Resolve support cases with multi-agent workflows | ODSP915

Carlos Camacho demonstrates a multi-agent workflow that diagnoses and resolves support cases across partner ecosystems, using the Agentic Partners Integration AI quickstarts. The session shows how agents share context, route intent, orchestrate MCP tools, and produce actionable outputs like Azure web app best-practices guidance, with audit logging via OpenTelemetry.

Overview

This Build 2026 session walks through a demo of an agentic, multi-agent system designed to help diagnose and resolve support cases across partner ecosystems.

What the demo is built on

• Agentic Partners Integration AI quickstarts
• Azure
• Red Hat OpenShift AI running on Azure Red Hat OpenShift (ARO)
• MCP and open-source components

Key concepts shown

Multi-agent collaboration

• Agents share context to work a support case end-to-end.
• A routing agent detects intent and delegates work to a specialized Azure support agent.

MCP-based tool orchestration

• The workflow includes deploying an MCP server from an MCP catalog.
• The system routes and executes MCP tools for new queries.

Output generation for support resolution

• The demo generates an output as a markdown file.
• Example output includes Azure web app best practices.

Observability and audit logging

• Audit logging is implemented using OpenTelemetry to track application events.

Access control / authorization behavior

• A scenario is shown where a user (Josh) lacks department membership, which prevents agents from responding.

Session metadata

• Event: Microsoft Build 2026
• Session code: ODSP915
• Level: Intermediate
• Language: English (US)
• Track: Agents & apps
• Speakers: Carlos Camacho (with an introduction mentioning Sharon from Red Hat)