🤖 AI Agent Runbooks

Welcome to AI Agent Runbooks — your go-to hub for building and deploying production-ready AI Agent solutions on Microsoft platforms. This repository provides structured runbooks, proven technical patterns, and real-world scenarios to accelerate your AI Agent delivery with confidence.

What is AI Agent Runbooks?

AI Agent Runbooks is an open, community-driven repository that provides hands-on implementation guides, validated technical patterns, and ready-to-use scenario blueprints for building enterprise-grade AI Agents using Microsoft technologies — including Copilot Studio, Azure AI Foundry, and Microsoft 365 Copilot.

Whether you're building your first agent or scaling across multiple business domains, this repository gives you the structure and guidance to deliver faster and smarter.

Repository Structure

📦 ai-agent-runbooks
├── 📂 00-overview/          → Understand the repository objectives, structure, and how to get started
├── 📂 01-scenarios/         → Explore pre-built AI Agent use cases and end-to-end implementations
├── 📂 02-patterns/          → Learn technical patterns and industry best practices
└── 📂 03-references/        → Access additional resources, templates, and reference materials

How to use this repository

Purpose	Description	Starting Path
📋 Overview	Understand the repository objectives, agent taxonomy, and navigation guide	00-overview/
🎯 Agent Scenarios	Explore pre-built use cases such as HR Onboarding Agent, Helpdesk Agent, and more	01-scenarios/
🏗️ Technical Patterns	Learn implementation patterns for Copilot Studio, Azure AI, and multi-agent orchestration	02-patterns/
📚 References	Access templates, guides, architecture references, and additional resources	03-references/

🚀 Quick Start

New here? Start with 00-overview/ to understand the structure and agent taxonomy.

Ready to build? Jump to 01-scenarios/ and explore real-world agent implementations like:

• 🤖 Autonomous Invoice Orchestration Agent — Autonomous agent that helps employees create payment request forms and get them approved by their manager
• 🧑‍💼 HR Onboarding Agent — Autonomous agent that helps new hires find HR-related information using a ServiceNow Knowledge Base
• 🔧 Dynamics 365 Monitoring Agent — AI-powered monitoring and diagnostics agent for D365 Finance & Supply Chain using Application Insights telemetry
• 🛒 Dynamics 365 PO Generation Agent — Autonomous agent that creates a purchase order autonomously or manually
• 🏢 Workplace Agent — Declarative agent that handles internal FAQ inquiries and application request workflows using SharePoint, Dataverse, and Power Automate
• 🧾 F&O User Onboarding Agent — Autonomous onboarding solution that automates user provisioning in Microsoft Entra ID and Dynamics 365 Finance & Operations, including approval workflows and role assignment.
• 🧩 M365 Agent — Collection of pre-built, deploy-ready Microsoft 365 Copilot agents (e.g., Plan My Day, Executive Briefing, Request Tracker) designed to accelerate common workplace scenarios and enable rapid customization.
• ⚙️ D365 F&SCM Compare & Copy Configurations Agent — AI-powered agent that compares and copies module configurations across Dynamics 365 Finance & Supply Chain Management environments, helping administrators identify configuration differences, validate settings, and maintain consistency using the Dynamics 365 ERP MCP server.
• 📊 DynPerf Performance Agent — AI-powered performance diagnostics agent that leverages DynamicsPerf data to help administrators analyze SQL query statistics, detect bottlenecks, and optimize Dynamics 365 environment performance with actionable, natural-language insights.
• 📈 ERP-D365-Inventory-Optimization-Agent — Interactive agent that generates reports on Inventory from D365 F&O for inventory managers and other stakeholders to get current information, predictions, and other analytics around their inventory data.
• 🎫 IT Service Desk Insights Agent — Declarative agent grounded on ServiceNow knowledge and catalog content that answers employee how-to questions and performs sentiment and theme analysis over ticket text, with permission-trimmed retrieval.
• 🧑‍💻 Employee Self-Service Agent — One conversational front door in Microsoft 365 Copilot for HR and IT support, grounded on the HRIS, ITSM, and SharePoint. Includes the qualification and readiness work that dominates this delivery.
• 🎟️ Copilot Licence Lifecycle Agent — Copilot Studio agent that manages the Microsoft 365 Copilot licence estate: inventory, dormancy detection, reclaim and reassignment, and waitlist management.
• 💼 CRM Account Planning Cowork Agent — Copilot Cowork connected to the CRM through a purpose-built MCP server, producing branded account plans, QBR decks, and pipeline models as Office artifacts.
• 🤝 Client Meeting Preparation Agent — Pre-meeting briefings and post-meeting file notes for advisers and relationship managers, assembled from the book of record, relationship history, and the client file. Built for regulated advisory work, with hard advice boundaries and information barrier enforcement.
• 📬 Email & Mailbox Triage Agent — Classifies, prioritises, routes, and drafts responses for high-volume shared mailboxes, with a draft-never-send boundary and a ticket handoff. The most frequently requested use case across the CAF portfolio.
• 🗓️ Meeting Intelligence Agent — Extracts decisions, actions, and owners from meeting transcripts under a hard evidence rule, deduplicates across recurring series, and tracks commitments to closure. Includes the consent position and the invented-action release gate.
• ⚖️ Contract & Legal Intelligence Agent — Clause extraction, risk flagging against documented standard positions, tracked-changes redlines, and evidence retrieval — with a hard legal-advice boundary and a qualified human on every decision.
• 📄 Document Processing & Extraction Agent — Classify, extract, validate, and route high-volume documents at production accuracy, with confidence-based human review and output shaped to whatever the downstream system actually accepts.

