Accountability as Infrastructure

The outage postmortem produced eleven action items. Everyone agreed they mattered. Six months later, at the next outage’s postmortem, someone pulled up the list. Zero were done.

Nobody failed to do their job because none of the items belonged to anyone. They belonged to “the team,” meaning the gap between people. The organization concluded it had an accountability problem and scheduled a workshop on ownership.

Wrong diagnosis. The team didn’t lack a trait; the organization lacked a structure. Accountability isn’t a virtue you hire for or a value you laminate. It’s infrastructure: a system of named ownership, explicit handoffs, and scheduled truth. Most companies never build it and then hold workshops on why no one feels responsible.

The Virtue Theory of Accountability

Ask most leaders how they ensure accountability, and the answers are all character-based: hire owners, promote the ones with “an ownership mindset,” and repeat the value at all-hands until it sticks. When accountability fails, the diagnosis is character too. Someone lacked follow-through. Someone didn’t step up. The fix is always more messaging or better people.

Call this the virtue theory of accountability. Notice what it requires: responsibility to materialize spontaneously, in the right person, at the right moment, without anyone deciding in advance whose it was. If your accountability depends on everyone feeling responsible independently, you don’t have accountability. You have hope and a values page.

The virtue theory also has a dark twin. When something breaks, and no structure says who owned what, the postmortem becomes an archaeology of intent: who touched it last, who should’ve known, who seemed insufficiently sorry. Blame’s what accountability looks like when nobody built it. The organizations most obsessed with finding the responsible party after the fact are, reliably, the ones that never named one before.

What the Military Builds Instead

People assume the military runs on accountability as intensity: standards, discipline, someone loud making sure. Years in Airborne and Special Forces units taught me the opposite, which I brought to engineering. Military accountability is boring. It’s the most thoroughly built structure I’ve operated in, and almost none of it depends on anyone’s character. Any veteran can name the exceptions; the structure’s the point.

Every mission has one named commander. Not a committee, not a partnership: one name decided before anything moves. Ownership transfers are ceremonies with verbal confirmation because ambiguity about who has command is how people die. Equipment runs on signature chains: you sign for what you hold, and “the squad’s rifle” isn’t a category that exists. The after-action review is calendared, formatted, and rank-suspended. It happens after every operation because it’s scheduled, not because anyone feels reflective. The suspension of rank isn’t warmth. It makes the truth affordable to say.

Here’s the part that transfers: the military built all this because it can’t staff for virtue. It runs on nineteen-year-olds and rotates everyone constantly. It’s an organization designed around ordinary people with permanent turnover, so accountability had to live in the structure since it could never be guaranteed to live in the person standing there this quarter.

That’s exactly the condition of a scaling engineering organization. Virtue works in a team of five that never changes. Add growth, churn, and reorgs, and every quarter your company’s partly strangers. Culture evaporates first, but structure survives turnover.

Software organizations dissolve ownership faster than most. Services multiply past the headcount. Matrix structures give everyone two managers and nobody the pager. On-call rotates, reorgs shuffle, and the engineer who understood the system is two teams away when it breaks. None of that’s anyone’s failure. It’s what scale does to accountability that nobody built.

The Load-Bearing Components

Four components, all buildable, none requiring anyone to become a better person.

One name per thing. Every service, decision, and action item carries exactly one owner (the DRI pattern, for those who want the searchable term). Two names is a negotiation. “Shared ownership” is the org-chart spelling of nobody’s job. Contribution can be shared. Accountability can’t. This feels harsh until you notice what it protects: the owner gets authority with exposure, and everyone else gets someone to hand problems to instead of a fog to shout into. The pairing isn’t optional. An owner without authority isn’t accountability; it’s blame with paperwork.

Handoffs are events, not drift. Ownership changes by explicit transfer with acknowledgment: a name out, a name in, both parties aware it happened. Most orphaned systems were never abandoned on purpose. They were handed off by reorg, resignation, or a Slack message no one confirmed. If your ownership records change mainly by erosion, you don’t have ownership records. You have sediment.

Scheduled truth. The after-action pattern, ported whole: retrospectives and postmortems that happen on the calendar, in a fixed format, with blame suspended. Blameless doesn’t mean ownerless. The postmortem doesn’t hunt culprits for causes; it names owners for actions. Suspending blame isn’t a kindness program. It’s the price of accurate information paid deliberately.

Consequences that are boring. Accountability collapses at both extremes: consequences that never arrive or consequences so dramatic everyone conspires to avoid triggering them. Infrastructure makes them small and inevitable. Action items carry a name and a date. The next review opens by reading the last review’s list. That’s the entire enforcement mechanism, and it’s astonishing what gets done when people know the list will be read. Repeated misses stop being a character issue and become a resourcing issue decided openly.

Boring Is the Proof

Run the opening postmortem again inside that framework. Eleven action items, eleven names, eleven dates. The next scheduled review opens with the list, which is mostly done; the two items that aren’t have their owners explaining what they need. No workshop, no soul-searching, no archaeology. Nothing about the people changed.

That’s how you know accountability has become infrastructure: it gets boring. Nobody notices the water main until it breaks, and nobody credits the ownership registry for the incident that resolved cleanly at 2 pm on a Tuesday. The virtue theory produces heroes, apologies, and repeat outages. The infrastructure produces quiet.

Stop hiring for it. Start building it.