The story below is invented. It involves an imagined hospital IT team, an imagined vendor and an imagined Tuesday, and if it resembles a real organisation, that is coincidence. It is here to illustrate AccessOps, the name a reference catalogue gives to the way requested, elevated access is granted, watched and withdrawn. Everything in it concerns administrative systems, not patient care, and the lessons come from published practice rather than from any single incident.
A quiet Tuesday and a stuck export
The operations team at a mid-sized cancer-care hospital had three people, one shared inbox and a long list of small requests. On this Tuesday the finance office reported that the nightly export from the appointments platform to the billing system had stalled. The fix needed a configuration change that only an administrator role could make, and the vendor’s support engineer offered to make it directly.
The team lead, whom we will simply call the lead, felt the familiar tug. The quickest way to resolve the matter was to create an administrator account for the engineer, mail the password and move on. The same shortcut had been taken the previous year. That account, as someone later discovered, was still enabled.
This time the team had recently read the Ops Atlas entry on the subject and had agreed on a rule: extra access is granted only against a stated reason and a fixed finish time.
Five fields on one page
The lead opened the request form and filled in five items: who the engineer was, what the role could touch, why it was needed, when it would end, and who was approving. The reason was specific: restore the stalled export. The window was three hours. The approver was the lead on duty, and the form would route anything longer to a second person.
Writing the form took about four minutes, which struck the finance office as slow until the team pointed out that the old method had involved a password in an email, a promise to disable the account “later” and no record at all. The Atlas describes this five-field record as the minimum useful unit of access management, and the team had found that nothing about it was ceremonial. Each field answered a question someone would otherwise ask in a panic six months later.
Eligible, then active
The engineer did not receive a permanent administrator account. The vendor’s identity was set up as eligible for the role, and the role was switched on only after the request was approved. It switched itself off when the three hours ended. The team had adopted what the Atlas calls the eligible-not-active pattern, and the person who set it up described the change as moving from handing someone a key to handing them a ticket that opens one door once.
There was a wrinkle. The first configuration the team tried used a tool that checks credentials only when a session starts. That meant a session opened at minute one would not be cut off at minute one hundred eighty. The Atlas reports a lab case where a remote session ran on for about twenty seconds past its credential’s expiry; for the hospital team, the practical lesson was to add a step that explicitly ended the engineer’s open sessions when the window closed. Expiry alone was not the whole story.
The fix, and the part nobody enjoys
The export resumed within the hour. The engineer logged off and the request closed. Then came the part teams are tempted to skip: a short review. The lead looked at the record and asked whether the three-hour window had been right. The work had used about forty minutes. That tallies with the Atlas’s own worked month, in which only a quarter of granted time was actually used, and it prompted the team to try a shorter default for vendor tasks of this kind.
They also checked what had changed. A short recording of the session, which the team had turned on for vendor access after reading the page’s advice, showed one configuration edit and a restart. Nothing surprising. The review took ten minutes and produced one note: reduce the default window to ninety minutes for routine vendor fixes.
One more detail deserves mention. Because the request form was a shared record, the finance office could see the status without phoning anyone. Waiting time, the sort of thing the Atlas tracks as time to access, fell noticeably once people could see the queue. The team noticed fewer chasing messages and fewer informal workarounds, which is the quiet payoff of a process that is fast enough that nobody is tempted to bypass it.
The old account
While preparing the request, the team had also found the administrator account from the previous year, still enabled, belonging to a contractor who had left long before. It had not been used for months. The team applied a rule the Atlas mentions from two sources: one review tool recommends a Deny decision once an account has gone thirty days without a sign-in, while a popular control framework marks anything idle for forty-five days as dormant. They disabled it that afternoon and added a monthly check.
No harm had come from the stray account. That is the awkward truth of this kind of work. Most of the time nothing happens, which is exactly why the risk goes unaddressed until the day it does.
Practising the emergency route
The Wednesday after, the team ran the exercise the Atlas recommends in its first thirty days: a drill of the emergency account. They treated it as a real event. Two cloud-only emergency accounts exist, protected with strong sign-in methods, and using one sends an alert to the whole team. The drill revealed that the alert went to a mailbox nobody read. They fixed the routing and made a note to test again in ninety days, matching the interval a major vendor recommends for such accounts.
The emergency accounts are, deliberately, always active. The team’s earlier instinct to disable them when not in use would have defeated the purpose, since the whole point is that they work on the day the normal sign-in does not.
What the team took away
- An extra-access request is five facts and an end time, written down before access is given.
- Switching access off is not the same as ending every session that began while it was on.
- Brief reviews of recent grants reveal whether the default window is too generous.
- Dormant accounts are a quiet liability and deserve a monthly look.
- Emergency accounts stay on, stay alarmed and get rehearsed.
Reading the source
The team’s reading list started with the Ops Atlas write-up on AccessOps, which also cautions that the term has no formal standard behind it and shares its name with several unrelated products. The practices it describes, however, are well established. Readers who want to go deeper into the Atlas’s scope and maturity steps will find them there; the framing here is a story, not a substitute for the reference.
For readers of this site, the stories that matter more are the ones families live through. If you are looking for background on questions to prepare when a diagnosis arrives, our guide on which questions to bring to the first conversations after a diagnosis is a sensible starting point, and the article about how treatment planning is usually organised at cancer centres explains the coordination that administrative teams quietly support. Behind both sits a team like the one above, keeping the keys few, short-lived and accounted for.