Need implementation patterns? Check out 02-patterns/ for proven technical approaches including Copilot Studio orchestration, RAG patterns, and multi-agent design.

• 🧠 Enterprise RAG Pattern — Multi-source, governed RAG pattern that grounds LLM responses in organizational data (SharePoint, OneDrive, Dataverse, Azure SQL, Fabric, Dynamics CRM) with document-level access control and full citation trails via Azure AI Search
• 🔀 Agentic Workflow Orchestration — Multi-agent orchestration pattern for coordinating handoffs between specialized agents, enabling sequential, concurrent, and hierarchical agent collaboration across complex business workflows
• 🧭 Declarative vs Custom Engine Agent — Decision framework for selecting between Declarative Agents and Copilot Studio Custom Engine Agents, covering three build methods (Agent Builder, Copilot Studio, Agents Toolkit), knowledge scale matrices, reference architectures, use-case scenario library, performance comparisons, and step-by-step implementation runbooks for both agent types and multi-agent composition
• 🔌 Copilot Connector Knowledge Onboarding — Getting third-party enterprise content (ServiceNow, Jira, Confluence, Salesforce, Workday) into the Microsoft Graph index with correct ACLs, including the permission validation evidence that unblocks security review
• 🛡️ Agent Governance & Rollout Control Plane — Separating agent creation from consumption, controlling sharing, establishing an approval route, and building the inventory and audit position that enterprise rollouts depend on
• 💳 Copilot Credits & Cost Control — Credit consumption rates, what enforcement actually stops, capacity allocation, per-agent caps, and chargeback design
• 🔬 Grounding & Response Quality Remediation — An ordered diagnostic ladder for agents that are built, connected, and still answering badly
• 🚢 Agent Publishing & Channel Deployment — The last mile: licensing prerequisites, channel-specific behaviour, authentication, and rendering across Microsoft 365 Copilot, Teams, SharePoint, and third-party channels
• 🔌 MCP Server Integration — Connecting third-party systems (Salesforce, Workday, ServiceNow, Databricks, Snowflake, GitHub) to Copilot Cowork, declarative agents, Copilot Studio, and Foundry via Model Context Protocol servers, including tool design, per-user identity, dynamic tool discovery, and gateway-mediated topologies for on-premises and private-network backends (APIM, Apigee, Workato, MuleSoft)
• 🎨 Branded Office Artifact Generation — Making agent-generated PowerPoint, Word, and Excel output brand-compliant, factually reliable, and safe to send, using the Organization Asset Library and an entry-point × model test matrix
• 🔄 Copilot Studio Migration & Modernisation — Moving an existing conversational or automation estate onto Copilot Studio: inventory and disposition, the pre-requisite work packages that stall most migrations, parallel-run cut-over, and decommissioning
• 🧬 Copilot Studio & Foundry Split Architecture — Deciding what belongs in Copilot Studio and what belongs in Foundry, how the halves communicate, and the ownership, cost, and latency consequences of the boundary
• 📑 Intelligent Document Processing Pipeline — Classify → extract → validate → route → output: extraction stack selection, confidence thresholds derived from measured data, validation against master data, and the downstream boundary when the target system has no API
• ✋ Human-in-the-Loop Review & Approval — Designing the gate between an agent's output and a consequence: gate placement, what the reviewer must be shown, confidence-based routing, review fatigue, and the audit record

Need reference material? 03-references/ holds the curated documentation index and the field-observed constraints that shape design.

• 📚 Agent Delivery Reference Library — Microsoft and non-Microsoft references organised by delivery question rather than by product
• ⚠️ Known Limitations & Workarounds — constraints observed repeatedly in the field, with the design responses that work

Key Features

🤖 Agent Runbooks – Step-by-step implementation guides for real-world AI Agent scenarios
🎯 Pre-built Scenarios – Ready-to-deploy use cases across HR, IT, Customer Service, and more
🏗️ Technical Patterns – Validated design patterns for Copilot Studio and Azure AI Foundry
🔗 Copilot Studio Native – First-class support for Microsoft Copilot Studio agent development
⚡ Accelerated Delivery – Structured runbooks to reduce time-to-production
📖 Comprehensive Guidance – End-to-end documentation from design to deployment

Support & Documentation

• 📖 Full documentation available in each section's README
• 💡 Scenario runbooks with step-by-step implementation instructions
• 🔧 Technical patterns with architecture diagrams and code samples
• 🎓 Best practices and lessons learned from real-world deployments

📣 Feedback & Issues

Found a bug? Have a question? We'd love to hear from you — no GitHub experience needed!

	Action	Description
🐛	Report a Bug	Something not working as expected? A broken link, incorrect step, or outdated information? Let us know and we'll look into it.
💬	Ask a Question	Not sure how something works? Ask the community and our team will help.

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA), which confirms that you have the right to, and do, grant us the rights to use your contribution. For more details, visit Contributor License Agreements.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and will update the PR accordingly (for example, by adding a status check or comment). Simply follow the instructions provided by the bot. You only need to complete this process once for all repositories that use our CLA.

This project has adopted the Microsoft Open Source Code of Conduct.

For more information, see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos must comply with Microsoft's Trademark & Brand Guidelines.

Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship.

Any use of third-party trademarks or logos is subject to the policies of those third parties.

---