Decisive AI · Vol. 5 · trust per-task
Trust Calibration Scorecard — batting average on every AI task
Trust is granted per-task, never per-tool. Log hits and misses, track Gut-Logic Deltas, and run quarterly reviews to expand autonomy where performance earns it — or claw it back where it slips.
Open the Trust Calibration Scorecard
Zero-Retention Policy:Your data is processed and purged. We don't train on your decisions.
Trust per-task, never per-tool.
Expand autonomy where trust is earned. Claw back where it isn't.
Trust without a test date isn't trust. It's nostalgia.
| Trend | Cost | Quick log | |||||
|---|---|---|---|---|---|---|---|
| AUGMENT | 2 | 100% | HOLD | → | Med | ||
| DELEGATE | 2 | 50% | HOLD | → | Low |
Export & backup
Runs in your browser — backup JSON so you never lose history.
About this tool
Why vendor trust scores mislead
A model that nails support routing can still butcher pricing recommendations. Per-task calibration is how you stop one good week from licensing bad judgment everywhere.
Expand, hold, claw back
High hit rate on a real sample with low failure cost? Candidate to move down one tier — with audits. Slipping performance or rising stakes? Move up one tier. No drama. Retune, re-test.
Built for the weekly habit
Quick-log hits from the dashboard. Install to your home screen. Export JSON backups. Pair with the Decision Rights Charter Builder so tier moves have a published boundary.