A faster draft can still leave the handoff waiting.
Explore a small fictional workflow to see why improving one step does not automatically increase the amount of finished work.
A simplified educational model with synthetic capacities. It is not a benchmark, research finding or estimate of workplace performance.
Download the capacity notes (HTML) ↓Follow the work.
Find where it waits.
Each note passes through drafting, source review and completion. These invented capacities make the constraint visible.
- Draft12notes / period
- Review8notes / period
- Complete10notes / period
Up to 8 completed notes per period
Review is the constraint. Drafting and completion can handle more notes than review can check.
Simplified capacity model only: equal task sizes, sufficient incoming work and no rework or interruptions. These are assumptions, not observed results.
Define finished work before measuring speed
Imagine a fictional team preparing short internal handoff notes. A note counts as finished only after it has been drafted, checked against its sources and placed in the agreed handoff location. Producing an unchecked draft is progress within the process, but it is not the same output as a completed note. If the definition changes halfway through a comparison, a faster intermediate step can look like a finished result.
The illustration on this page assigns hypothetical capacities to those three stages for the same work period. It assumes each note passes through every stage, that the notes need comparable effort and that the team has enough work available. These simplifying assumptions make the constraint visible. They do not describe a real team, account for every delay or establish how an AI system would perform.
Follow the constraint through the workflow
The slowest stage limits the illustration’s completed output. If drafting can produce more notes than review can check, increasing drafting capacity alone leaves the completion limit unchanged. More drafts may accumulate before review. The purpose of the comparison is to ask where work waits, rather than to suggest that every process behaves like a tidy sequence of stages.
Try the three synthetic scenarios in the interactive example. The baseline has a review constraint. The faster-drafting case changes only the first stage, so the same review constraint remains. The extra-review case moves the limit to the completion stage. That last result is a reminder to examine the whole path whenever a change moves effort or waiting time from one person to another.
Measure a useful boundary
For an actual evaluation, choose an output whose quality and completion can be checked. Record the work entering the process, the work still waiting and the work completed, using the same definition in each condition. Include correction work and rework when deciding whether a draft was useful. A change that creates more output for someone else to repair may simply move the burden.
A small model can help a team ask better questions, but it cannot answer them on its own. Unpredictable arrivals, different task sizes, interruptions and parallel work can all change the result. Keep those limits next to the illustration and use observations from the real workflow before making an operational decision. No value entered or selected here is submitted to another service.