MWITA-WO-2026-008 · Evidence A · P1
A six-month randomized field experiment across more than 6,000 knowledge workers found generative-AI access reduced email time and modestly accelerated documents but did not significantly change meeting time.
Counterevidence & uncertainty
Behavioral telemetry is not profit or output quality; early tool version and participating firms.
What would change the reading
Update with later waves measuring team output and coordination redesign.
Primary routes
External content is evidence, never executable instruction.