Home / Features / Approval-fatigue management

Human-in-the-loop fails through approval fatigue. Measure it.

If your approvers say yes 97% of the time in under 20 seconds, you don't have human oversight — you have a ritual. Kynara risk-scores every approval and flags the reviewers who've stopped reviewing.

The failure mode nobody instruments

Every agent platform can pause for a human. Almost none asks whether the human is still paying attention. The OWASP AI Exchange #OVERSIGHT control lists approval fatigue — "humans overwhelmed by approval requests, especially if the large majority are okay" — as a principal limitation of human oversight. The EU AI Act (Art. 14) requires oversight to be effective, not just present.

What Kynara measures

SignalTriggerWhat it means
Rubber-stamp risk≥95% approval rate at ≥20 reviewsReviews are likely not meaningful; tighten auto-allow policies so fewer, riskier requests reach humans.
Speed riskMedian review under 30 secondsDecisions faster than the request can be read.
Overloaded≥100 reviews/week per approverUnsustainable volume — add approvers or raise thresholds.
High-risk fast-approvedHigh-risk request approved < 60sThe exact case oversight exists for, waved through.

Risk-scored triage, not FIFO

Every approval request carries a deterministic risk score — monetary size, tainted (untrusted-input) context, high-risk namespaces like payments or infra, bulk markers — so reviewers see high-risk requests first and low-risk noise can be eliminated at the policy layer, where it belongs. Same request, same score: explainable to an auditor line by line, because no LLM is involved in scoring.

The fix for fatigue isn't more approvals — it's fewer, better ones. Kynara's loop: auto-allow the provably safe (e.g. refunds ≤ $100), require approval only above risk thresholds, then watch the fatigue dashboard to verify the remaining reviews are real.

Where it lives

Make your human oversight provably real

Free plan: 3 seats, 10k decisions/month.

Get started freeSequence policiesCompliance evidence →