What does human attention per outcome make visible?#
The work no delivery metric counts: reviewing agent output, preparing briefs, catching up after time away and fixing what came back wrong. An organisation can double its pull request volume while its human attention per outcome rises, a loss that looks like a gain on most AI productivity dashboards. Herbert Simon put it plainly in 1971: a wealth of information creates a poverty of attention. We propose counting it in accountable senior minutes per delivered outcome.
Why is human attention per outcome never a measure of individuals?#
Because every input is a property of the system: brief quality, validation reliability, captured context. Using the number to compare people turns a diagnostic into a threat, and the data stops being true.
Human attention per outcome sits beside DORA metrics for delivery and FinOps for AI for spend. Token counts, pull requests and agent utilisation stay diagnostics: optimise any of them and you simply get more of it.