Walk into any decent parts room and you will find tags on everything: serial numbers, batch codes, shelf-life dates, a certificate that says where the component came from. Nobody finds that fussy. A bolt with no paperwork is a bolt you cannot use. Now picture the office computers behind that parts room. They are full of accounts that belong to no person at all: the connection between the inventory tool and the accounting package, the script that emails supplier orders, the key that lets a phone app read stock levels. Those robot accounts rarely get tags. Managing them, along with the human ones, is what IdentityOps is about, and this explainer keeps it short and practical.
The plain definition
IdentityOps is the routine upkeep of every identity your organisation depends on: who or what exists, how each proves it is genuine, and when it should stop existing. The Ops Atlas, which I am drawing from, adds that the word is still young and that nobody owns a universal definition. Fine. For a parts business the working version is simple enough to put on a whiteboard: know every account, know who owns it, and shut it down on schedule.
Two kinds of identity are in scope. Human ones are staff, contractors and customers with logins. Non-human ones, shortened to NHIs, are the service accounts, API keys, tokens and certificates that let software talk to software. The Atlas gives the second group a lot of attention because they are the ones that quietly pile up.
Why machine accounts go wrong
Think about how a part ends up forgotten on a shelf. Someone ordered it for a specific job, the job changed, nobody scrapped it, and years later it sits there with a faded label. Machine accounts follow the same life story. A developer or an office manager creates one for a quick integration, copies the key into a settings file, and moves on. Nobody records an owner or an end date. The key never changes, because changing it might break something and no one remembers what.
The Atlas lists a few classic habits that make this worse: several people sharing one service login so nobody can tell who did what, keys with no expiry, and credentials stored in places such as personal file storage. It also references a community top-ten list for non-human identities published in 2025, which is a handy checklist if you want a second opinion on your own setup.
A parts-room lifecycle for robot accounts
The Atlas describes a lifecycle with five stages. Here they are in the language of a stores clerk:
- Take stock. Walk the shelves. List every machine account in every tool you use, not only the ones you remember creating.
- Tag it. For each account, record an owner, the purpose, an expiry date and how often it gets reviewed. If you cannot fill in the owner column, that account is your first problem.
- Prefer short shelf life. Where the tool supports it, issue credentials that expire quickly instead of ones that live for years.
- Rotate on a calendar. Replace keys on a fixed schedule, the way you replace consumables, and test the replacement before you retire the old one.
- Scrap with a record. When an account is no longer needed, disable it, wait a defined period to see whether anything screams, then delete it, and keep proof you did so.
That “disable, wait, delete” rhythm is worth copying. It is the same idea as quarantining a doubtful part before you scrap it: you give the business a chance to say “wait, we still use that” before the action is irreversible.
A caution about the tools you already have
One practical note from the Atlas deserves emphasis because it trips up well-meaning people. A built-in report from a large cloud provider can list the age of keys, but the Atlas points out that it only covers the first two keys per user and it lists users rather than roles. In other words, a tidy-looking report can hide the very accounts you most need to see. Always ask what a report leaves out before you trust the green ticks.
Sweep for dormant accounts
Another routine that suits a parts business is the dormant sweep. Every so often you pull a list of logins that have sat untouched for a chosen stretch, and the Atlas suggests picking a figure from three to six months, then you switch off whatever is idle. It also lists sensible exceptions: staff on a formal leave of absence, emergency accounts kept sealed for the worst day, and machine accounts whose owner registered a longer review cadence on purpose. Without the exceptions, the sweep creates outages. With them, it is a satisfying housekeeping job.
A real-world warning without the drama
The Atlas recounts a series of intrusions into customer environments at a cloud data vendor during 2024. As the summary has it, the shared gaps were no second sign-in factor, passwords and keys left untouched for up to four years, and no limits on which networks could connect; about 165 organisations received warnings. I mention it not to alarm anyone but because the failure list is so mundane. Each item is a paperwork issue: no second factor, no rotation date, no allow list. A stores clerk would recognise it instantly as an inventory-control failure.
Human accounts need the same tag-and-track habit
It is easy to focus on machines and forget that people come and go in a parts business too: counter staff, warehouse temps, seasonal drivers. The Atlas stresses a leaver sequence that starts from the personnel record and moves through disabling the central account, revoking live sessions and telling connected apps to switch the user off, with timestamps recorded at each step. It also stresses that “disabled” and “offboarded” are not the same thing. An account can be switched off in the main system and still be alive in a supplier portal or a spreadsheet nobody reviewed.
A good habit is to keep a short list of every place a person might hold a login, and to check that list when someone leaves. Do it for the person who ordered fasteners, and for the one who closed the stock count.
A worked example from the counter
Suppose a small distributor connects its stock-control tool to a courier’s booking system through a key that a former employee set up three years ago. Nobody knows the key exists until the courier changes its terms and the integration suddenly fails. With a register in place, the story is calmer: the key’s row shows an owner, a purpose and a review date, so the replacement is a scheduled job rather than a scramble. Without one, the business discovers its dependencies by breaking them. The register costs an hour a month; the scramble costs a day of unshipped orders.
What to do on Monday morning
- Open a spreadsheet and list every system your business logs into. One row each.
- Add a column for the machine accounts inside each one, an owner for each, and a date by which each should be reviewed.
- Pick the three oldest keys and replace them this month.
- Choose a dormancy window and schedule the first sweep.
- Stop any two people from sharing one login. Give each their own.
Nothing here calls for a purchase. It needs an owner and a calendar. If you later want to see how larger organisations structure the programme, with metrics, standards and maturity levels, the entry titled IdentityOps in the Ops Atlas is a thorough place to start.
Related reading from this blog
Procurement habits and identity habits overlap more than you would think. If you are tightening your buying process, you might like this list of pitfalls when purchasing control-system components, and for a refresher on the hardware that holds everything together there is a full guide to fasteners and fittings.
The no-nonsense summary
A part without a record is a liability, and an account without an owner is the same thing in digital form. Tag it, track it, rotate it, retire it. Do that for people and for software, and most of what the industry calls identity operations is already in place. The rest, the standards, the dashboards and the drills, is refinement on top of a solid habit.

