October 11, 2026 · Timothy O’Brien
Who Owns the System When the People Are Gone
Capacity systems get forgotten and abandoned for one reason: ownership drifts. Teams merge and disappear, people leave and join, and every tool you bought still works — but none of it helps if nobody is responsible for what the systems consume.
Yesterday’s post was about the money: rack-scale GPU systems, where the rack is the cheap part and the building sets the rest. Today’s post is simpler, and it is the other half of managing capacity, cost, and scale: all the technology solutions in the world won’t help you if you lose track of who is responsible for each system you run.
The one constant is change
If you have operated a system to manage capacity, cost, and scale for a thousand internal customers, you already know the one constant: change. Teams merge. Teams split. Teams disappear. People leave the company and people join the company. The systems stay on the bill. A system without an owner stops being a system and becomes a cost.
The median full-time worker in the United States had been with their employer 4.1 years as of January 2026, per the Bureau of Labor Statistics.1 Run a system for ten years and its owners will have turned over at least twice. The system doesn’t care. It keeps consuming and keeps being charged, and every year the question “who do we ask about it?” gets harder to answer.
What forgotten looks like
Forgotten systems are not dramatic. They are quiet. A service that was a priority three years ago still holds its full allocation. A storage volume left behind by a terminated instance keeps getting charged exactly as if it were attached — AWS’s own guidance notes that unattached, orphaned volumes incur the same charges as attached ones.2 A workload moved to a new region, and the old stack was never retired.
None of it is dramatic, and none of it goes away by itself. The technology is working as designed. What went away is the person who could explain the system — and nobody noticed, because nobody was assigned to notice.
Effective ownership, and the test for it
The dangerous number is what percentage of your capacity is effectively owned. Ownership is one of the biggest challenges in managing capacity and scale, and effective ownership is a stronger claim than a name in a spreadsheet.
A system is effectively owned when somebody, today, could answer for it. The test: if it broke tomorrow, or if somebody showed up and said “we’re taking this back, because nobody has vouched for it” — could you, within the hour, name the person who would answer for the business impact? If the honest answer is no, it is not effectively owned. It is a quiet system, and a quiet system is a question (chapter 1).
Note what the test is not. It is not “is there a name in the field.” A name in a field is documentation. Ownership is being able to answer the break-or-take-it-away question — which means the name has to be re-earned, not just recorded once.
Four responses that work
Reaffirm ownership on a calendar. Force everyone who consumes capacity to reaffirm ownership and responsibility for what they consume, every year. If you cannot get a name back, you have found a gap while it is still only a cost question.
Run automatic gap detection. Something automatic should identify the gaps: unattached volumes, idle allocations, and any resource whose owner has left the team or the company. Detection is the cheap part. The expensive part is the defined process behind it — who inherits a resource when its owner is gone, and who decides when nobody volunteers.
Look ahead at reorganizations. If a change in the structure or composition of the organization is coming, look ahead and ask how it will affect ownership. A reorganization that moves three hundred people without an ownership handover takes your ownership map apart in an afternoon. Reorganizations are the largest single driver of ownership drift, and the only one with a calendar — you can see them coming, so ask the ownership question before the announcement, not after.
Measure it. Report what percentage of your capacity is effectively owned, every period, next to utilization and spend. Utilization without ownership tells you a system is being used; it does not tell you anyone would notice if it stopped.
Why this beats every tool
A utilization dashboard is wonderful, and it answers a different question. A dashboard can tell you a system is idle; it cannot tell you whether idleness is a decommission candidate or a once-a-quarter settlement file. Only a person can answer that, and the whole discipline of managing capacity, cost, and scale is mostly the discipline of knowing which person to ask.
That is why the ownership percentage is worth reporting next to the money. If you have supported a thousand internal customers through years of reorganizations, you have watched perfectly good measurement and allocation machinery keep running while the answers behind it went stale. The machinery does not fix this. People, reaffirming, on a calendar, is what fixes it.
Sources
U.S. Bureau of Labor Statistics, “Employee Tenure (news release), Table 1,” January 2026 collection: Employee Tenure News Release — median tenure with current employer, all workers 16 and over: 4.1 years in January 2026 (4.2 in 2016, 3.9 in 2024).↩︎
AWS Prescriptive Guidance: Delete unattached Amazon EBS volumes — unattached (orphaned) EBS volumes incur the same charges as attached volumes, based on provisioned size and type; accumulation of unused volumes can result in significant costs over time.↩︎
