The Model is not the Moat!

It's been a fun last few weeks. Working with the team, clients, partners and more on new ways of working.

Part of it was in London, with amazing people who have recently joined us. They understand media planning, and they understand what a client actually needs. That kind of understanding takes years to build. Decades, even.

Part of it was around the country with clients working on their specific use cases and needs. What great looks like with

Part of it with my technology and ops team in Spain, in Barcelona and Valencia, whose grasp of AI, technology and operations is as good as I have worked with anywhere.

And part of it was in Greece, on Kefalonia, with Amplifyd , our partners and now our clients. Whatever you want to know about search, social or direct response, they know it.

Good people. Real talent. Plus some sunshine and stunning locations.

But what made it one of the best months of my career was how it all fit together. Different talents, codified and magnified by AI.

Start with the model. There is real value in it. A PhD in every subject, in your pocket. But you can swap it for a different one next month and barely feel the difference. So the value does not really live there. We have spent a while as an industry circling where it does live. It is no longer hard to see.

Look at what the people closest to the models are doing. OpenAI's most telling recent move was not a new model. It was launching a deployment company, backed by billions, and buying an AI consultancy, Tomoro, with around 150 engineers, to staff it. Those engineers do not build models. They sit inside businesses and make the models actually work. The same logic is, in part, why Publicis has just agreed to pay 2.2 billion dollars for LiveRamp.

You can see it lower down too, in the run of agent tools arriving right now. Codex and Cowork from the labs. OpenClaw and Hermes, which I have built myself in my spare time. Every one of them is a layer wrapped around a model, not a model. The model is becoming the commodity. The layer around it is where the value sits. Being able to build that layer is a vital capability now, and still a rare one.

For a business, that layer has a clearer name. It is a brain. The value is in connecting great people and ideas to a shared one. A second brain for a person, an institutional brain for a company. An AI that holds the context of your business, and the institutional knowledge your best people spent careers building. A room of the right people in the right place produces something none of them would alone. A team wired into a brain like that does the same.

For that brain to deliver, a few things have to be true of it.

It has to be owned. The brain is core IP. It is the codified memory of how your business actually works, and it should belong to you the way your brand and your customer list do. You can rent the build. You cannot rent the outcome. Our partners in Greece have their own, earned the hard way. We are building others, owned outright, for top clients in gambling, travel and retail. A brain you rent walks out of the door when the contract ends, and takes everything it learned about you with it.

It has to be portable. No lock-in, to anyone. Not an agency, not a model provider, not a CRM, not a software house. If your brain only runs inside someone else's product, you do not really own it. You are holding it on loan.

It cannot stay walled off. Your marketing brain will not be the only one. In time there is a finance brain, a product brain, an HR brain. The marketing one has to be built so it can connect to them, because the best decisions sit across those lines, not inside any one of them.

It has to be used, and not only by the experts. Used in some way by everyone in the department, because that is what makes a team stronger. Each person gets the part of the brain relevant to them, shaped to how they work. Everyone's wisdom deserves a way to feed in. Everyone's intent deserves an engine behind it. A brain only a few people touch is one that barely learns.

And it needs layers. The core, the data and rules you have to get right, changes slowly and under proper governance. The edge, where the everyday learning lives, things like user transcripts, updates fast. The layers are not all the same, and should not be governed as if they were.

Where does all of this live? That is for each business to decide. A database. One of the newer tools like Notion. Markdown files in a GitHub repo. Google Drive. Obsidian, for the personal brains. The home matters far less than the principle. It is yours, and it is portable enough that it never gets trapped inside someone else's product.

So how do humans and AI split the work? A recent MIT study put numbers on it. AI is good at tasks and much weaker at whole jobs. What it does well is the easy part. It holds more context than any human brain can, and with the right skills it knows how to do the work. We have built those skills into our system, and the gap to what a person manages alone is embarrassing. I have met brilliant media planners, but nobody is equally sharp across every channel in every market. An AI with the right context can be.

The hard part stays human. The person brings the intent, the responsibility for whether the work lands, and the wisdom to know whether what the AI just said is true for the business, or just very well marketed. I have gone deep on that division before, so I will leave it there.

Codifying our own brain at TAU, and then helping clients grow theirs, is the most rewarding work I have done. I get to spend the next stretch of my career on it. And far from being a threat to the people who do this work, learning how to build it is the opposite. It is liberating.

The model is not the differentiator. Everyone can buy the same one, and before long everyone will. What you build around it is the difference. And that has to be yours.