Jaybird DXP — the tool that shipped every other tool
An internal headless content platform is usually the force multiplier nobody writes a case study about. Here's how ours came to be.
Most of what I shipped through the Jaybird Group years lived in public — client websites, broadcast smart-TV apps, consumer streaming platforms. One of the things I’m proudest of, though, was something the clients never saw directly. Internally we called it the DXP. It was the tool that shipped every other tool.
The name was a stretch. “Digital Experience Platform” is the kind of phrase you find in the glossy pages of a sales deck. What the DXP actually was, under the hood, was a headless content management and automation system that let Jaybird spin up, manage, and maintain client sites faster than anyone paying us for those sites would have believed was possible.
Why we built it
You could, in theory, run an agency of our shape on off-the-shelf tools. WordPress for some clients. Contentful or Sanity for the picky ones. Custom one-offs for the rest. I’ve watched agencies try that path. It ends the same way every time:
- Two or three content models per vendor, none of them identical
- Per-client pipelines that break whenever a dependency updates
- A pile of admin UIs clients have to learn separately
- Support burden that scales linearly with client count
The DXP was the decision to stop doing that. Instead of juggling five CMS philosophies, we made our own one and put every client on it.
What was actually in it
Under the DXP name, I ended up building or co-owning a collection of pieces that fit together more like a small operating system than a CMS:
- A content model engine. Node.js + MongoDB.
Schemas defined in code, migrated through
migrate-mongo, versioned, rollback-safe. Every client’s content lived in the same engine with their own schema. - An admin interface. Clients and Jaybird staff used the same web app to manage their content — brand assets, copy, product catalogs, video metadata, whatever the site needed.
- A delivery API. Every client website, mobile app, and smart-TV app pulled content from DXP over a consistent REST surface. One learning curve, many front-ends.
- Automation lambdas. A growing set of AWS Lambdas wired into the content graph: data-to-PDF rendering for specsheet generation, contact-form normalization, webhook ingestion, scheduled content pushes. The system did things on its own, not just when asked.
- Infrastructure as code. CloudFormation templates for staging and production, so a new DXP environment could be stood up in a morning.
- A CI/CD pipeline. Bitbucket Pipelines, Docker images, ECS deployment. Shipping a DXP update was a single commit.
None of these pieces were exotic. The trick was that they were all the same pieces, shared across clients, improved every week, and owned end-to-end by a small team.
The leverage, in numbers
I don’t have exact counts to share. What I can tell you is the direction of the numbers:
- New client onboarding went from weeks to days.
- Fixes that touched content modeling landed in every client at once instead of being forgotten on older projects.
- Client-side training shrank to one conversation. Every client learned the same admin, so documentation compounded.
- Jaybird could take on more projects with the same headcount, because the per-project setup cost had collapsed.
Why this essay lives on the Cybind site
Because it’s the single clearest example I can give of what I actually do when I’m allowed to invest in the right thing.
The DXP wasn’t glamorous. No one was going to blog about “how we built the headless CMS.” But it’s what made every other Jaybird thing possible — Ken’s monorepo, WFMZ’s CMS, Locast’s ops tooling, all the less famous client sites that shipped on time because the boring infrastructure wasn’t the hard part.
Cybind is built on the same reflex. Most of the interesting things one senior engineer can ship alone in 2026, with modern AI as leverage, are only interesting because the tools underneath them were built with care five years earlier. If I do the Cybind job right, most of what I actually ship for the next decade will be flashy products standing on top of unflashy platforms. That’s the shape of engineering that compounds.