How to use AI to think better and work more creatively

Executive overview

Most people treat AI as an output machine — something that writes things for you. The real leverage is using it as a thinking partner: for brainstorming, stress-testing assumptions, and seeing problems from angles you wouldn't reach alone.

AI doesn't replace expert judgment. It amplifies it — especially for non-technical workers who know their domain and can direct the tool with precision.

The quality of your output is determined by the quality of your prompt, not the capability of the model.

AI as a cognitive tool, not just an automation engine

• AI is a cognitive artifact — like a map, it can make you smarter if used actively, not just consumed passively
• GPS analogy: passive use creates dependency; active use builds capability
• AI will change not just how we work but how we think
• The question isn't "what can AI automate?" but "how can it improve products, services, and relationships?"
• Offloading tedious tasks (insurance notes, scheduling, data summarisation) frees humans for higher-value presence and judgment

AI and creativity

• Creativity depends on quantity: one Stanford study found you need ~2,000 ideas to reach one successful product
• Humans are socially inhibited in brainstorming; AI has no social stakes and generates without filtering
• Prompt for volume: ask for 50 or 500 ideas, then iterate — "less tofu, more of this"
• AI is a great mimic: give it style examples and it will adapt to tone, audience, and voice quickly
• AI surprises in both directions — it can generate creative leaps and fail at things that seem simple

Where to insert AI in your workflow

• AI is useful at multiple points: drafting, brainstorming, outlining, editing, and audience-testing
• Use it to surface blind spots: "Here's what I've written — what am I missing?"
• Role-play difficult conversations: upload performance reviews or emails, ask AI to play the person you're dealing with
• Stress-test plans: "Here's my six-month business plan — what could go wrong?"
• Test copy for specific audiences: "How will Oklahomans read this differently than Texans?"
• Analyse qualitative data: AI handles large volumes of narrative feedback that don't reduce to numbers
• AI can now read emotional dynamics in meeting transcripts — who interrupted whom, who seemed distressed

Prompting well

• Generic prompts get generic answers; specificity is everything
• Good prompts tend to be longer — include context, style examples, constraints, and process instructions
• Treat AI like a naive intern: explicit instructions matter; assume nothing is obvious
• Useful additions: "do this one step at a time", "if you can't find a real reference, skip it", "start over if these criteria aren't met"
• Odd but effective: framing, tone, and even unusual openers can shift response quality significantly
• Iterate: if the first three prompts disappoint, the prompt is probably underspecified

Getting started and building the habit

• Start with the best available models: Claude, ChatGPT, Gemini — the free tiers are significantly weaker than paid
• For evidence-based questions, use specialised tools (e.g. Consensus) that only draw from published research
• Use other people's prompt libraries to understand what's possible before writing your own
• 88% of AI productivity gains at work come from non-technical workers — this is not an IT problem
• Moderna gave access to all employees; 25% were using it daily in highly varied ways within months
• The right organisational culture: encourage individual experimentation and sharing of what works
• Don't block access over data concerns — people will use personal devices anyway, and that's where learning happens

On bias, judgment, and expert oversight

• AI was better than junior analysts at predicting quarterly earnings — partly because it had no emotional reaction to figures like Elon Musk
• AI bias exists, but it's easier to surface and correct than human bias: ask it "what contradictory evidence exists for my position?"
• AI is consistently better than novices across legal, medical, and financial domains; not yet better than experienced experts
• Everyone needs to become an expert — AI raises the floor, not the ceiling

Holding uncertainty well

• Avoid fixed pro/con positions on AI; the technology and its consequences change faster than any settled view
• AI will produce both clear benefits and real harms — social media is the model, not the exception
• The real risk is premature closure: dismissing a use case as bad before understanding how it could be good
• Daily recalibration is appropriate; certainty is a liability in a fast-moving landscape