Writing on what it actually takes.

The most uncomfortable truth in ResearchOps right now — and a challenge to our own community.

Most design system post-mortems blame the library. Wrong diagnosis. Engineering doesn't resist well-designed systems — it resists systems where the token contract keeps changing, where a designer can push a breaking change without review, where nothing maps to their platform's release cadence.

That's not a craft problem. It's a trust problem. And at this stage of generative tooling adoption, it has stopped being academic.

Evidence across REA, Atlassian, and Bupa — three orgs at three maturity levels — on what engineering co-ownership actually looks like, and what AI has exposed about the design systems that were never really systems.

DesignOps came out of 2018–2020 as an answer to a specific problem — take the operational admin off designers so they could do the work. Six years later that framing has stopped being useful.

The operational scaffolding layer is the part generative tooling absorbs first. What survives is governance, capability development, and platform-grade measurement — work that a service team can't do.

On the five classic pillars, what's been reshaped, what's been exposed, and why experience debt belongs on the same register as engineering debt.

If design isn't embedded in how your organisation prioritises work — RICE scoring, OKR setting, investment planning — it will always be overridden by product and technology pressure.

Not because the work isn't good. Because it was never part of the decision.

The organisations that get design right don't just hire better designers. They change where design sits in the prioritisation conversation — and what evidence it brings to the table.