The team in this story is invented — a composite of small clinic IT groups everywhere, with details changed and no real operator implied — but every practice described is drawn from documented public material, including how the Atlas frames SecOps: a working contract, signed between the team that watches for attacks and the team that keeps the systems running. Read it as a precise fiction: what three sites with four IT staff did in ninety days when they stopped treating security findings as weather.
Three sites, one spreadsheet, and a scan that saw nothing
Meridian Orthopaedic Group ran two surgical centres and a rehabilitation clinic on a shared record system. Its security tooling consisted of a vulnerability scanner installed two years earlier and an export someone pasted into a spreadsheet every fortnight. The spreadsheet had 640 rows. Nobody owned rows; people volunteered for them, occasionally, the way neighbours water a plant. The reveal came during a routine insurance questionnaire, when an external assessor asked Meridian to prove its internet-facing systems were scanned monthly. They had been — or the scanner had run. What nobody had checked was whether the scanner could see. A certificate on the perimeter appliance had expired roughly ten months earlier, and the scan reports had been politely blank on the asset class that mattered most.
The parallel they could not ignore
The practice manager, preparing the remediation story, reread the public record on Equifax’s 2017 breach and found the mirror uncomfortable. The GAO’s account notes an outdated recipient list for US-CERT notices, a scan that failed to catch the Struts flaw on the affected server, and an expired certificate — some ten months old — that blinded inspection of traffic for about 76 days while attackers worked undetected. A House report later counted more than 300 expired certificates across the estate, though the record notes it came from majority staff rather than the full committee. Meridian’s incident was microscopic by comparison and harmed no patient. But the mechanism — a lapsed certificate quietly disabling the sight, and no owner noticing — was identical, and the team wrote that sentence into the first slide of its proposal.
Month one: the SecOps agreement, in one page
The new operations lead resisted the temptation to buy anything. Instead the group signed a one-page working agreement with six clauses:
- hardening baselines chosen from public benchmarks, tailored with departures recorded;
- one register for every finding, wherever it originated;
- a person’s name and a resourced date attached to each entry;
- exceptions that expire and must be re-earned on review;
- change windows for security work, agreed with each site in advance;
- one monthly page of numbers, read jointly by both sides.
Then came the unglamorous part that made it real — owner fields added to the CMDB records, so routing could be mechanical rather than heroic. The lead’s note to the directors was characteristically dry: the agreement cost one meeting and a page; the spreadsheet had cost two years.
Month two: a pilot ring on the smallest site
The rehabilitation clinic, smallest and least critical, became the pilot ring. The first tailored baseline ran there in a logging posture — rules recording what they would have changed, changing nothing — for three supervised weeks. The register, meanwhile, was wired to route findings automatically from the scanner into owner queues; by the end of the month roughly 80 percent of new findings placed themselves, in line with what maturity guidance expects of an automated first level. The logging data earned its keep at the three-week evidence review, where two rules were re-tailored before they could have caused a single outage — precisely the return the quiet phase exists to buy. The team also ran its tabletop: a two-hour rehearsal of a perimeter certificate expiring, walked end to end — who checks expiry, on what schedule, who gets called, what the fallback is. The exercise found the gap the real incident had found, which is the point of rehearsing.
Month three: the first scorecard
Ninety days in, the first joint scorecard went to the directors, and it was deliberately unflattering. Telemetry coverage sat below target for the acquired clinic’s inherited systems; several exceptions inherited from the spreadsheet era had no review dates and were re-earned or retired; the disruption counter recorded one security change that had briefly slowed appointment scheduling — logged, owned, and used as the argument for extending pilot rings before wider enforcement. Staged enforcement of the first baseline began on production systems only after the pilot’s evidence review, ring by ring. The group had, in maturity terms, walked from improvised to managed inside a quarter: routing automated, a scorecard held for three consecutive months, and no finding entering the estate without a name attached.
Three habits that made the quarter stick
- The fortnightly spreadsheet export was retired on day one — one register, no shadow copies feeding side conversations.
- Every exception review got a calendar invitation with the system owner present in the room, not merely copied on a thread.
- The disruption counter was reported beside remediation speed with equal prominence, keeping the hardening agenda as accountable as the fixing agenda.
None of these habits appears in a product brochure, and none required procurement. They are the small mechanics that turn a signed page into behaviour, which is where most agreements quietly fail.
Where the SecOps ladder goes next
The documented maturity path gives the group its next targets, and they are specific rather than aspirational. Level three is measured, not felt: fewer than one exception in ten past its review date, remediation tracked per cell of asset importance and urgency, and the scorecard boring enough to read in four minutes. Level four is adaptive — baselines revised as platforms change, telemetry treated as a covered asset class, the agreement edited like any other living document. What the quarter actually purchased was not tooling. It was the conversion of goodwill into obligations, which is what allows a four-person team to make promises a director can rely on. Teams running equipment fleets will recognise the discipline from managing and scaling working robot fleets: every unit accounted for, every fault owned, inspection after intervention. The architecture side — where the contract sits among the wider operations disciplines — is set out in the unified architecture of XOps, and for clinics that would rather buy execution than build it, expert engineering support for software and cloud operations shows what that handoff looks like. The moral of the composite is plain: Meridian did not need a bigger budget in month one. It needed a page, an owner field, and one honest scorecard — all three of which were cheaper than the certificate that started the story.
