A property portfolio and a cloud estate have more in common than either side admits. Both are collections of assets spread across locations, each with its own keys, tenants, contractors and maintenance schedule, and in both cases the expensive surprises come from things nobody checked. Aviation property owners, hangar developers and estate managers increasingly store leases, drawings, tenant details and access records in cloud accounts. This guide walks through a CloudSecOps routine in seven steps, translated into property-management terms. It is general operational information, not legal, investment or compliance advice.
Step 1: Draw the boundary and name the keyholders
Start with a schedule of every cloud account, subscription or project your business holds, and next to each, the person who answers for it. A managing agent would never run a building without knowing who holds the master key; apply the same standard here. The Ops Atlas entry on CloudSecOps treats this register as the foundation, and it is blunt that a finding without an owner will not get fixed.
While you are at it, mark what lies outside this exercise. Contract review, audit paperwork and software-code scanning are separate jobs. Keeping them apart stops the project from ballooning.
Step 2: Understand the lease between you and the provider
When you rent a unit, the landlord keeps the structure sound and you manage what happens inside. Cloud providers split duties the same way and call it shared responsibility. They secure the platform; you secure your configuration, user permissions and stored files. Write the split on one page so that new staff do not assume the landlord covers everything.
Step 3: Install the building’s incident book
Every managed property has an incident book. The cloud equivalent is a central archive of audit logs, kept where the people who administer the accounts cannot alter it. Do this early, because evidence cannot be created after the event.
Check how long each source keeps its history. Basic event history on one major provider runs to ninety days, with object-level events requiring opt-in and an extra charge. Another provider keeps sign-in records for seven days on a free plan and thirty on paid ones, and does not apply the longer period to past events. Our commentary on simplifying complicated property decisions makes a related point about gathering information before choosing: records first, decisions second.
Step 4: Run the first inspection and sort the snag list
Switch on a posture scanner. An open-source tool is a perfectly respectable place to begin. The first run will produce a long snag list, as any building survey does. Do not try to fix everything. Rank by seriousness, assign a named owner to each critical item within a working day, and agree a due date.
Record which benchmark edition the scanner tests against. Industry benchmarks are numbered and revised, and some provider-built checkers support only older editions. Note the version in your report so that comparisons next quarter mean something.
Step 5: Handle waivers like temporary access permits
Some findings reflect a legitimate business need. A contractor needs temporary access, a partner needs a shared folder. Record each as an exception in a register: reason, owner, expiry. The Atlas caps exceptions at ninety days and suggests a weekly glance at the list. It also lists a quiet warning sign: an exception that has been renewed without being closed. That is a waiver that has turned into a permanent hole.
Step 6: Trial new rules in one building before the whole estate
Preventive rules, such as one blocking public storage, are the cloud equivalent of a building regulation. Pilot them in a low-risk test group first. This matters because on some platforms organisation-wide policies have no evaluate-only mode, so the only preview you get is a trial group. Remember too that those policies may not reach the management account or service-linked roles, so your most privileged identities need separate protection.
Avoid imposing a deny-everything policy at the top on day one. It will stall routine work and prompt a stampede of exceptions that weakens the rule from within.
Step 7: Close by re-inspecting and keep score
After a repair, send the inspector back. In cloud terms, rescan and close the finding only when the scanner agrees; a comment saying “done” is not proof. Then keep a short scorecard.
- How many of your live accounts sit beneath today’s guardrails, as a percentage.
- The share of account-and-region combinations being monitored.
- How long critical findings take to close, reported with the oldest ones still open so the tail is visible.
- Waivers already past their end date, set against all waivers still open.
Beware dashboards that flatter. Certain tools show a control as satisfied once its finding is simply hidden, which is rather like a surveyor accepting a repair photo instead of re-visiting the site.
Keys and locks: a short aside
Encryption keys guard tenant and lease documents, so compile a list showing every key and the systems relying on it. Deleting a key is built to be slow, with waits from seven days up to ninety depending on the service. A misconception worth correcting: changing a key on a schedule does not alter the protection of files already encrypted with it.
A worked picture of the first quarter
Suppose a hangar-development firm follows steps one to four. In the first fortnight it discovers an account created for a feasibility study that nobody owns, and assigns it to the project lead. By week four, its logs flow into a protected archive and its first scan has produced a ranked list of thirty items, of which five are serious and each has a name beside it. In the second month, the exceptions register acquires its first three entries, each with an expiry. By the end of the quarter, the firm can state in one sentence how many accounts are covered by its rules and which ones are not, which is more than it could say in January.
The example is invented, but the shape is typical. Progress shows up first as clarity, not as fewer findings, and that is a healthy sign.
When the full programme is too much
If your cloud footprint is a pair of accounts managed by one small team, the Atlas notes Microsoft’s suggestion to combine platform and security roles, and a lighter routine of protected logs, named owners and rescans is likely enough. A boutique estate manager can reasonably stop after steps one to four and revisit when the footprint grows.
Client-facing firms also benefit from treating this as part of service quality. In our look at personalised service in luxury property, we argue that trust is built in small, repeated, unglamorous acts. Careful stewardship of tenant and owner data belongs on that list.

