AI is No Longer About the Model – It’s About the Car Around the Engine

I recently attended a partner event where the conversations around AI had a completely different tone than just a year ago. It's no longer a discussion about which language model is best. It's a discussion about what you build around the model. And the more I've thought about it since, the more convinced I am that this is where the actual competitive advantage will lie in the coming years.

Models Are No Longer the Important Part

For the past two years, the AI market has focused almost exclusively on language models. GPT, Claude, Gemini. Who has the most parameters? The best benchmark? That was the dominating question.

But something has changed. Those working most with AI in practice are reaching the same conclusion: models are becoming interchangeable. They are good. They are rapidly getting better. And they are rapidly becoming more like each other.

This isn't a theoretical argument. According to IDC, 88 percent of enterprise AI projects never reach production – and the reason is almost never that the model is too weak. It's that the layer around the model – integrations, governance, context, security – isn't in place. That's where projects die.

One of the speakers at the event used a metaphor that stuck with me. He likened language models to engines in a car. Powerful, yes. But fundamentally comparable components. The actual value, he said, lies not in the engine but in the car around the engine. He called it "the harness."

What Is a "Harness"?

It's everything that surrounds an AI model when it's actually supposed to do something useful in a business: workflows, tools the agent can use, memory of previous work, access to organizational data, brand voice, approval flows, governance, and security. The entire context that turns an AI model from an intelligent chat function into something that actually creates value day after day.

That's why a good harness is more important than a good model. The model becomes an interchangeable component. The harness becomes the platform.

This isn't just a metaphor that suited a speaker on a stage. The concept of "harness engineering" has emerged over the past year as its own discipline within AI engineering, and is now described by several leading AI teams as the most important skill in the agentic AI era.

That's Where the Organizational Memory Lives

This changes where competitive advantage is built over time.

A model can be replaced. When a better one is released, you just swap the component, keep the harness, and immediately get a better AI without rebuilding the rest. But the harness is something else. It's where the organization's collective intelligence is stored: what has worked before, how the brand actually sounds, what approval flows exist, what data should be consulted before a decision is made. It is built up over time and cannot be easily copied.

One of the speakers phrased it like this: "Harness value compounds over time. Models reset." That sentence says something important about how competitive advantage is built in the AI era. Not in the component, but in the context. Not in what everyone can buy, but in what only you have built up.

What It Means for the Marketing Organization

When we talk about "AI for the enterprise," it is often the marketing organization that becomes the first real user. That's where content is created, campaigns are run, performance is measured, and the brand is delivered. Workflows that can be dramatically accelerated by AI—but only if the AI is embedded in the actual work, not alongside it.

What I heard between the lines at the event was a vision of a future where the role of the marketing department changes. The classic model where marketing orders from IT and optimizes at the margins is breaking down – not because it's bad, but because it's too slow for what AI makes possible. BCG describes it as a shift where the CMO becomes a "chief growth architect" – responsible for the entire orchestrated workflow where humans and AI agents collaborate toward business goals.

In the digital platform of the future, the CMS is no longer the center. It's still important, but it's the AI layer that becomes the new center—where workflows are orchestrated, where content is created and optimized, where decisions are prepared. CMS becomes one of several components the AI layer works with, not the driving mechanism.

This changes who needs to be involved in the platform discussion. It's no longer enough for technology and IT to decide. The marketing organization also needs to be at the table – because it's their workflows the AI should support, and their context that should be built into the harness.

My Key Takeaways

The most important investment in AI is no longer which model you choose. It's how you build the context around it. And since the value in the harness accumulates over time, companies that start early gain a structural advantage that will be difficult to catch up with.

That's why "starting now" is meaningful even when you don't immediately see the ROI. You are building a headstart that wins in the long run.

A concrete example of how this is taking shape in practice – and one of the reasons I personally became more convinced after the event – is what Optimizely is doing with Opal. It's easy to see it as "AI in the CMS," but what's interesting from a harness perspective is what they've built under the surface: skills (equivalent to the agent's tools), organizational memory, instructions that guide the agent based on the company's context, governance layers. And what really ties it to the harness thesis is the contextual awareness – the agent is aware of previous articles, what visitors are looking at, what they appreciate. That's where organizational memory actually becomes useful in a decision, not just stored somewhere. It's worth looking at regardless of whether you have the Optimizely portfolio today or not, because it shows concretely where enterprise platforms are headed.

We work with these questions in our customer deliveries daily, and it's clear that more CMOs and digital decision-makers are facing the same type of strategic choices. Feel free to reach out if you'd like to discuss further.