THE COST OF USEFUL AI / LABS 06 AND 07
Which decision does the evidence support?
Fictional planning examples. Calculations stay in your browser. No provider calls or input uploads.
Price the error profile
The policies start from identical drafts. Flagged drafts are reviewed; good drafts stay good. Review recovery applies to failed drafts sent to review. Generation, gate and review are operating resource costs; the failure consequence is separate and must not duplicate costs already counted. Setup, shared operation and review delay are excluded. Expected counts may be fractional.
drafts / period
% passing before review
% of failed drafts released
% of good drafts flagged
USD / draft
USD / draft
minutes / reviewed draft
USD / hour
% of reviewed failures repaired
USD / final failure
YOUR SCENARIO
- Selective operating cost
- $300.00
- Selective expected consequence
- $800.00
- Selective combined total
- $1,100.00
- Automatic-release combined total
- $2,100.00
- Universal-review combined total
- $1,600.00
- Selective final failures
- 40
- Selective review hours
- 4
- Quality among automatic releases
- 95.45%
The lowest combined cost can still violate an outcome or process requirement. Apply mandatory constraints, check uncertainty and measure review capacity before choosing a policy.
Try the printed exercise
Change the failure consequence from $20 to $50. Selective review rises to $2,300 while universal review stays at $1,600 under perfect recovery. Then reduce recovery to 99% and inspect the remaining failures.