AB-100: Agentic AI Business Solutions Architect

O'Reilly Live Learning Course | 4 Hours | Microsoft AB-100 Exam Prep + Real-World Agentic AI

Design, deploy, and govern multi-agent AI business solutions across the Microsoft stack -- and walk in prepared to pass Microsoft's AB-100 certification exam on the first attempt.

This repo has three sides:

1. Live training plan -- the canonical 4-hour plan and per-hour breakdowns in docs/course-plan-april-2026.md, plus the built slide deck in docs/ab100-course-deck-april-2026.pptx.
2. Hour 3 Deploy POC -- a working Foundry agent + MCP + ACA + APIM stack in src/. Five bash scripts spin it up, one tears it down. See src/README.md.
3. AB-100 Cert Buddy -- a GitHub Copilot agent in .github/ that generates exam-realistic practice questions, short architect-grade labs, and personalized study plans, all grounded in Microsoft Learn via MCP. Open the repo in VS Code and invoke @ab100-cert-buddy-agent in Copilot Chat. See the Cert Buddy quick start below.

Course overview

Hour	Theme	AB-100 Domain
1	Plan AI Solutions	Plan AI-powered business solutions (25-30%)
2	Design AI Solutions	Design AI-powered business solutions (25-30%)
3	Deploy AI Solutions	Deploy -- implementation & operations (~20-25%)
4	Security & Exam Mastery	Deploy -- security & governance (~20-25%)

Prerequisites

• A Microsoft 365 tenant with Microsoft 365 Copilot and Microsoft Copilot Studio licensing (a developer tenant or trial works for the demos).
• A Microsoft Power Platform environment for agent authoring and a Microsoft Foundry project for the model and Foundry Tools demos.
• Working knowledge of one Microsoft business application stack -- Dynamics 365 customer experience and service, Dynamics 365 finance and supply chain, or Microsoft Power Platform -- consistent with the AB-100 prerequisite associate certifications.
• Comfort with prompt engineering basics, ALM concepts, and baseline cloud security vocabulary (identity, RBAC, data residency, audit logging).

Full prerequisite detail: docs/course-plan-april-2026.md.

Repository structure

ab100/
├── README.md                 # This file
├── CLAUDE.md                 # Claude Code instructions
├── CODE_OF_CONDUCT.md        # Community guidelines
├── contributing.md           # Contribution guide
├── SECURITY.md               # Security policy
├── LICENSE                   # MIT License
├── .markdownlint.json        # Markdown linting config
├── .editorconfig             # Editor formatting consistency
│
├── .github/                  # Cert Buddy agent + skills + prompts
│   ├── agents/
│   │   └── ab100-cert-buddy-agent.agent.md
│   ├── skills/
│   │   ├── ab100-item-creator/SKILL.md
│   │   ├── ab100-lab-creator/SKILL.md
│   │   └── ab100-study-planner/SKILL.md
│   ├── prompts/
│   │   ├── ab100-quiz.prompt.md
│   │   ├── ab100-lab.prompt.md
│   │   └── ab100-plan.prompt.md
│   ├── copilot-instructions.md
│   └── workflows/validate.yml
│
├── .vscode/                  # MCP server + recommended extensions
│   ├── mcp.json
│   └── extensions.json
│
├── docs/                                   # Course plan, exam reference, slide deck
│   ├── course-plan-april-2026.md           # Canonical 4-hour plan
│   ├── ab100-exam-objectives.md            # Verbatim Microsoft Learn skills measured
│   ├── ab100-portals-and-products.md       # Architect-grade portal/product index
│   ├── ab100-metrics-dashboards-reports.md # Highest-yield dashboards and reports
│   ├── ab100-study-resources.md            # Curated public AB-100 content
│   ├── microsoft-certification-policies.md
│   ├── pearson-vue-registration.md
│   ├── ab100-poc-architecture.html         # Standalone POC architecture diagram
│   ├── ab100-course-deck-april-2026.pptx   # Built slide deck (canonical artifact)
│   ├── deck-build/                         # HTML slide sources + Node build pipeline
│   └── learning-paths/architect-agentic-ai/  # Local mirror of MS Learn path (11 modules)
│
├── images/cover.png          # Course cover image
└── src/                      # Hour 3 Deploy POC -- Foundry agent + MCP + ACA + APIM
    ├── README.md             # Five-command demo + teardown
    ├── app/                  # Python MCP server (FastMCP) wrapping a Foundry agent
    ├── infra/                # Modular Bicep (UAMI, Logs, ACR, ACA, APIM)
    ├── policies/             # APIM inbound policy
    └── scripts/              # 00-login -> 99-teardown bash demo flow

