> For the complete documentation index, see [llms.txt](https://dyns.ntd.asia/power-dynamics/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://dyns.ntd.asia/power-dynamics/artificial-intelligence/study-copilot-studio/mission.md). # Mission The learning plan for the Copilot Studio course — mission, sources, and full session map. Course complete as of Session 8.3 (Groups 1–8). ## Why this, in my own frame Anchor: [What Is AI, Really? — Thinking About the Goal of "Human-Like"](https://dyns.ntd.asia/power-dynamics/artificial-intelligence/what-is-ai-really) That piece argues knowledge alone "makes you a library," not something human-like. Skill is what sits between wanting something and having it, built through repetition, failure, and watching whether something you built actually holds up. It names agentic AI — systems that act, not just answer — as the current attempt at that second half. This course is built on that premise. It doesn't stop at "the agent can answer questions" (the library). It pushes through to "the agent can act, decide, and coordinate with other agents in a live system" (the skill) — and every session ends with something built and tested, not just explained. {% hint style="success" %} **Course complete.** With Session 8.3 (the capstone), all eight groups of the original learning plan are taught. The Northwind Outfitters agent now runs conversational, autonomous, and multi-agent mechanisms together, wrapped in the governance layer from Group 7 and tested by the evaluation surfaces from Group 8. A possible Group 9 (exam-readiness / certification depth) remains unsized — see the Session estimate section below. {% endhint %} ## Practice scenario All hands-on exercises build one running example: a customer-facing support agent for a fictional retailer, **Northwind Outfitters** — order status, FAQs, returns and exchanges. It naturally exercises knowledge grounding (FAQs), tools and actions (order lookup), generative answers, autonomous triggers (proactive return follow-ups), and multi-agent handoff (billing vs. product support) — the same shape as the enterprise scenarios in Microsoft's own certification material. As of 6.5, that handoff spans five live connections: two connected Copilot Studio agents, a Foundry agent, an A2A agent, and a Fabric data agent, tested together. **8.3 wired all of it — the conversational core, the autonomous return-follow-up trigger, and the five-agent handoff — into one cross-group test case for the first time**, and re-confirmed, via a live re-fetch rather than memory, that three previously-flagged gaps (Fabric data agent evaluation, multi-agent visibility in the Activity page, DLP's reach into external agent connections) are all still open. ## Primary sources * Official course skeleton (launches 9/18/2026, Intermediate): [AB-620T00-A course page](https://learn.microsoft.com/en-us/training/courses/ab-620t00) * Exam skills outline (usable now, same domains as the course): [Study guide for Exam AB-620](https://learn.microsoft.com/en-us/credentials/certifications/resources/study-guides/ab-620) * Certification overview: [Microsoft Certified: AI Agent Builder Associate](https://learn.microsoft.com/en-us/credentials/certifications/ai-agent-builder-associate/) * Foundational learning paths filling the beginner gap AB-620 assumes: [Create and publish agents](https://learn.microsoft.com/training/paths/work-power-virtual-agents/) · [Create agents in Copilot Studio](https://learn.microsoft.com/training/paths/create-extend-custom-copilots-microsoft-copilot-studio/) · [Agent in a Day](https://learn.microsoft.com/training/paths/agents-online-workshop/) · [Extend Microsoft 365 Copilot](https://learn.microsoft.com/training/paths/extend-microsoft-365-copilot-studio/) * Product docs (verified per lesson): [Copilot Studio documentation](https://learn.microsoft.com/microsoft-copilot-studio/) ## Session estimate **\~28–30 sessions** was the running estimate once Group 8 was sized (before 8.1 ran) — 3 sessions instead of the original plan's 2, since researching 8.1 alone surfaced seven distinct test methods and a genuine harness-shaped fork in the evaluation surface itself, dense enough to earn its own session split from conversational evaluation, automation, and the capstone build. **With 8.3 taught, the plan lands at 29 sessions total (1.1–1.2, 2.1–2.4, 3.1–3.3, 4.1–4.5, 5.1–5.3, 6.1–6.5, 7.1–7.4, 8.1–8.3) — every group sized and closed as it was reached, per the standing Session-Sizing Engine rule that fired identically before every group since Group 2.