The problem
Organizations have different tolerances for automated relationship updates and different capacities for reviewing proposed graph changes.
What EKOS does
Administrators set minimum-persist and auto-accept thresholds, inspect the effective values after defaults, and return to platform defaults by clearing an override.
- Minimum-persist and auto-accept thresholds
- Tenant overrides with visible effective values
- Validation of the discard, review, and accept range
- Reversible return to platform defaults
- Mandatory review for high-risk relationship types
How it works
The minimum-persist threshold sets the floor below which a weak relationship proposal is discarded outright.
The auto-accept threshold sets the point above which a sufficiently confident, lower-risk proposal can enter the graph under policy.
Everything between the two thresholds goes to the Review Queue for a human decision.
Administrators see tenant overrides beside the effective values, and clearing an override returns the tenant to the platform default.
Validation keeps the discard, review, and accept ranges coherent, and conflict or supersession proposals require review regardless of either value.
Business outcome
Each tenant can make the tradeoff between review volume and policy-based acceptance explicit, observable, and reversible.
Proof and human control
The settings preserve a valid discard, review, and accept range, while conflict and supersession proposals remain mandatory-review items regardless of threshold.
Example in practice
A tenant lowers its auto-accept threshold after reviewing queue quality, reducing routine review volume while keeping conflict and supersession decisions in the manual path.
Scope today
These settings govern relationship review thresholds for later analysis. They are not a broad AI governance console and do not retroactively reclassify earlier proposals.