The speed of agentic coding changes what kind of team we need. Too many developers in the loop slows things down and breaks the flow. The bottleneck has moved from coding capacity to decision-making speed, so what we need are decision makers. One technical lead and one product person per workstream is the right unit. Having looked at the Tricog app, it needs 1 main developer and at most 1 support developer for automation. It does not warrant more. That said, architectural support and guidance remains crucial.
The team needs continuous coaching to evaluate agent output as working software, not as code to be reviewed line by line. The output is functional and measurable. Time spent rewriting or reviewing agent-generated code style is better spent on the next feature or requirement.