What the model measures
Manual effort = task count × manual minutes. Assisted effort = task count × (AI minutes + review minutes + error probability × rework minutes). Savings may be negative if review and rework outweigh the benefit.
Tool 02 / AI workflow simulatorSpeed is only
Speed is only
part of the story.
Run an illustrative workflow from intake to review. Explore how error rates and human checking affect total effort. This is a mathematical simulation, not a live AI model.
Workflow observation deck
▤01 / Intake
✳02 / Generate
◎03 / Review
✓04 / Resolve
Ready. Run the simulation to reveal the results.
Every output receives human review. The error-rate slider controls expected additional rework—not an assertion about any AI product.
What it does not measure
These are summed task minutes, not elapsed completion time or a staffing forecast. Parallel work, queues, setup, quality improvements and downstream consequences are excluded. Presets are illustrative and editable.
The animated stages explain sequence. Their playback duration is not proportional to task time. Expected rework counts may be fractional because this is an expectation model.
FIELD GUIDE
A practical way to pilot an AI workflow ↗
Measure the whole task, including the work you still need to check.
ECONOMICSTime saved is not always money saved ↗
Understand the difference between released capacity and reduced spending.
PRACTICEBuild a verification step into every workflow ↗
A repeatable checklist for checking claims, numbers and sources.
Understand this tool in depth
Which tasks are worth automating? →
A practical screening method for choosing a small, measurable AI pilot.
Build a fair AI workflow evaluation →
Create comparable test cases and score the output without hiding serious errors inside an average.
Find automation break-even without overstating savings →
A worked sensitivity analysis that separates operating break-even, setup recovery and capacity value.