Quick start

1. Clone the repository

   git clone https://github.com/timothywarner-org/ab100.git
   

2. Read the course plan: docs/course-plan-april-2026.md -- per-hour learning objectives, teaching flow, demos, exercises, and Microsoft Learn module mapping.

3. Open the slide deck: docs/ab100-course-deck-april-2026.pptx.

4. Run the Hour 3 POC: src/README.md -- five bash scripts spin up Foundry agent + MCP + ACA + APIM in your subscription.

5. Use the Cert Buddy: open the repo in VS Code and invoke @ab100-cert-buddy-agent in Copilot Chat.

Hour summaries

Hour 1: Plan AI Solutions

Requirements analysis, AI strategy on the Cloud Adoption Framework, multi-agent platform selection across Microsoft 365 Copilot, Microsoft Copilot Studio, and Microsoft Foundry, plus ROI and build-buy-extend analysis. Maps to AB-100 Domain 1 (25-30%). See Hour 1 in the course plan.

Hour 2: Design AI Solutions

Task, autonomous, and prompt-and-response agents, Microsoft Copilot Studio topics and agent flows, Microsoft Foundry Tools, MCP and A2A extensibility, Computer Use, voice mode, and Microsoft Power Platform Well-Architected Framework alignment for intelligent workloads. Maps to AB-100 Domain 2 (25-30%). See Hour 2 in the course plan.

Hour 3: Deploy AI Solutions

Monitoring and telemetry interpretation, agent testing strategy, and ALM for agents, connectors, custom models, and Dynamics 365 AI. Live demo materializes as the Foundry-agent-as-MCP-server stack in src/. Maps to AB-100 Domain 3 implementation & operations subset (~20-25%).

Hour 4: Security & Exam Mastery

AI security, governance, prompt manipulation defense, data residency, audit trails, Responsible AI review, plus AB-100 exam logistics, item formats, and final-week study plan. Builds on the Hour 3 POC by layering validate-jwt, named-values-from-Key-Vault, azure-openai-token-limit, and Defender for APIs on the same artifact. Maps to AB-100 Domain 3 security & governance subset (~20-25%) plus exam readiness.

Course schedule

Time	Activity
0:00 -- 0:50	Hour 1 -- Plan AI Solutions
0:50 -- 1:00	Q&A + Break
1:00 -- 1:50	Hour 2 -- Design AI Solutions
1:50 -- 2:00	Q&A + Break
2:00 -- 2:50	Hour 3 -- Deploy AI Solutions
2:50 -- 3:00	Q&A + Break
3:00 -- 3:50	Hour 4 -- Security & Exam Mastery
3:50 -- 4:00	Wrap-up, resources, next steps

Full plan in docs/course-plan-april-2026.md.

Microsoft Learn path companion

This course is sequenced to align with the Microsoft Learn path Architect AI solutions for business productivity. The path's 11 modules are mirrored locally with per-module summaries and AB-100 domain mapping at docs/learning-paths/architect-agentic-ai/. Drill in between live sessions to deepen any topic introduced during an hour.

