Insights / Case studies / Automotive & Mobility
A global automotive OEMAgentic warranty triage, from pilot to production.
INDUSTRYAutomotive & Mobility
CAPABILITIESAgentic AI · Data & AI Platforms · Governance
ENGAGEMENTDelivery partner + managed pod
DURATION7 months to production
01 — Challenge
Warranty claims were triaged by hand across dozens of dealer and quality systems, delaying root-cause detection.
An internal pilot had shown a language model could classify claim narratives accurately. It had not shown how to connect to production systems, who would approve a supplier chargeback, or how to keep the model honest as part numbers changed.
02 — Approach
01Assess the workflowMapped the claim path end to end, from dealer submission to supplier recovery, and ranked the steps by value and risk.
02Architect for controlDefined risk tiers: agents classify and cluster autonomously; chargebacks above a threshold require a quality engineer’s approval.
03Build on the data platformIntegrated claims, parts, and supplier data on Databricks, with agents orchestrated on Azure and connected to SAP.
04Staff and runA managed pod in Detroit and Pune owns evaluation, monitoring, and monthly model updates.
03 — Tech stack
Microsoft AzureAzure OpenAIDatabricksUnity CatalogSAP S/4HANAServiceNowPythonLangGraphOpenTelemetry
04 — Outcomes
40%faster claim resolution
21days earlier supplier-defect detection
65%of claims triaged without manual touch
05 — Related work
Financial ServicesA top-10 US bankA governance operating model, an LLMOps platform, and a managed delivery pod to run it.
12use cases in production in 9 months
Aerospace & DefenseA Fortune 500 aerospace manufacturerA retrieval-based assistant for MRO technicians on a secure cloud tenant, built and staffed by a dedicated pod.35%reduction in time-to-diagnosis
Your program
Have a pilot that should be in production?
Tell us where it stalled. We’ll tell you what it would take.