AI Agents

Intelligent Incident Routing Agent

Ford Motor Company — GCP · Google ADK · ServiceNow · Microsoft Teams

150–250 tickets/week routed automatically across 4 business verticals

200+Tickets routed per weekAcross Manufacturing, Supply Chain, Quality, Finance

ZeroMissed SLAs since deploymentPreviously a recurring risk

10xFaster mean time to resolutionFrom days → minutes to hours

4Business verticals coveredSingle agent, unified routing logic

The Problem

Ford's data platform team supports four major business verticals — Manufacturing, Supply Chain, Quality, and Finance — each generating a continuous stream of ServiceNow tickets: incidents, change requests, problems, and demand requests. The combined volume reached 150–250 tickets per week.

The existing process required someone to manually monitor the team queue at all times, read each ticket, identify the right owner based on product and domain knowledge, and manually assign it. Response times ranged from a full day for simple issues to ten days for complex ones. SLA breaches were a recurring risk, and the cost of a missed ticket could cascade into supplier disruptions, downstream dashboard failures, and reporting outages affecting critical business decisions.

The Solution

We deployed an agentic AI system built with Google ADK, hosted on Cloud Run, and integrated with both ServiceNow and Microsoft Teams via Power Automate and Apigee.

When a new ticket arrives in ServiceNow, the agent is triggered via API. It parses the ticket content, enriches context by querying internal knowledge about products, data sources, and ownership mappings, and generates three ranked remediation recommendations tailored to the specific issue. The agent then sends a targeted ping directly in Microsoft Teams to the engineer who owns the affected system — not a broadcast to the whole team.

For complex issues where automated recommendations are insufficient, the assigned engineer can initiate a live chat session directly with the agent from within Teams, working through the problem collaboratively with AI assistance in real time.

System Flow

→01Ticket ArrivesServiceNow incident / change / problem

→02Context EnrichmentAgent queries ownership maps, source metadata

→03AnalysisRoot cause inference, priority classification

→04Recommendations3 ranked remediation steps generated

→05Teams NotificationDirect ping to exact product owner

06Live Chat (optional)Engineer chats with agent to resolve complex issues

Tech Stack

GCP Cloud Run

Vertex AI

Docker

Python

Google ADK

ServiceNow API

Microsoft Teams

Power Automate

Apigee

Cloud Scheduler

Key Engineering Challenges

Intelligent ownership routing

Routing a ticket to the exact right engineer — not just the right team — required building and maintaining a dynamic ownership graph mapping products, data sources, and pipelines to individual engineers. The agent needed to reason over this graph contextually, accounting for on-call rotations and escalation paths, with high enough confidence to act autonomously.

Apigee + Cloud Run integration

Publishing the agent as an org-accessible endpoint through Ford's internal Apigee gateway involved navigating complex API security policies, custom auth flows, and network topology constraints specific to the enterprise environment. Getting Cloud Run to serve reliably behind Apigee under variable load required careful configuration of timeout, concurrency, and health check settings.

Teams-native chat interface

Building a conversational interface directly inside Microsoft Teams — rather than routing users to an external tool — required Power Automate flows that could maintain session state across a multi-turn conversation. The challenge was preserving context between messages so the agent could reason over a full diagnostic thread, not just respond to isolated prompts.