WORKFLOW TIME REVIEW
Use privately. Do not include confidential task material.

Task and acceptance standard:
Comparison window and repetition count (N):
Basis: planning estimates / observed times (describe how collected)
Usual effort per task (M), minutes:
Assisted preparation and production per task (A), minutes:
Review and repair per assisted task (R), minutes:
One-time setup (S), minutes:
Upkeep across the comparison window (K), minutes:

Compare the same work at the same acceptable quality. Use a consistent
measure of effort. Count waiting only when it prevents other useful work;
do not count a minute twice. Include failed attempts and cleanup.

Usual total = N * M
Assisted total = S + K + N * (A + R)
Net time saved = N * (M - A - R) - S - K

If M - A - R is positive, no net time loss begins at the ceiling of
(S + K) / (M - A - R), with at least one repetition.
Strictly positive savings begin at floor((S + K) / (M - A - R)) + 1.
These thresholds hold setup and upkeep fixed. Recalculate if they grow.
If the per-task difference is zero or negative, repeating the workflow
does not create positive time savings at those costs.

Quality check: not yet checked / met / not met
What was checked, against which source or independent example?
What failed or needed repair?
Privacy, money, reliability, enjoyment, and effects on other people:
Uncertainty and what would change the decision:
Decision: keep / adjust / stop / gather more observations
Next review point:

An estimate is not a measured outcome. A before/after observation alone
is not proof that AI caused the change. This is a time comparison, not
financial ROI, a safety assessment, or a quality score.

Nolan Kido | https://nolankido.com/technology/ai-time-savings/