How to think with AI without losing your mind
Can you still sharpen your brain with the temptation of AI?
I've been rethinking what it means to think with AI. And what it means to think alone with the knowledge that AI is in the wings.
It's a hot baked challenge facing professionals who earn their bread using their loaf: whether you're a PhD scientist shaping research papers, an economics analyst or an AI comms consultant.
If we don't put in the work of thinking and delegate it all to AI, will we become redundant in the robo-revolution?
Is AI turning your white-collar job grey?
Mechanisation and automation, in past technology waves, were associated with the demise of low-skilled work. Self-checkout tills (the devil's own retail enshittification) replaced cheery assistants.
But before then, technology replaced skilled workers. Most notably, the English Midlands' finest revolutionaries, the Luddites. They broke the weaving machines during the Industrial Revolution. Not, as history unkindly told, because they rejected the march of progress, but to fight for their livelihoods, hollowed out with the arrival of click-clack mechanised machines.
The Luddites were imprisoned, shot and executed. They ultimately didn't succeed in reversing mechanisation, but they did slow the tide.
Now, as fast as you can summarise that 200-page report, AI is coming for so-called 'knowledge workers'.
At first, they came for the bookkeepers.
I did not speak out as I was not a bookkeeper.
Then they came for the healthcare administrators.
And that poem isn't all that long, so we don't have time to stand on the sidelines wondering whether an AI chatbot will improve our IQ or not (clue: it won't).
Is AI brain rot really the threat?
Will AI deskill and demote us to voice-prompting aides, gently corralling an army of AI agents who do the deep thinking and analysis work?
Will leaning too hard on AI cause brain rot so we forget our education, identities and sense of self?
Will AI mean there's no need for 'workers' and a universal basic income from generous tech industry patrons gift us a life of leisure?
Quick answer on that last one: No.
Longer answer: Economist Keynes in 1930 predicted a life of leisure where rapid industrialisation meant that we would only work a 15-hour week. How’s that working out for you? With a gently shrinking population in many nations, there will be plenty of things for people to do.
Brain rot is real. A 2025 Microsoft and Carnegie Mellon University paper said AI is causing "the deterioration of cognitive faculties". Lean harder on AI tools, and you disengage with the information. A recent MIT research study pitted students with and without AI to write a SAT essay (university entry test). Those who leaned fully on AI hardly remembered a thing they wrote.
Professor Ethan Mollick says in his argument against AI causing brain rot:
"AI doesn't damage our brains, but unthinking use can damage our thinking. What's at stake isn't our neurons but our habits of mind."
We need to remind ourselves that thinking is a habit.
And unthinking is a bad habit.
The singularity isn't coming for us (yet)
AI agents are still silly robots. Experiments gallantly playing out in public by Anthropic show that managing the stock and sales of a simple vending machine is out of the grasp of today's AI agents. The automaton invented fictional people, personified itself and claimed to be wearing a garish blazer.
AI tools that lack guardrails, like Grok's attempts to solve world problems by channelling a certain 1930s German dictator, are frankly terrifying.
Why doesn't automation work yet? Us pesky humans. We're unpredictable. Klarna's failed AI-first sales team (rewind to real people), or McDonald's annoying and hackable AI recruiter prove that people are the most unpredictable variable in the data set.
We're unashamed outliers.
Going back to the origins of today's AI chatbots, they don't really 'understand'. They're drawn from natural language processing, guessing what is pleasing and likely based on prior patterns. More than a monkey with a typewriter, and a little more sophisticated now than the stochastic parrot previously claimed by AI researchers.
Perhaps a better analogy is a cat that speaks. It's intuitive, analytical, and has wily ways of luring you to bend to its will. But it's still just a cat that speaks. Like animated alley cat Top Cat (or Boss Cat for UK folks due to a cat food advertising dispute).
Train your brain to work with AI
If you're worried the thinking and interesting bit of your job will get eroded by AI, you're in a self-fulfilling spiral down.
