The network had grown to 15 outpatient centres running scanners from different brands, and AI had arrived one purchase at a time. Radiology leadership would trial a promising point tool, sign a contract, and hand IT another bespoke integration project. By the time we were engaged, seven such tools were live, each with its own interface, its own support arrangement and its own way of moving images.

The clinical value was leaking away at the last step. Findings from most of the tools never reached the radiologist worklist: results sat in separate viewers and portals that nobody had time to open mid-shift. A radiologist reading a study had no reliable way of knowing which algorithms had already looked at it, or what they had found.

Governance was thinner still. Once a tool went live, nobody monitored how its models performed, so a quietly degrading algorithm would simply keep producing output with no one watching. And every new algorithm the network wanted to adopt meant around four months of bespoke integration work before a single study flowed through it.

We designed the platform around one principle: vendor-neutral routing. Imaging studies from every scanner, regardless of brand, flow into a single pipeline that identifies each study and dispatches it to the appropriate third-party algorithms. When a result comes back, the platform writes it directly into the PACS worklist, so the radiologist sees AI findings in the same place they read the study.

Around the pipeline we built a central console that turns governance from an afterthought into the default. Onboarding a new algorithm is a configuration exercise rather than a bespoke project, access control decides who can see and manage what, uptime and drift monitoring watch every model continuously, and audit logging records what each algorithm received and returned.

The result is one governed pipeline in place of seven separate silos. The network keeps the freedom to adopt algorithms from any vendor, but every tool now enters service, runs and eventually retires through the same managed path, with the same monitoring and the same audit trail sitting behind all of it.