** Group-by-group sizing history: Group 2 added one spaced-review session over its original plan (2.4); Group 3 added one grounding-and-citations session (3.2); Group 4 added one closing review/build session (4.5); Group 5 added none — the only group whose original 3-session split held with no change; Group 6 added one closing consolidation session (6.5) alongside a genuine scope correction (six addable agent types, not four or five); Group 7 was sized at 4 sessions, up from 2, when identity/security and responsible AI turned out to be genuinely separate mental models bundled into one session name; Group 8 was sized at 3 sessions, up from 2, for the reason above. AB-620 itself is a 3-day intermediate course that assumes hands-on agent-building experience already in place. Groups 1–2 (6 sessions) closed that gap. From Group 3 on, sessions mapped directly onto AB-620's three skill domains — Plan & configure (30–35%), Integrate & extend (40–45%), Test & manage (20–25%) — broken into single-concept lessons small enough to verify and practice rather than compressed into 24 hours of lecture. **A Group 9 (exam-readiness / certification depth beyond working proficiency) remains a possible, not-yet-sized extension** — named as an option back in the original session-estimate reasoning, for anyone who wants to go from working proficiency to full exam readiness for the *Microsoft Certified: AI Agent Builder Associate* credential. ## Session map | Group | # | Session | Status | | --------------------------------------------------- | --- | --------------------------------------------------------------------- | -------- | | 1 — Foundations & Orientation (complete) | 1.1 | What Copilot Studio actually is | ✅ Taught | | | 1.2 | How an agent decides | ✅ Taught | | 2 — Core Building Blocks (complete) | 2.1 | Topics & the authoring canvas | ✅ Taught | | | 2.2 | Entities and variables | ✅ Taught | | | 2.3 | Generative answers & instructions | ✅ Taught | | | 2.4 | Review & retrieval | ✅ Taught | | 3 — Knowledge & Grounding (complete) | 3.1 | Knowledge sources fundamentals (files, SharePoint, websites) | ✅ Taught | | | 3.2 | Grounding mechanics & citations | ✅ Taught | | | 3.3 | Azure AI Search & custom knowledge sources | ✅ Taught | | 4 — Tools & Actions (complete) | 4.1 | Connectors & Power Platform actions | ✅ Taught | | | 4.2 | Agent flows: human-in-the-loop & error handling | ✅ Taught | | | 4.3 | Prompts and REST APIs: off the connector path | ✅ Taught | | | 4.4 | Adaptive cards & response formatting | ✅ Taught | | | 4.5 | Review & build (Northwind tool chain) | ✅ Taught | | 5 — Agentic Behavior: Autonomous Agents (complete) | 5.1 | Triggers, boundaries, and guardrails | ✅ Taught | | | 5.2 | Build & deploy an autonomous agent | ✅ Taught | | | 5.3 | Monitoring and governing autonomy | ✅ Taught | | 6 — Multi-Agent Orchestration (complete) | 6.1 | Multi-agent design patterns & best practices | ✅ Taught | | | 6.2 | Building child & connected agents (Northwind Billing/Product-Support) | ✅ Taught | | | 6.3 | Integrating a Foundry agent (preview) | ✅ Taught | | | 6.4 | Agent2Agent (A2A) protocol & Fabric data agents | ✅ Taught | | | 6.5 | Review & build: Northwind multi-agent handoff | ✅ Taught | | 7 — Enterprise Integration & Governance (complete) | 7.1 | Identity and Security | ✅ Taught | | | 7.2 | Responsible AI | ✅ Taught | | | 7.3 | ALM Foundations: solutions & environment variables | ✅ Taught | | | 7.4 | ALM Automation: pipelines & CI/CD | ✅ Taught | | 8 — Testing & Capstone (complete — course complete) | 8.1 | Evaluating agents: test sets & evaluation methods | ✅ Taught | | | 8.2 | Conversational evaluation, automation & the GitHub Copilot harness | ✅ Taught | | | 8.3 | Capstone: the full Northwind Outfitters agent | ✅ Taught |