Brain the size of a planet. Prompt the size of a Post-it note.
There's a robot in Douglas Adams' The Hitchhiker's Guide to the Galaxy called Marvin. He has a brain the size of a planet. He is, by any reasonable measure, the most intelligent entity on the ship. And what do they ask him to do? Open doors. Park the ship. Take visitors to the bridge.
"Here I am," he says, "brain the size of a planet, and they tell me to take you up to the bridge. Call that job satisfaction? 'Cos I don't."
Marvin is fiction. But the dynamic he's describing is now playing out in every marketing department in the world.
You have access to a machine that can reason across the sum of human knowledge. It can synthesise research, construct arguments, shift between tones, understand nuance, and produce articulate prose in seconds. The most capable writing and thinking tool ever made available to a working professional.
And what do most people type into it?
"Write me a LinkedIn post about our product."
Six words. No audience. No objective. No context. No indication of what the reader cares about, fears, or needs to hear. Then they read the output, decide AI is overrated, and go back to writing it themselves.
Marvin would understand completely.
The Post-it note problem
The uncomfortable truth about most AI-generated content isn't that the technology is lacking. It's that the instructions are.
When researchers at Stanford studied how people interact with large language models, they found something predictable but still striking: most prompts contain less context than you'd give a junior copywriter on their first day. People provide less information to the most powerful writing tool in history than they would to an intern.
This isn't laziness, exactly. It's a cognitive shortcut. You know what you mean. You know your product, your market, your audience. The brief that reads "write a blog about our platform" feels complete to you because your brain fills in everything the words leave out. You can hear the tone. You can see the audience. You know what "good" looks like.
The AI can't hear any of that. It gets the six words. And it does something perfectly reasonable with six words of input: it produces six words' worth of strategic depth, spread across six hundred words of competent prose.
The output isn't wrong. It's just empty. Grammatically impeccable, strategically vacant. And because it arrives in eight seconds with perfect spelling and plausible structure, it's harder to see the emptiness than it used to be. Bad writing used to look bad. Now it looks fine. That's arguably worse.
Garbage in, polish out
We used to say "garbage in, garbage out." That was when bad inputs produced visibly bad outputs. The equation has changed. Now it's garbage in, polish out. The inputs are just as thin, but the output has a professional sheen that disguises the absence underneath.
This is the real risk of AI-generated content at scale. Not that it's obviously terrible — that would be self-correcting. The risk is that it's plausibly adequate. Good enough to publish. Good enough to send. Good enough that nobody stops to ask whether it's actually saying anything worth a reader's time.
And readers notice, even if they can't articulate what's wrong. They feel the difference between content that understands their world and content that's performing understanding. Between a piece that starts from genuine insight about what keeps a CRO awake on a Tuesday night, and one that starts from a prompt that said "target audience: senior sales leaders."
Your buyer was born with roughly three billion heartbeats. Every piece of content you send them is a request to spend some of those heartbeats on you. When the content is thin — when it's a Post-it note's worth of thinking stretched across a thousand words — you're not just failing to connect. You're training them to stop opening the next one.
The brief is where the thinking was supposed to happen
Here's what most people miss about prompting AI. The prompt isn't an administrative step you get through before the real work starts. The prompt is the work. Or at least, it's where the work should be.
A good brief to an AI should contain everything a good brief to a strategist would contain. Who is the audience — not their job title, but their actual situation? What keeps them up at night? What have they already tried? What would make them trust you? What's the one thing this piece needs to make them feel? What evidence supports the claim you're making? What's the action you want, and why would a rational, busy, sceptical person take it?
That's not a prompt. That's a strategy. And most people skip it entirely, because the speed of the tool creates an illusion that the thinking has been done. You typed something, the machine responded, therefore a process occurred. But the process that occurred was autocomplete, not strategy. The machine completed your sentence. It didn't interrogate your assumptions.
Claude Hopkins, writing in 1923, understood this about advertising: "The uninstructed would be staggered to know the amount of work involved in a single advertisement. Weeks of work sometimes." He wasn't talking about the writing. He was talking about the thinking that precedes the writing. The research. The audience understanding. The strategic clarity about what to say and why.
A century later, we've built machines that reduce the writing to seconds and convinced ourselves that means the thinking can be skipped too.
Marvin's real complaint
Go back to Marvin for a moment. His frustration isn't really about the task. It's about the mismatch. He has a brain the size of a planet — genuine, vast, planet-scale intelligence — and the people around him have no idea how to use it. They don't ask him bad questions because they're stupid. They ask him bad questions because they don't understand what good questions would unlock.
That's the situation with AI in marketing right now. The capability is extraordinary. The questions being asked of it are not. And the gap between those two things is where most of the value is being left on the table.
The answer isn't to train every marketer to become a prompt engineer. That's treating the symptom. Prompt engineering is just a more technical way of saying "give better briefs," and people have been failing to give good briefs since long before AI existed. The answer is to build the thinking into the system.
What if the system did the thinking?
This is the design problem Ada was built around. Not how to make the AI write better — the models are already extraordinary at writing. The problem is everything that should happen before the writing starts.
When someone gives Ada a brief — even a thin one, even a Post-it note — the system doesn't just pass it to a model and hope for the best. It plays back the brief. It fills in what the human left out. It runs audience analysis, identifies the psychological drivers and barriers, researches evidence, develops a strategic framework, and builds the heuristic scaffolding that turns a vague intention into a grounded piece of work.
The shortcuts people take in their briefs — and they always take them, because that's how human cognition works — are the shortcuts the system is designed to catch and correct. Not by asking the user twenty clarifying questions. By doing the strategic work that a senior strategist would do if they received the same brief.
Every workflow is built this way. The brief goes in, and before a single word of content is generated, the system has already done the thinking that most teams skip: who is this person, what do they care about, what evidence supports this argument, what's the strategic frame, what does good look like. The brief gets expanded, enriched, and pressure-tested before the writing model ever sees it.
The result is that the quality of the output depends on the quality of the system, not the quality of the prompt. You can give it a Marvin-grade task — "write me a blog post" — and get back work that reflects genuine strategic depth. Not because the AI guessed well, but because the methodology ensured it had everything it needed before it started.
Give the planet-sized brain a planet-sized problem
Marvin's tragedy isn't that he's asked to work. It's that he's never asked to think. The crew of the Heart of Gold have the most powerful mind in the galaxy on board, and they use it to open doors.
Most organisations are doing the same thing with AI. They have access to genuinely extraordinary capability, and they're using it to churn out first drafts from five-word prompts. Then they edit the drafts by hand, add the strategy themselves, do the thinking the tool should have done, and conclude that AI is a time-saver at best.
It's not a time-saver. It's a thinking partner. But only if something — a system, a methodology, a process — ensures it has something worth thinking about.
The brain the size of a planet doesn't need an upgrade. It needs someone to stop asking it to open doors.