How to use AI as a strategic thinking companion, not a search tool

Executive overview

Most people use AI like a search engine — type a query, get an answer, walk away. That's the wrong model. The brain literally pattern-matches a chat interface to Google, triggering the same command-response behaviour.

The shift is treating AI as a companion: something you have long, iterative conversations with that knows your goals, your context, and your preferences. When it does, every output is targeted — not generic.

Conor Grennan (Chief AI Architect, NYU Stern; founder of AI Mindset) argues this is a behavioural problem, not a knowledge problem. The fix isn't learning more tools — it's changing how you engage with the ones you already have.

The biggest unlock is not a better prompt — it's a longer conversation.

Why people stay stuck at surface-level AI use

• "Who's using AI?" gets every hand up. "Who uses 40+ formulas in Excel?" gets three.
• The same gap exists with AI: everyone's "using it," almost nobody has made it change how they work.
• Power users engage 30+ times a day, across home and work, in extended conversations.
• FOMO about new tools and prompt libraries is a distraction — the underlying models are good enough already.
• Neural pathway automation makes your brain default to search-engine behaviour; you have to consciously override it.

How to actually engage with AI

• Start by telling it your strategic goals, constraints, and preferences — every answer then targets those.
• Ask it to push back: "poke holes in my argument," "what am I missing?" produces more value than asking for answers.
• Write your first draft yourself, then hand it to AI: "make this more readable" or "what's wrong with this?"
• Use it like a co-CEO — someone brilliant who doesn't know your company yet but will refine outputs with you through iteration.
• Transfer memory between models by asking your primary LLM to produce a 10-page document of everything it knows about you, then paste that into another tool.

Choosing tools and managing accuracy

• Don't optimise for the best tool — optimise for the one that feels natural for the task.
• Claude (Opus): strong for writing. Gemini: deep research and Google integration. ChatGPT: everyday go-to, strong memory. Copilot: enterprise integrations.
• AI is a process machine, not an answer machine. Break any decision into steps; AI helps at each step.
• For hallucinations: treat it like a smart friend, not a calculator. If you need precision, verify at the source — don't cross-check with another AI.
• Accuracy matters proportionally: brainstorming tolerates errors; anything consequential needs verification.

Using AI to get hired and stand out internally

• Saying "I know AI" in an interview is table stakes — everyone claims it.
• The differentiator: walk in with a worked example of how you'd reinvent a process in that specific role, using AI.
• The kicker: show how the whole team could use that new process — not just you.
• Reinventing a process gets noticed far more than just doing more work.
• Inside any organisation, become the person who makes AI workflows repeatable and transferable — that's the IP.
• Think of yourself as a "Chief AI Architect" of your team or function.

The job market and entry-level risk

• Dario Amodei (Anthropic): even with no further AI advances, the technology could wipe out 25% of entry-level white-collar jobs.
• A team of 10 can realistically be reduced to 2 — not because AI replaces everyone, but because one person can do far more.
• The structural problem: entry-level roles are how people learn to do the work that qualifies them for senior roles. Eliminating those rungs is unsolved.
• Domain expertise still matters: knowing what quality looks like is what lets you steer AI to extraordinary output.
• Companies have been slow to change hiring processes — the window to stand out is open now.

Entrepreneurship and process-as-IP

• AI removes the traditional barriers to entrepreneurship: you no longer need to know how to code, market, or run operations.
• The opportunity: use AI to invent a repeatable process in a domain you understand, then sell that process.
• "Parallel resume" approach: keep your job, but build something alongside it — even one workshop or client validates the model.
• Inside a company: propose to eliminate two of your team's five headcount by redesigning the workflow. That gets attention.
• The holy grail is a process that's agnostic to who runs it — repeatable by anyone, not just you.

AI in education and emerging markets

• Universities aren't going away: they provide human development, not just information transfer. The middle layer (undifferentiated online courses) is at risk.
• AI as a personalised tutor can reach the lowest common denominator of any classroom and still tailor to each student — something teachers structurally cannot do.
• In low-income countries with no access to tutors or doctors, mobile AI access is transformative: immediate medical triage, personalised education.
• Critical thinking is at risk if AI is used only to get answers — the same risk Cliffs Notes always posed. Used to go deeper, it can enhance critical thinking.