Are we thinking, or just asking

Richard Wright
2 days ago
5 min read

Richard Wright | Wright Thinking

There's a scene that plays out in offices everywhere right now, and it looks nothing like the danger it actually is. Someone opens a chat window, types a half-formed sentence — "write me a digital strategy for a luxury car brand" — and within seconds has three paragraphs of confident, well-structured, entirely plausible prose. They tidy the language, drop it into a deck, and present it as their thinking.

Nobody in the room can see the joins. That's exactly the problem.

We've Been Here Before

I've watched this industry get turned upside down more than once, and there's a pattern that keeps repeating: a powerful new tool arrives, the old guard panics that a whole generation is about to lose the ability to think for itself, and everyone argues past each other for a decade.

We had this exact fight about calculators in schools in the 1980s. Parents wrote anguished letters warning that children would stop understanding mathematics and simply trust whatever number the machine produced — that without the mental practice, nobody would even know if the answer was wrong. They weren't entirely wrong. But they weren't entirely right either.

Calculators didn't destroy mathematical thinking.

What they did was change what kind of thinking schools actually needed to teach. The tool was never the problem. Skipping the thought that should happen before you reach for the tool — that was always the problem. Generative AI is the same fight, at professional scale, with considerably higher stakes than an exam grade.

What The Research Is Actually Telling Us

I don't reach for academic studies lightly in these posts, but the evidence coming out over the last eighteen months is too consistent to ignore.

A 2025 study by Swiss researcher Michael Gerlich looked directly at what happens when people offload cognitive effort to AI tools. The pattern was stark: as reliance on AI increased, critical thinking measurably decreased — and younger, less experienced participants showed the steepest decline. The correlation between offloading and reduced critical thinking was strong enough that it's hard to write off as noise.

Then came the MIT Media Lab study — probably the most talked-about piece of research in this space all year. Participants wrote essays in three groups: one using ChatGPT, one using a traditional search engine, and one using nothing but their own head. EEG monitoring showed the brain-only group had the highest neural connectivity, especially in the regions tied to memory, concentration and problem-solving.

The ChatGPT group showed the weakest.

Worse, the effect didn't switch off when the study ended — participants who'd leaned on the tool for four months were still showing sluggish engagement afterwards.

Separate research into knowledge workers found the same dynamic in professional settings: the more people trusted AI output, the less mental effort they believed critical thinking required — and the more likely they were to accept an AI's solution even when it was visibly flawed.

None of this means the tool is bad. It means the thinking that used to happen automatically now has to happen deliberately.

The Trap Specifically

In immersive experiences, CGI and creative production, this trap has a very particular shape. A brief comes in. Someone under time pressure skips the bit where they interrogate the brand, the market position, the budget reality, the client's actual problem, and instead prompts for a strategy cold. What comes back reads well. It's grammatically flawless and structurally tidy. But it carries no institutional knowledge, no point of view, and no understanding of the client sat across the table.

And critically — if the person presenting it doesn't genuinely understand what it says, they can't defend a single line of it when someone in the room pushes back.

Visual quality has always been the non-negotiable floor in my studios. The same principle now applies to thinking quality. A strategy built on AI-default thinking looks like every other AI-default strategy — and the work that flows from it looks like everything else too.

In a craft that lives or dies on distinctiveness, that's not a small risk.

The Fix Isn't Less AI — It's More Thinking, Earlier

The quality of what comes out is a direct function of the quality of what goes in before you type a word. A prompt that actually earns a useful answer needs:

Context — who you are, who the client is, what the real goal is, what constraints actually exist
Requirements — what the output needs to do, and in what form
An invitation to be questioned — the best prompts ask the AI to interrogate you back before it starts. Almost nobody does this, and it's the single most useful habit you can build, because it forces you to confront what you haven't yet decided
Mandatories and guardrails — what must be in, what must never be in, what tone and register is non-negotiable
A reference point for quality — if you know what good looks like, show it

Version One Is a Draft, Not a Delivery

I'll be honest about my own process here: I regularly go through six, seven, eight versions of something before I'm happy to put my name to it. That's not inefficiency — that's the work. The gap between the first version and the eighth is the entire gap between generic and right.

AI is a genuinely brilliant collaborator.

It's fast, it's extraordinarily well-read, and it has absolutely no knowledge of your specific client, your specific market, or your specific point of view. If you don't supply that, it fills the gap with the average of everything it's ever seen — and average was never what any client paid us for.

Where This Leaves Us

I'm optimistic about this tool, the same way I was optimistic about real-time engines when everyone else thought Unreal belonged strictly to gaming. New capability, used well, has always moved this industry forward. But capability was never a substitute for judgement, and it still isn't.