AI & Intelligence
warrantini's AI layer participates directly in claims and rules workflows — triage, assessment, resolution, escalation, rule recommendations, and defect detection.
What the AI layer does
AI participates directly in six decision points across claims and warranty rules. Except for claims triage, each one works in the background as a suggestion engine — a human still makes the call. Details on each below.
Claims Triage
Runs when a claim is submitted. Analyzes the claim and recommends a classification. By default it's suggestion-only. A tenant can opt in to Auto-Triage, which lets the AI automatically approve or reject claims once its confidence clears a threshold the merchant sets — a separate threshold for auto-approve and for auto-reject. Auto-Triage is off by default.
Assessment Intelligence
Suggests an assessment of the claim — what the AI believes is going on with the product — for a reviewer to consider. On by default; suggestion-only, no automatic action.
Resolution Recommendation
Recommends how a claim should be resolved (repair, replace, refund, and so on). On by default; suggestion-only.
Escalation Intelligence
Runs on a schedule against claims sitting in submitted or in-review status, weighing claim complexity and safety signals. Unlike the other decision points, this one can escalate a claim on its own once it judges escalation warranted — it's a plain on/off toggle rather than a merchant-set confidence threshold, and it's on by default. A hard time limit escalates a claim regardless of the AI's judgment if it's been waiting too long.
Rule Recommendations
Suggests new or adjusted warranty coverage rules based on patterns in claims data. On by default; recommendations are reviewed and accepted or rejected from the Intelligence dashboard, never applied automatically.
Defect Pattern Detection
Looks across claims for clusters and unusual spikes that might indicate a product defect. On by default; surfaces patterns for review and doesn't take action on its own.
The Intelligence dashboard
A separate screen (Dashboard → Intelligence), distinct from the per-claim AI suggestions that show up inside a claim. It has three views:
- Overview — a summary of AI activity.
- Decision log — a browsable record of every AI decision made, across all six decision types.
- Patterns — defect clusters and unusual spikes/anomalies. Rule recommendations can be accepted or rejected from here.
If AI is unavailable
AI failures never block core workflows. Claim and registration operations succeed regardless of whether the AI layer is reachable — triage, assessment, and the rest degrade gracefully rather than holding up the underlying operation.