Instructor

Tim Warner -- Microsoft MVP (Azure AI and Cloud/Datacenter Management), Microsoft Certified Trainer

• LinkedIn
• Website
• O'Reilly Author Page

AB-100 Cert Buddy

An AI-powered study companion for the Microsoft AB-100: Agentic AI Business Solutions Architect certification, built entirely on GitHub Copilot agents. The buddy lives in .github/ as agent definitions, skill specs, prompt templates, and an MCP server configuration that turns GitHub Copilot Chat into an interactive exam coach.

The agent generates exam-realistic practice questions with two-phase interactive evaluation, short hands-on labs (15-25 minutes) across agent authoring, design walkthroughs, ALM, and governance, and personalized study plans based on your self-assessed confidence across the three AB-100 domains. All content is original and grounded in Microsoft Learn via MCP.

Cert Buddy prerequisites

Requirement	Purpose
VS Code	IDE with Copilot Chat support
GitHub Copilot Chat extension	Runs the agent and skills inside VS Code
Microsoft 365 tenant	Required for Copilot Studio and Microsoft 365 Copilot labs
Copilot Studio trial	Start a trial -- required for hands-on Copilot Studio labs
Azure subscription (optional)	Only needed for Microsoft Foundry labs that create Azure resources

No API keys, Node.js, or Python are required. The Microsoft Learn MCP server is a free hosted service with no sign-up.

Cert Buddy quick start

1. Open the folder in VS Code:

   code ab100
   

2. The Microsoft Learn MCP server is defined in .vscode/mcp.json and auto-configures when the workspace loads.

3. Open GitHub Copilot Chat and invoke the agent:

   @ab100-cert-buddy-agent Generate 5 practice questions on Copilot Studio extensibility with MCP
   

   Or use the slash commands. Type / in Copilot Chat:

   /ab100-quiz
   /ab100-lab
   /ab100-plan
   

Cert Buddy skills

Skill	File	What it does
ab100-item-creator	.github/skills/ab100-item-creator/SKILL.md	Exam-realistic AB-100 practice questions with two-phase interactive delivery (question -> wait -> evaluate).
ab100-lab-creator	.github/skills/ab100-lab-creator/SKILL.md	15-25 minute self-validating labs across agent authoring, design walkthroughs, ALM, and governance.
ab100-study-planner	.github/skills/ab100-study-planner/SKILL.md	Personalized study plans based on confidence across the three AB-100 domains, with real Microsoft Learn module links.

Cert Buddy MCP server

Server ID	Technology	Purpose
ab100buddy-mslearn	https://learn.microsoft.com/api/mcp (HTTP)	Free Microsoft Learn MCP server -- no API keys, no sign-up. Provides microsoft_docs_search, microsoft_docs_fetch, and microsoft_code_sample_search. Grounds all cert-buddy content in official Microsoft documentation.

Cert Buddy key rules

• Grounded in Microsoft Learn. Every question and lab is grounded in official Learn docs via MCP before any other source.
• Current terminology only. Legacy names like "Azure AI Studio" and "Power Virtual Agents" are silently replaced with "Microsoft Foundry" and "Microsoft Copilot Studio." See the full rename table in .github/copilot-instructions.md.
• Original content only. The agent never recreates, paraphrases, or references real exam questions, braindumps, or leaked content.
• No contractions. No trick wording. Distractors must be real.
• Answer randomization. Correct answer position is distributed across A/B/C/D.
• Company randomization. Scenarios draw from the WWL-approved Fictitious Names List embedded directly in the agent and skill files -- always the entire company name, never just Contoso.
• Labs include cleanup. Every lab ends with exact rollback or deletion steps.
• Microsoft style. Items follow the Microsoft Worldwide Learning Exam Writing Style Guide; prose follows the Microsoft Writing Style Guide. Both rule sets are inlined into the agent and skill files.

Disclaimer

This is an unofficial study guide. Always refer to the official Microsoft exam page for the most current information.

License

MIT License. See LICENSE for details.

---

Questions? Open an issue or reach out via the course chat during live sessions.