AI Skills

Detect Claim Anomaly

Warranty claim volumes make it difficult to catch irregularities through manual review alone. The Detect Claim Anomaly skill brings AI-powered inspection into the warranty claims workflow, invoking the Anomaly Detection Rule store app to flag claims that require further investigation before adjudication.

The skill analyzes incoming claims against historical data, supporting documentation, and images to surface anomalies across eight detection categories:

Duplicate claim detection — Identifies claims already submitted for the same repair or service event, preventing duplicate payouts.
Parts mismatch detection — Flags claims where listed parts do not match supporting documentation or attached images, catching transcription errors or inflated submissions.
Image reuse detection — Detects claims that include a photograph previously submitted with another claim, a common indicator of fraudulent or copy-paste submissions.
Dealer pattern detection — Identifies clusters of similar claims submitted by the same dealer within a specified period, surfacing unusual submission behavior.
Cross-asset customer claim detection — Flags similar claims submitted for the same customer across multiple assets, highlighting potential systemic issues or coordinated claim activity.
Product usage mismatch detection — Compares product usage data on the current claim against prior claim history for the same asset, catching inconsistencies in reported mileage, hours, or operating conditions.
Threshold mismatch detection — Detects repair claims that exceed threshold limits specified for the causal parts.
Inflated claim detection — Detects claim amounts that are significantly higher than the average for similar claims.

Enhance Non-Conformance Description

Unstructured or incomplete issue descriptions are a leading cause of clarification delays in quality workflows. This skill evaluates free-text issue descriptions submitted by field teams and service partners, highlights missing details, and suggests improved versions for clarity and completeness — ensuring non-conformances are actionable from the point of submission.

Recommend Quality Assignment and Actions

Quality teams managing high volumes of non-conformances face pressure to triage quickly and act consistently. This skill helps by intelligently triaging incoming NCs, recommending appropriate NC tasks, and suggesting correction and containment actions — drawing on historical data and current NC context to improve consistency, reduce manual effort, and accelerate resolution timelines.

AI Agents

Create Recall Corrective Actions Agent

Recall campaigns often begin with dense technical documents — repair instructions, service bulletins, engineering notices — that must be manually reviewed and translated into structured corrective actions. This agent automates that process.

The agent analyzes uploaded documents, extracts the required corrective actions, and identifies the associated parts and labor for each action. It then creates the corrective action records and their corresponding charges directly within the recall campaign, reducing manual effort and minimizing transcription errors.

Capabilities:

Analyzes repair instructions, bulletins, and related documents to extract corrective action details
Identifies required parts and labor associated with each corrective action
Automatically creates corrective action records and corrective action charges within the campaign

Other Capabilities

Plan and Execute Recall Campaign Phases

Planning a recall rollout across thousands of impacted assets — while accounting for parts constraints and servicing capacity — is one of the most complex tasks a recall manager faces. This agent turns that process from weeks of manual planning into an AI-driven recommendation.

The agent evaluates the full picture: impacted assets added to the campaign, corrective actions (whether AI-generated or manually created), calculated parts requirements based on corrective action specifications and relevant asset counts, and parts availability data uploaded by the recall manager. Using this information, the agent determines which assets can be serviced and when, then structures the campaign into phases and sub-phases accordingly.

Capabilities:

Analyzes impacted assets, corrective actions, parts requirements, and parts availability to build an informed rollout plan
Determines asset servicing readiness based on parts supply against demand
Generates campaign phases and sub-phases that align execution to real-world parts constraints

Auto-categorize Repair Claims

High claim volumes make consistent, manual adjudication decisions difficult to sustain at scale. This agent automatically evaluates incoming warranty claims and applies configurable business rules to categorize each claim — as Approved, Partially Approved, Send Back to Dealer, Potential Fraud, or Rejected — while cross-validating claim details to catch discrepancies before adjudication.

Capabilities:

Applies configurable categorization rules to incoming warranty claims, delivering consistent adjudication decisions at scale
Cross-validates claim details to surface discrepancies and reduce back-and-forth with dealers
Reduces manual review workload for Claims Agents, freeing them to focus on exceptions and edge cases

GenAI Summary

Summarize Repair Claim

Claims agents review dozens of claims each day — evaluating repair details, causal parts, labor and service charges, and any prior back-and-forth before making an adjudication decision. This skill generates an AI-powered summary of the repair claim, giving agents a clear, structured snapshot of the claim so they can make faster, more informed decisions without reading through every detail manually.

The summary covers the full disposition picture of the claim:

Approved items — Lists what was approved and the corresponding approved amount.
Partially approved items — Lists what was partially approved and the approved amount.
Sent back and resolved items — Captures what was initially sent back to the dealer, what was subsequently modified, and what was ultimately approved.
Rejected items — Lists what was rejected and the corresponding rejected amount.
Claim totals — Provides the total claimed amount, total approved, and total rejected in a single view.

The summary is surfaced directly within the claim record and can be refreshed as claim details are updated, ensuring agents always have an accurate, up-to-date picture of the claim at a glance.