Composable architecture without content discipline is just expensive flexibility

I have sat in a lot of transformation kick-offs. The pattern is usually the same. The agency presents the target architecture: headless, composable, MACH-aligned. The client nods. The platform decision has been made, the engineering team is ready, the timeline is ambitious. Discovery starts next week.

Nobody has yet asked what happens to the content.

Not in the sense of who writes it. In the sense of: how is it structured, how does it reuse across channels, how does the editorial team maintain it after the agency exits, and how does it need to be shaped for the AI applications already being built toward.

Composable architecture without structured content discipline is just expensive flexibility.

MACH gives you the ability to assemble and reassemble content across channels and surfaces. That is genuinely transformative. But the ability to do something and the content being in the right shape to do it are two different things. A composable platform with page-shaped content is still page-shaped content. Faster to deploy, more expensive to maintain, and no more AI-ready than what it replaced.

The failure mode I see most often: content models designed to serve the visual design rather than the other way around. A designer produces page templates. Each template becomes a content type. Each variation becomes a separate variant rather than a shared component. It looks clean in Figma. It looks reasonable in the first sprint review.

By the third sub-brand, the content model has tripled in size. The editor experience is a dropdown of near-identical options nobody can distinguish. Personalisation has nothing reliable to attach to. The migration that was scoped for eight weeks is now twelve because every piece of content has to be manually matched to a variant that almost but does not quite fit its source.

And when the client asks whether their content estate is AI-ready, the honest answer is not yet. Because AI applications consuming structured content need reliable types, clean taxonomies, and consistent field-level data. What they have is page templates dressed up as a content model.

The upstream content strategy work, discovery synthesis tied to business objectives, content modelling from the editorial side, taxonomy designed for findability and AI-readiness, governance built for the real team, migration with editorial validation as the deployment gate, is the layer that makes composable architecture actually deliver. It does not replace platform engineering. It is what platform engineering depends on.

For agencies: this is the next layer of value that clients are starting to ask for. For in-house teams: it is the layer most likely to be under-scoped and deferred to a later phase. Later usually means never, or an expensive remediation after go-live.

Both sides of the conversation are asking the same question in different ways. The answer to both is the same: structured content discipline, applied upstream, before the platform is built.

Want the full version? The complete piece covers what the upstream content strategy work actually involves across five disciplines, with the AI-readiness argument in full.