help you think, not just type. Then build from there.
Acknowledgement
This article was prepared with the assistance of AI tools for research, structuring, and image generation. Some links in this article may be affiliate links. There is no additional cost to you. A commission may be earned if you make a purchase through these links.
