Over the past few weeks, I interviewed with several AI marketing startups. One conversation in particular stayed with me, not because of the interview itself, but because of a product philosophy that I couldn't stop thinking about afterward.
Many AI marketing products today seem to share the same underlying assumption: marketing is fundamentally a knowledge problem. If we can collect enough best practices, document enough workflows, write comprehensive playbooks, and feed them into large language models, then marketing can gradually become autonomous.
At first, I couldn't explain why that assumption felt incomplete. The more I thought about it, the more I realized that my discomfort had very little to do with AI itself. It had everything to do with how we think expertise is formed.
Throughout my career, I've written plenty of documentation myself, including playbooks, experiment frameworks, measurement methodologies, campaign guidelines, and operational processes. I believe documentation is incredibly valuable. It preserves organizational knowledge, aligns teams, and gives people a common language to work with. Good documentation makes organizations more scalable.
What documentation does not do, however, is turn someone into an expert.
A marketer can read an entire playbook on experimentation without knowing when an experiment is worth running. They can memorize every platform best practice and still optimize toward the wrong business objective. They can understand the framework perfectly and struggle the moment reality no longer resembles the examples in the documentation.
The gap isn't intelligence. It's judgment.
Over time, I've come to believe that marketing has always been both science and art. The scientific side is increasingly machine-friendly. Attribution, measurement, experimentation, budget optimization, forecasting, and automation all operate on structured data with relatively clear objectives. I have little doubt that AI will continue transforming this part of marketing.
The artistic side is different.
Taste, context, timing, consumer psychology, and business judgment don't exist as isolated facts waiting to be written down. They emerge through years of interacting with an environment that is constantly changing. Competitors launch new products. Consumer behavior shifts. Auctions fluctuate. Creative fatigue appears. Economic conditions change. Every decision is made within a moving system rather than a static one.
This is why I've started questioning an assumption that seems increasingly common in the AI era: that sufficiently detailed language can fully capture expertise.
Recently, I've been reading perspectives from researchers working on world models, many of whom argue that intelligence cannot arise from language alone. Human understanding develops through continuous interaction with the world, not simply through processing text. Whether or not that ultimately proves to be the path toward more capable AI is still an open question. I'm certainly not an AI researcher, and I don't pretend to know which technical direction will win.
What I do find fascinating is that this debate reminded me of something much older.
There is a Chinese expression that says, 只可意会，不可言传
, meaning that some things can be understood through experience but cannot be fully expressed in words.
The more I think about marketing, the more I feel this captures something important.
Playbooks can explain what usually works. They struggle to explain why it works in one situation but fails in another. They can document knowledge, but they cannot fully transfer judgment. Language can describe experience remarkably well, but describing experience is not the same as having experience.
Ironically, I didn't arrive at this conclusion by studying AI.
I arrived at it through therapy.
Over the past several months, I've been working through trauma, and I noticed something unexpected. Every time I processed another old emotional pattern, new ideas seemed to emerge almost effortlessly. They weren't ideas I had consciously tried to invent. Instead, they felt as though they had always been there, waiting for enough mental space to connect.
That experience changed the way I think about creativity.
For a long time, I assumed creativity came from learning more, including reading more books, collecting more knowledge, building more frameworks. Now I'm beginning to wonder whether creativity is less about accumulation and more about integration. Perhaps our best ideas don't always come from adding something new. Sometimes they appear only after we remove what has been occupying our attention all along.
In an unexpected way, this has also changed how I think about AI.
Maybe the future of AI marketing isn't simply about building larger playbooks or writing better prompts. Maybe the more interesting question is which parts of expertise can truly be encoded into language, and which parts only emerge through interaction, adaptation, and lived experience.
I don't know the answer.
Perhaps future AI systems will narrow that gap. Perhaps entirely new paradigms will emerge beyond today's language models.
But for now, I find myself returning to a much simpler idea.
Knowledge can be documented. Experience has to be lived.
Judgment, whether in marketing, creativity, or life, may always begin where language reaches its limits.