Engineering: measure the work after review
Use a small, inspectable example before you rely on a larger workflow.
All prefilled examples are synthetic. These tools are original editorial exercises, with arithmetic checked where applicable. No client results or professional endorsement are implied.
Include review and rework in the estimate
The hypothetical month compares the same work scope. Baseline preparation is 200 hours with 20 hours of review and rework. The AI-assisted estimate is 140 preparation hours and 40 review/rework hours.
What the worked result means
Total modeled effort falls from 220 to 180 hours. Forty hours have an internal value of $6,000 at the supplied rate. Subtracting $5,000 recurring and setup costs leaves $1,000 modeled capacity value. None of those figures is observed revenue, payroll savings or a causal estimate of AI productivity.
Equal failure counts do not establish equal quality: severity, sample size and scope may differ. Do not use this calculation alone to justify a staffing change.
Run a trial you can interpret
Define eligible tasks, the timing rule and quality conditions before the trial. Record preparation, reviewer time, rework, accepted outcomes, tool versions and missing observations. Include concurrent work in the timing rules. Compare comparable categories and keep learning/order effects visible.
The blank log collects actual observations. It is not prefilled with successful trial results. SPACE describes productivity across multiple dimensions; METR's February 2026 update explains why task selection and timing can distort apparent gains; DORA reports how verification can move work to reviewers.
Download the working files
- Editable pilot cost and trial workbook
- Blank observation log, CSV
- Calculated demonstration results, JSON
What supports this approach?
The references support specific principles such as verification, explicit assumptions and measuring review. They do not validate these exact worksheets or certify your results.
- Cost Estimating and Assessment Guide: Best Practices for Developing and Managing Program Costs
- The SPACE of Developer Productivity: There's more to it than you think
- We are Changing our Developer Productivity Experiment Design
- Balancing AI tensions: Moving from AI adoption to effective SDLC use
Read the task guide
