The AI-Ready Healthcare Infrastructure Campaign: Connecting Clinical AI to the Hardware Underneath It
Hospital AI Needs Somewhere To Run. This Campaign Connected the Clinical Story to the Hardware Story.
The correction was the interesting part, not the initial build.
What was measured.
Three readings from a six site campaign. Two of them are about things that did not happen, which is where the value actually landed.
The value was not publishing five more articles. It was catching the wrong destination before the wrong site absorbed the content.
Refusing to link a domain that does not resolve is a boring check. It is also the difference between a topical cluster and a set of dead ends.
Backups per round is the honest measure of surface area. A cluster campaign is only as good as the specific site each spoke lands on.
The short version.
IBM was baking AI into Power hardware and IBM i tooling at the same time healthcare AI properties in the portfolio had no way to talk about the infrastructure, data locality, and governance that clinical AI actually depends on. The campaign turned that gap into a linked cluster across six sites, built around one central hub, corrected a wrong target subdomain mid-build, and deliberately left out a healthcare-related domain that wouldn't resolve rather than publish a broken link.
Explain it to me like I'm ten.
A hospital that wants AI also needs computers strong enough to run it. We connected our health websites to our IBM computer websites so people researching either one can find both. Partway through we caught ourselves about to publish on the wrong site, and we skipped linking to a website that doesn't actually work instead of pretending it does.
What was the system?
AIHealthcareNow.com and AIMedicineNow.com talk to a healthcare audience about clinical AI and governance. AS400IBMSystem.com, Power11.AS400IBMSystem.com, and AS400Software.com talk to an IT audience about IBM Power hardware and IBM i modernization. Nothing connected the two conversations, even though the honest answer to "can my hospital run this AI system" is an infrastructure answer as much as a clinical one. The campaign's job was to build that connective tissue: a central hub, real spoke articles on each property framed for that property's actual audience, and cross-links a reader can actually click through, instead of passing mentions.
What changed?
AIHealthcareNow.com became the hub, publishing at /library/healthcare-operations-ai/ai-ready-healthcare-infrastructure/ and pinned to the homepage and category page. Five spoke pieces followed: a hospital-infrastructure article on AIHealthcareNow itself, a clinical-governance article on AIMedicineNow, a Power 11 healthcare IBM i upgrade-planning article on AS400IBMSystem, a Power S1112 local-inference article on Power11.AS400IBMSystem, and an IBM Bob healthcare-modernization article on AS400Software, plus an update to an earlier AS400IBMSystems Power 11 overview. Category and hub copy was updated across all six properties so the new pieces had real entry points instead of sitting as standalone URLs. Federation, sitemap, and LLM outputs were rebuilt for every touched site. A targeted 65-file backup was taken before anything shipped.
The build hit a real correction mid-stream. The original target for the Power S1112 healthcare piece was power11.as400ibmsystem.com. That was wrong; the intended property was the separate power11.as400system.com subdomain. Rather than force the AS400IBMSystem site's per-file content architecture onto a site that doesn't use it, the follow-up session adapted the article into that subdomain's actual data/articles.json collection pattern, added a matching Healthcare AI glossary entry with structured external context links to AIMedicineNow's clinical AI content, AIMedicineNow's governance article, AIMedicineToday's research library, and the AIHealthcareNow hub, and fixed two live-review issues along the way: spec tables running too narrow on mobile, and glossary content links inheriting plain black text instead of the site's usual link styling.
What did the evidence show?
The correction was the interesting part, not the initial build. Two things surfaced only because a human looked at the live pages afterward, not because a check caught them.
First, the target subdomain was simply wrong at the start of the second session, power11.as400ibmsystem.com instead of power11.as400system.com. The two names look almost identical, but they belong to two real, separately run sites in the same constellation, and building on the wrong one would have shipped a whole article to the wrong property. Catching it meant re-adapting the content to a different site's actual data structure rather than copy-pasting the same JSON shape across.
Second, a domain called AIHealthcareToday.com looked like an obvious link target for the new Healthcare AI glossary entry: healthcare-adjacent name, right topic. It didn't resolve, not from local DNS, not from the VPS. Rather than publish it anyway on the assumption it would come online eventually, it was left out. A campaign built around connecting real infrastructure to real clinical need doesn't get to publish a link to nothing and call it thorough.
After the campaign, AIHealthcareNow carried 28 public items across 19 sitemap URLs, AIMedicineNow 141 items across 140 URLs, AS400IBMSystem 55 across 55, the original Power11.AS400IBMSystem 46 across 46, and AS400Software 361 across 339. The corrected Power11.AS400System build brought that subdomain's sitemap to 64 URLs and its federation catalog to 83 public items, including a glossary layer that grew to 22 terms. Those are inventory counts, not ranking or traffic numbers. Whether the cluster earns discovery is a separate question, and the paired tracking entry exists to answer it with data instead of assumptions.
What can be reused?
A topical cluster across six properties is only as good as the specific site each spoke lands on. The value in this campaign wasn't publishing five more articles. It was catching a wrong subdomain before the wrong site absorbed the content, and refusing to link a domain that didn't actually answer when called. Both of those are small, boring checks. Both of them were the difference between a real cluster and a cluster that looks real until someone clicks through.
Keep Browsing the Case Library
Move from this evidence file back into the full proof system.