A charter operation hands out access all day long. The crew gets the flight plan, the handler gets the passenger manifest, the caterer gets the dietary notes and the accountant gets the invoice. Mixing those up is a nuisance at best and a serious privacy problem at worst. The digital twin of this discipline is IAMOps, and the eight practical items below translate it into plain actions for a small aviation business that lives on shared cloud accounts and booking tools.
1. Treat permissions as a document you maintain
The starting idea comes from the xOps School Atlas entry covering IAMOps, which labels the word an informal one and applies it to what it calls authorisation as code. In everyday words, that means the rules about who may do what are written down, kept in one place, reviewed before they change, and tested. If your current system is “the admin clicks around when someone asks,” you have found your first improvement.
2. Give each role only what the job needs
A dispatcher needs to see trip requests and update aircraft availability. A dispatcher does not need to edit payment settings. This least-privilege idea dates back to a 1975 paper, and the Atlas treats it as the intellectual root of the whole topic. The cheapest first move is to list every role in your booking, accounting and email systems and ask of each permission, “what job would break without this?”
3. Tell grants and ceilings apart
Some rules give access; others merely cap it. An operator’s certificate is a ceiling on what the company may do; it does not by itself win any trip. In cloud accounts, the Atlas says organisation-wide limits behave the same way: they never hand out access, and a clear denial beats a grant wherever it applies. Knowing which kind you are editing prevents hours of head-scratching.
4. Retire long-lived keys
Static keys are the lost gate-code of the digital world. The Atlas points to a late-2025 cloud security survey where year-old keys were still live for 59% of AWS users and 55% of Google service accounts. When a program has to reach a cloud service, prefer a short-lived pass obtained on demand over a secret stored in a file. The Atlas names this move, federation in place of permanent keys, as one of the highest-value steps.
5. Compare what is granted with what is used
Run a utilisation check. In one of the Atlas’s sample calculations, a single role could perform 180 actions yet exercised only 27, a utilisation of 15%. Do the same for your busiest roles, then trim in small stages and write down every removal so it can be reversed quickly if a forgotten task surfaces. A trimmed permission that is easy to restore costs you very little; one that nobody recorded can cost a morning of confusion. Watch out for rare tasks: an annual compliance export or a once-a-quarter fuel reconciliation can look unused in a short window and then fail the day you need it. Ask the owner first.
Our readers who like to understand how planning works behind the scenes may enjoy the explanation of the mechanics of empty-leg private flights, which is another case where careful scheduling rules make the system hold together.
6. Test guardrails before you trust them
The Atlas describes a nasty failure mode: some policy engines skip a rule that errors, so a guardrail you believed was blocking something silently stops working. Another is a negated condition that matches when the data it checks is missing. Both are caught by testing in a sandbox and by validating the rule against a schema. Treat a new guardrail like a new checklist item: rehearse it before the first live trip.
7. Keep an exception register
There will always be special cases, such as a contractor who needs wider access during a software migration. Record who approved it, why it exists and when it ends. The Atlas includes an exception register in its ninety-day milestones. Without an end date, exceptions become permanent residents.
8. Measure a few things monthly
Pick a short list and keep it steady: how many credentials are past the 90-day mark, the number of denied requests following each release, the fraction of wildcard allow statements (the Atlas’s example had 168 out of 1,400, or 12%), and how long a legitimate request takes to be granted. In that same example the wait for a new grant lengthened by half a day after cleanup, and nobody minded. Slower is not always worse if the result is safer.
Bonus: write down who owns each identity
Every account, role and automated job should have a name next to it, preferably a person and a backup. The Atlas recounts a January 2024 incident at a large software company in which an old test application, still carrying elevated rights, became the way in. Nobody was responsible for it, so nobody noticed. In a charter business that could be a long-forgotten booking integration still holding read access to client lists. An owner column in a simple sheet costs nothing and answers the first question anyone asks during an incident: whose is this?
Add a review date to that sheet and revisit it at the same time you review crew currency or aircraft inspections. Aligning the two routines makes it easier to keep both going.
Putting the eight into a calendar
- Week one to four: name owners, list roles and keys, and look for anything reachable from outside your organisation.
- Week five to eight: move policies into a repository, add automated checks, and trim one over-generous role.
- Week nine to twelve: federate half your remaining automated jobs, run the first monthly review and decide on a tool for evaluating rules.
If your business also thinks about repositioning flights and shared-seat options, the second article, a walkthrough of empty-leg flights for private jet passengers, adds useful background about how operators stretch their schedules.
None of this requires a large budget. The Atlas’s cost example prices a provider’s unused-access scan at around $170 a month for 850 roles, and its custom policy checks for a few hundred changes at about $4. The expensive part is attention, and the eight items above are a way of spending it on purpose.

