Classification: measurement method; current Pro 100 human recovery and review minutes are UNKNOWN. This article reports no time-saved result.
View the full infographic → · View this article
“Autonomous” does not mean that a workflow used no human time. Setup, review, approvals, authentication, retry decisions and recovery can all matter. Measure these separately from machine runtime and provider wait so that a cost or capacity comparison has a meaningful denominator.
Define the window and count accepted work
Choose one start/end window and one job class. Count unique jobs only after the stated verifier accepts the result. Record incomplete attempts and duplicates; do not divide by all scheduled jobs and call them successful output.
Record the time components
- Human active minutes for setup, prompt/context preparation, review and recovery.
- Machine active runtime, separately from provider/resource waiting.
- Wait duration by cause, such as resource limits or authentication.
- Retries, rework and the verified final outcome.
A descriptive ratio is total human active minutes divided by verified jobs. Machine minutes per verified job can be reported separately. Total elapsed time is not the sum of machine time if jobs run concurrently; state whether a metric is per-job or wall-clock.
What remains unknown here
The current Pro 100 evidence has no instrumented human reauthentication minutes, useful machine hours or account-wide verified job count. The £89 charge is user-recorded and its billing window is not verified. Therefore cost per verified job, human-to-machine time ratio and output per hour remain not calculated.
The Context Cost Calculator keeps measured-job and useful-hour inputs blank until supplied. A dedicated human-vs-machine calculator remains planned; neither tool turns missing measurements into zero.