The engineering team is treated as the default unit of accountability, i.e., this group is responsible for this work. This has never been true. Teams are not rewarded or penalized for the success or failure of a project. Individuals are. The team was a workaround for a specific constraint. Software grew too large and too complex for one person to build, test, deploy, and maintain alone. Once you accept that premise, everything else follows. Specialists, handoffs, shared ownership, standups, retros. The team was a solution to a scale problem, not a given.
What the Team Was Solving
That solution carried real costs. Accountability diffused across the group. Coordination overhead compounded with every new hire. Credit diluted, and so did blame. The scoreboard drifted from “are you clear on what you need to do” to “are we directionally aligned.” Organizations accepted the costs because the alternative, one person carrying the whole scope, wasn’t viable. Complexity demanded headcount. Headcount demanded structure. Structure demanded ceremony. The tax was paid every sprint, and nobody mistook it for a feature.
What AI Changed
AI moved the capacity ceiling. A single engineer with competent tooling can now build, test, deploy, and maintain scope that required a squad two years ago. Boilerplate, test coverage, documentation, migration scripts, dashboards, runbooks. Work that used to demand a second or third set of hands compresses into one operator’s workflow. Capacity is no longer the constraint.
The bottleneck has shifted from individual capacity to individual judgment. What to build, what not to build, what “done” looks like, when to ship, where the hidden cost lives, which AI output to trust and which to throw out. Groups can inform judgment, but they don’t own decisions. At some point, one person decides or no one does. That was always true. Headcount let us pretend otherwise.
Collaboration Survives. Shared Ownership Does Not.
Nothing about this eliminates collaboration. Engineers still review each other’s designs, trade context, challenge decisions, mentor juniors, warn each other off bad paths. They do it as peers consulting a named owner, not as co-signers of the deliverable.
That’s what a cohort is. A community of subject-matter experts that provides learning, camaraderie, and peer pressure without blurring who carries the outcome. The social benefits of teams survive the transition. The diffused accountability does not, because the architecture no longer requires it.
Why the Frameworks Compound
Solo ownership is where Outcomes Over Optics and Collaborate by Contract stop fighting the structure.
OOO demands individual-attributable results. The team model made that architecturally hard. You could always hide behind team velocity, team morale, team maturity. Named ownership makes evaluation clean. The outcome has a face on it.
CBC at the ground level, ideally, involves one person per agreement, with explicit deliverables and measurable outcomes. The old-school team model pushes CBC into compromise: multiple signatories, shared deliverables, and ambiguous accountability are still possible, albeit less so. The AI-augmented named owner is the first common configuration where the framework’s ideal and the operational default line up. The agreement names one person. One person owns the outcome. One person is evaluated on it.
”Team” Becomes a Label, Not a Structure
Some work requires coordination. Platform rewrites that cross several domains. Regulated work that mandates separation of duties. Large migrations where the surface area exceeds any one operator’s reach. Incidents that span multiple systems. None of that requires a team as an operating unit. It requires named owners whose scopes overlap enough that they need to talk to each other.
“Team” in this model is descriptive language. It’s a label for a group of individuals working in the same domain or on adjacent systems. It describes proximity, not shared accountability. The billing team is the people who own billing-related work. It is not a unit that owns “billing” collectively with diffused responsibility and a shared velocity metric.
When coordination is needed, name the owners, define the interfaces between their scopes, and let them coordinate as peers. If one scope genuinely can’t be owned by one person, break it smaller until it can. “Because we have always staffed it this way” is not an answer.
Complexity Moved. It Did Not Vanish.
The job hasn’t gotten easier. It has changed shape. Complexity used to live in implementation surface area. It now lives in scope definition, evaluation of machine-generated work, and the taste required to accept, reject, ship, or kill it. The solo operator isn’t a lone genius. They are a single accountable owner whose leverage makes the old coordination tax optional instead of mandatory.
Capacity used to be the scarce resource. Judgment always was. Headcount obscured that fact. Now it does not.
The Organizational Trade
The organization that keeps defaulting to teams will lose ground to the one that defaults to named owners backed by cohorts. Not because teams are bad. Because the coordination tax is no longer free to ignore. When the alternative was impossible, the tax was invisible. It is now visible, and every time a scope goes to a squad that one person could have owned, the organization pays it.
The future of workstream assignment is the named operator and the cohort that sharpens them. “Team” is just a word for the people who work nearby. Not a planning unit. Not an accountability structure. A label.