During a stay in Greece recently, I was reminded daily that once you step out of the high-tech bubble, the pace of change in life and business is leisurely.
Most organisations are still working through their digital transformation. If you can get yours and your team's work (ideally the whole organisation) focused on where AI is heading, you'll be well-placed for the next wave.
Here’s some ways to adapt to get your mind working with AI.
Step 1: Educate yourself
First, consider AI literacy: working out what AI can and can't do for your job. Accept AI is coming for whatever repetitive or information-led tasks you sometimes do, and figure out how to collaborate with the tools to give you a little leg up or shortcut.
Reality check: it's not going to give 95% efficiency gains (silly AI hype, Mr Altman). It's incremental wins. Just some Microsoft Clippy-style inspiration like, "have you thought about doing it like this?" really helps if you work in a team of one, as it seems we increasingly do.
Step 2: Don't press the button
The magical homework button tempts our lazy ape brain on a sunny afternoon. But AI tools can limit us if we're at the crucial stage of learning something new.
AI tools are primed to spit out 'answers' and 'information'. However, when trying to understand something novel, such as how AI reviews brand content, published sources may lack authority or depth. Often, I need to do the hard graphite myself to learn by reading, writing and speaking.
Step 3: Define your value
Deepseek and OpenAI's deep research mode is a gear shift. It can spit out plausible research data in hours or minutes. Ladder in a RAG model (which pulls in data from your knowledge sources) and AI can do a lot of first-level analysis, partly why entry-level knowledge and research jobs are at serious risk.
But we will always need those who understand market specifics. Double down on niche verticals that aren't well-published. Provide a layer of specialist analysis or deep-domain knowledge. Talk to experts. Content creation must shift beyond basic 'how-to' information to expert-led insight.
Step 4: Stand out or camouflage?
If standing out is important, like shaping a brand in a competitive space (which space isn't), being the same is bad. Very bad.
You need daring creatives, writers and designers to play querdenker – the maverick that challenges the status quo. As I write, I see the term has been usurped by Germany's anti-vaxxers. Proof that context awareness is everything.
Don't strip out or replace those specialist creatives and subject experts, or, like Klarna, repent at leisure when your market share plummets.
If blending in suits, and you’re all about efficiency, then AI is a golden ticket to a smoother life. Focus on how you can track and measure tasks, celebrate the wins like an automated invoice pay and chase (happy solopreneur joy).
Step 5: Add a second brain
AI vendors are all vying for you to choose their platform as your 'second brain'. Why? Data is knowledge. And knowledge is platform lock-in.
It's why Google, custodians of your Gmail and search history since high school, can craft a better personal letter than ChatGPT. It's why Perplexity is stealing a march from Google with Comet, a search engine that makes sense of your 28 unread tabs and search history.
You will need to lean on your second brain. Choose wisely.
Work with the bots. Don't BE a bot.
ChatGPT-made business plans meet the sniff test, but interrogate how you would deliver it, and it won't taste right. Only you know the complexity of the irregular world you, your organisation and your customers inhabit.
Understand the value of your noggin. Smooth along the grain to surface richer insights faster, or rub against it and create work that stands out from the 50 shades of beige of AI-generated 'content'.
Snuggle that feline AI on your laptop. I've been back and forth with Claude on this article. It's given me some useful edits. And some bits, like this, I've put back in.
To be contrary. And human.
Interrogate your AI drafts like a spy.
Ask your bot: Why did you choose this over that? Try it the other way. It's not your friend. It's OK to be a mean girl.
Sad because it's true:
If the future is people producing words that no one has written, that no one reads, and are re-processed and regurgitated by AI, we’ll all soon be out of a job.
Think about AI as a conduit to do more of what you do well and love. (Still not many openings in the UK for theremin players, sadly).
Report back to your boss about what gains you've made. What did and didn't work with your nascent tests?
This should keep your skills in demand as the smartest, AI-enhanced human in the loop.