Understanding Workflow Health
Clean exit rates, unverified completions, and the period-over-period signals that tell you whether your AI coding workflow is improving or degrading.
What workflow health measures
Workflow health is about the quality of the process, not the quality of any individual output. A run can produce correct code and still represent an unhealthy workflow signal if it exited without verification, required manual override, or produced a review burden that was not anticipated.
The report tracks four primary signals: clean exit rate, average session time, unverified exits, and estimated time saved. Together these tell you whether your team's AI workflow is becoming more reliable and efficient over time, or whether friction is accumulating.
Clean exit rate
The most important metric. A clean exit means the run produced a verified receipt: tests passed, scope was confirmed, and the output is ready for the next step without requiring a human review before it can be trusted.
A clean exit rate below 80% is a signal worth investigating. It does not always mean the work is wrong. It often means the validation step is not completing, which may indicate task instructions that are unclear, test coverage that is not adequate, or scope declarations that are too broad.
Clean exit rate is a process quality metric, not a correctness metric. Code can be correct and still fail to exit cleanly if verification is not configured to match the task type.
Average session time
This is the median duration from run start to clean exit. Shorter is not always better: complex tasks should take longer. What you are looking for is stability. If average session time is increasing across comparable task types, something in the process is adding friction.
Common causes of session time increases: context that has to be rebuilt manually on each run (see Context Continuity report), scope that is too broad requiring more validation steps, or task instructions that are not giving Avorelo enough signal to route efficiently.
Unverified exits
These are runs that completed but did not produce a verified receipt. They appear as review needed in the What Got Done report. Each one represents a gap between the work that was done and the confidence that the work was correct.
A small number of unverified exits per period is expected: exploratory runs, draft work, and interrupted sessions will not always produce clean exits. A trend of increasing unverified exits is worth addressing before it accumulates.
Period over period
The trend table shows how each metric changed from the previous period. Up arrows on clean exit rate and time saved are positive. Up arrows on unverified exits and scope drift events are negative. Down arrows on average session time (shorter sessions for comparable work) are positive.
The trend table is more useful than any single period's snapshot. One period with high unverified exits may be a noisy sprint. A consistent trend across three periods is a structural issue.
Open Workflow Health report
View run quality and friction signals for your team