Input Brain vs Output Brain
Ever notice the words people use when they brag about their AI usage?
“I entered the right set of prompts and boom I just wrote ten blog posts, howboutthat?”
“Watch me enter in some made up ICP data and vibe code a dozen software tools that will make me four million dollars in ARR. Boom.”
In other words, “Here is what I produced. And here is how fast I produced it.”
For things like blog posts, social media posts, videos, reports, designs, and vibe-coded apps — the cost of creation is down to zero and the time commitment is as fast as you can prompt. This framing of AI as an output machine, understandably, has rubbed most thoughtful professionals the wrong way.
Okay, I won’t speak for you, but it rubs me the wrong way. These people have a magic thinking machine and they use it to make mistakes, but faster. And I’ve spent the past few months scoffing at people who frame this life-changing technology in such shallow terms.
Amateurs are focused on the outputs. But creative professionals in the AI age need to be focused on the inputs. The creation of the thing is secondary to what the thing is drawing from.
Inputs over outputs
It’s understandable how this happens. The people making these “workflows” are often the people who have never been responsible for their creation. I’m not a programmer, but I know my vibe coded one-off tool is not as robust or considered as a veteran software engineer would make. The people “one shotting” marketing have often never successfully marketed anything in their life.
Marketing is not “producing assets” any more than cooking is pulling things out of an oven.
Someone who, say, writes a blog post, doesn’t just write a blog post. They think deeply about what they want to say. They explore the idea by writing and deleting ideas. They talk to other people to get feedback and flesh out their ideas. They think through second and third order effects of their ideas and put their thumb on the rhetorical scales to ensure the important points are remembered.
They’ve only ever seen the output, so that’s what they focus on.
We can’t be precious about “craft”
Ok, so focusing on outputs is bad. So then what? Successful creative and marketing leaders in the AI age focus more on getting things like first-party data and less on kicking and screaming when someone uses AI to write a blog post.
At first glance, this may be disheartening. “What do you mean my writing or video abilities mean nothing in the AI age?” It’s not that the craft of the output does not matter. It does. But it does not matter more than the inputs. The ideas matter more than the execution because the execution is the price of a GPT subscription.
This is not some departure from the pre-AI world. There are novelists whose prose is forgettable but storytelling is captivating and they sell lots of books. There are journalists who are experts at finding scoops but the writing is secondary. We’ve long rewarded the premise and the idea before the execution.
Note the “before” and not “instead of.” It’s an order of operations thing. Get the premise and the idea right first and you can iterate on the execution.
To thrive in the AI platform shift is to ignore the endless calls to generate more and more and instead spend time honing your inputs first. Put aside your “output brain” and focus on the inputs.
Here’s what that looks like:
| Output brain | Input brain |
|---|---|
| Spend time producing volume | Spend time defining what good looks like |
| Train agents on general best practices | Train agents on what is uniquely you |
| Measure what’s made | Measure its impact |
| I did this faster than you | I did this more effectively than you |
| Research is a waste of time | Research is my edge |
| The benefits of my work decay over time | The benefits of my work compound over time |
It’s why for my clients, I don’t touch AI until we’ve established things like the ideal sales motion, unique POV, category defining terms, and available unique data sets. You need to buy quality ingredients to make a good dish. Or, more precisely, you need to use ingredients only you could ever have access to.
Any AI workflow I make has at minimum:
- A large corpus of examples to emulate. This is often successful or ideal things that the client has produced previously, but in some cases it can be other or similar works to model on.
- A tone and style guide. One of the best outcomes of AI is that I no longer have to yell at mid-level creative team members to use the style guide. It’s baked into their tools and keeps everyone using the same language in the same way.
- Principles for success. Every channel and every company has these. These are the things that a long-time creative manager starts to intuit about the audience. It needs to feel like X. It can’t ever make the reader think of Y. They love it when we address Z. I gather these largely through feedback from the team and, eventually, audience.
- Results from previous outputs. The workflow or agent should learn from itself. This could be edits from the creative team or analytics from a given publishing platform (like, say, open rates). These are feedback into the “principles for success” layer.
- Scheduled human reviews. Always have a human review your workflows and occasionally blow them up, otherwise you’re at risk of a form of platform or audience capture. Additionally, the models and their connectors improve quickly.
This is the bare minimum. Other worthy inputs include transcripts with customers, sales enablement, product usage data, survey data, and relevant third-party data.
A team that gathers this will outperform a team that just pulls some Claude Skill off the shelf or uses a vanilla AI agent.
Why the input brain works
This takes a bit of a steely resolve as you’ll look around and see people “shipping.” But what they are shipping is often not tied to a larger premise and is an amalgamation of the mediocre middle scraped by their AI.
The reason for this is the tinker’s dilemma: it can feel more satisfying to establish some whiz-bang AI workflow than to do the foundational input work. Founders with an engineering background have long run into a version of this, where they spend time with their heads down in Figma files versus doing the hard thing that actually finds product market fit: talking to customers.
Marketing is extremely difficult and requires grappling with some bracing truths about your company. What sets us apart? Why do people buy from us? Is our product even any good? What is the best way to talk about this product? How can we set up a system that compounds over time rather than just producing random acts of marketing?
Focusing on your inputs forces you to grapple with these questions. And when you do, the long term health of your company will be greatly improved.
If you want to play business, focus on outputs. If you want to actually run a business, focus on the inputs.