Old wine in a new bottle?
The promise sounded familiar: separate your content from its presentation. Give your content team a clean, structured environment. Decouple the back end from the front end.
We actually built our own company website on a headless CMS to test the idea. The result? A website. Just a website. A harder-to-build website, with more complexity under the hood, that looked and functioned exactly like what we could have built with a traditional CMS.
And composable architecture? Connecting two systems via an API and calling it revolutionary felt like a stretch. We had been doing best-of-breed for years: let the CMS do what a CMS does well, let the CRM do what a CRM does well, connect them where it makes sense. We just never called it composable.
My scepticism was not irrational. In 2018, the market was not yet mature enough to deliver on the promise. The technology was real, but the narrative around it was often oversold. Vendors were getting ahead of what the ecosystem could actually support.
The crack in the wall
The shift happened gradually, and it came from an unexpected direction: AI.
About three years ago I got involved in building conversational assistants, systems that use natural language, built around content and knowledge, without a predefined flow. And the first question that came up was: where does the content come from?
Suddenly headless made sense in a way it never had before. A chatbot or knowledge assistant is a head. Not a visual head, no webpage, no mobile app, but a head nonetheless. It consumes content in its purest form. Structured, clean, channel-agnostic content. Exactly what a headless CMS is designed to deliver.
And then something else clicked. That same assistant could sit inside a composable architecture and do more than just answer questions. It could suggest translations in the PIM. It could propose SEO tags inside the CMS. A capability added to the stack without rebuilding the foundation.
That was when I stopped seeing headless and composable as a technical preference and started seeing it as a strategic decision.
The moment I knew
I was preparing a talk about the future of customer experience, specifically about what it might look like to hand control to an AI agent entirely. Start with a blank canvas. Make the context, the content, the building blocks available. Let the agent decide how to assemble a webpage based on what it knows about the user.
During the preparation, the importance of the underlying architecture became impossible to ignore. This kind of experience is only possible if your content is structured and accessible. If your capabilities are modular and connectable. If your foundation is solid.
After the talk, a client asked: "This is obviously the future. We're not ready for this yet, are we?"
I looked at them and said: "Nobody is fully ready for this yet, including the agents themselves. But you have already built a very strong foundation with your headless and composable setup."
They were proud. They were sitting next to prospects. That moment was worth more than any slide deck.
Later, that same type of architecture enabled one of our clients to let their marketers generate draft blog posts based on existing CMS content, product data from the PIM, and additional context, all accessible via MCP connections. Not a future scenario. Happening now.
What I still tell clients honestly
Becoming a believer does not mean becoming an evangelist. There are things I stay critical about.
- 1
Headless is not automatically cheaper. This is one of the most persistent myths, partly fuelled by vendors who lead with lower licence costs at the entry level. Yes, you can start smaller. But you also have to build the building blocks: the design system, the component library, the content types. In my experience, the initial investment for a headless and composable platform is often higher than a traditional setup. The payoff comes later, when adding a new channel, a new market, a new capability goes faster and costs less because the foundation is already there.
- 2
"No vendor lock-in" is not a real argument. You sign a contract with a vendor. You are locked in by definition, because let's be honest, nobody switches CMS every year. What composable does give you is risk distribution. You are not dependent on a single vendor for everything. That is a real advantage, but it is different from the promise of total freedom.
- 3
Not every client is ready. Headless and composable require a different mindset. Content becomes a product. Authors work less visually. You need a strong design system and component library before things start to feel intuitive. Some clients need to make that mental shift first. And some clients simply do not need this level of architecture. A straightforward informational website without multichannel ambitions or AI integration is a perfectly valid situation. Pushing headless and composable on a client who is not ready, or does not need it, is not good advice. It is just selling.
Why it matters more now than ever
I am convinced that headless and composable will become increasingly decisive in the years ahead, not less.
LLM vendors are not going to keep stuffing new content into their models indefinitely. Those models get better because they can do more, not because they have ingested more data. Real-time content retrieval via MCP connections, via agents with specific expertise, is where this is heading. A company that has already built an MCP layer on top of a headless and composable platform has a meaningful head start.
More importantly: people are increasingly asking instead of searching. As a brand, you need to think about how your content can be served in fundamentally new ways. Not just a chatbot. Not just a chat interface. A new kind of UI, one where an AI agent gets a blank canvas, receives the building blocks, understands the user's context, and decides how to construct the answer.
That is not science fiction. It is the logical next step for anyone who has already built the right foundation.
The question that stays with me
A client once asked me: "Why would we still build a website if AI is becoming so important?" We had a long conversation. I explained our vision. We got the project, because after that conversation, they understood that the right architecture was the way forward, not a way to avoid the future.
That is the moment I feel the difference. Not when I explain the technology. When I see the click happen. When someone goes from "I don't see why" to "now I get it."
That is the journey I made myself, starting in 2018. I am glad I kept my eyes open.