blog / privilio

blog

Privilio: what changed when we rebuilt our delivery around agents

For fifteen years we built WordPress projects the same way: senior engineers, by hand, in sequence. We got good at it — a hundred-plus projects good. Which is exactly what made Privilio interesting: when you've done something one way a hundred times, you have a real baseline to measure change against.

The setup

Privilio (formerly Rehab Care) — a private physiotherapy clinic and physiotherapist education centre in Košice, Slovakia, rebranding under a new name and domain. Brief: migrate a legacy Joomla site to WordPress under the new brand — without losing a decade of SEO equity or a 4.9-star, 380-review Google reputation. Hard requirement from the owner: every piece of content editable by her team in WP admin, no developer needed. Constraint that shaped everything: a ~62,000-character technical SEO specification by external SEO consultant Pavel Ungr — custom structured data, a multi-sitemap architecture, image SEO rules, Slovak-inflected filter URLs built from a 3,304-keyword analysis, and a 735-row redirect plan. Privilio was the first project where we committed fully: agent-orchestrated delivery end to end, including design. No parallel manual track, no safety-net team. Same quality bar as always, same type of project as dozens before it — and a stopwatch.

What we kept

Everything that mattered. Our WordPress conventions — the way we structure themes, harden plugins, handle performance — didn't get thrown out; they got encoded as skills our agents apply without being reminded. Fifteen years of craft became configuration. The senior engineer still owned architecture and every ship decision.

What changed

Nothing mutated without a rehearsal. 77 purpose-built runners followed a read → dry → apply → verify cycle; every apply was preceded by a dry plan and followed by a verification suite that had to pass completely. Agents built, agents attacked, seniors judged. Parallel build agents with strict file ownership, adversarial review agents prompted to refute their work, and a human senior as the merge gate — with bulk operations permission-locked so only a person could pull the trigger. Same-day turnaround became the norm. The client's change batches — and all 9 findings from Pavel Ungr's independent 262-URL pre-launch crawl — were fixed and deployed the day they arrived. The fix window was scheduled for four days; it took one.

The headline numbers, measured against our own pre-agentic scope estimate: a 48–72 person-day scope, delivered in roughly 16 active build days — about a quarter of the estimated effort. And the part we didn't expect: quality went up — some 780 recorded verification checks, unit-tested SEO modules, and security hardening at a depth a manual budget never justifies on a project this size, because agents don't get tired of writing tests.

What this means if you're buying

The honest version: this is one project, measured against our own history — not a laboratory study. But it's a real before/after with fifteen years of "before." It's why our quotes look the way they do, and why we now run every project on this model. Launch-ready in 9.5 weeks — content-complete staging, independent SEO audit passed, DNS switch scheduled days after handover. A client who owns her site — everything editable in WP admin, backed by an 8-section admin manual and a worked content-entry guide we wrote for her team.

The full case study →