Learning curriculum report · Current as of July 1, 2026
Author: UngerAI

Copilot Studio & Microsoft Foundry: A Beginner's Learning Curriculum

A structured, hands-on path to take any beginner from zero to confident builder: Microsoft Copilot Studio from 100 → 400 (including agent-to-agent / multi-agent orchestration), and Microsoft Foundry from 100 → 200. Every item links to official Microsoft Learn content, with a phased plan, 2026-aligned certifications, and capstone projects.

For beginners Author UngerAI Pace ~16 weeks @ 6 hrs/wk Effort ≈104 hours
Developer building AI agents on a laptop
From zero to multi-agent builder — on official Microsoft Learn rails.

Goals & how this curriculum works

The goal: learn Copilot Studio (agents, and agent-to-agent) from 100 to 300 and 400 level, and Microsoft Foundry from 100 to 200. This plan honors that order — it front-loads the low-code Copilot Studio path, layers in Foundry fundamentals, then climbs into advanced and multi-agent territory. No prior certifications are assumed.

Two tracksCopilot Studio (low-code agents) and Microsoft Foundry (pro-code generative AI), taught in a deliberate order.
Official onlyEvery link points to Microsoft Learn docs, learning paths, or certifications — no third-party detours.
Agent-firstThe through-line is agents: build one, publish it, extend it, then orchestrate many.
Build to learnEach level pairs with a concrete capstone project so the skills stick.

Level legend

Levels describe depth, not difficulty for its own sake. Each builds on the last.

100 Foundations

Concepts, vocabulary, first agent. "What is it and why."

200 Build & publish

Topics, knowledge, actions, testing, shipping to channels.

300 Advanced

Autonomous agents, tools, MCP, agent flows, governance, evaluation, ALM.

400 Expert / multi-agent

Agent-to-agent orchestration, connected & child agents, pro-code SDK, enterprise governance.

Visual Analytics

Confidence scores reflect how directly each claim is supported by official Microsoft sources, plus practical interpretation. They are qualitative, not benchmark results.

Official Learn coverage

96% confidence. Both core tracks are anchored by official 6-module Microsoft Learn paths.

2026 cert alignment

90% confidence. Targets the new AI-901 and AI-103 exams; some dates are beta/planned.

Multi-agent depth

92% confidence. Copilot Studio and Agent Framework both document agent-to-agent orchestration.

Hour estimates (pacing)

75% confidence. Typical ranges; actual pace varies with prior experience.

Estimated effort by phase

Approximate study hours (reading + hands-on labs). Certification prep is interwoven, not separate.

Phase 1 · Foundations
12 h
Phase 2 · CS Build/Publish
18 h
Phase 3 · Foundry Apps
18 h
Phase 4 · CS Advanced
28 h
Phase 5 · CS Multi-agent
28 h

Total ≈ 104 hours over ~16 weeks. Confidence: Medium — hour estimates are typical ranges; actual pace varies with prior experience.

Competency coverage by level

Competency100200300400
No-code agent authoring (topics, knowledge)
Actions, connectors & Power Automate flows
Autonomous agents, tools & MCP
Governance, ALM & evaluation
Agent-to-agent / multi-agent orchestration
Pro-code (Agents SDK, Agent Framework)

Important: 2026 certification refresh

Microsoft is rolling its AI certifications to new, agent-centric exams. This directly affects what you should study toward.

⚠ Two foundational AI exams retire June 30, 2026 — target the replacements Confidence: High

AI-900 (Azure AI Fundamentals) → replaced by AI-901 — the new version explicitly covers building AI apps & agents with Microsoft Foundry. Ideal for the Foundry 100 goal.

AI-102 (Azure AI Engineer Associate) → replaced by AI-103: Azure AI Apps and Agents Developer Associate — built around generative + agentic architectures and multi-agent orchestration in Foundry. This is the perfect "stretch" cert after Foundry 200.

Because this plan assumes a fresh start in 2026, it points at the new exams (AI-901, AI-103) rather than the retiring ones, while still using the current, evergreen Microsoft Learn modules to prepare.

Phased roadmap

A suggested order at a sustainable ~6 hours/week. Phases can compress if you have more time — an existing Power Platform background will accelerate Phases 1–2.

PHASE 1

Foundations

Copilot Studio 100 + Foundry 100. Vocabulary, first agent, AI basics.

Weeks 1–2 · ~12 hrs
PHASE 2

Build & publish agents

Copilot Studio 200 learning path end-to-end + first published agent.

Weeks 3–5 · ~18 hrs
PHASE 3

Foundry gen-AI apps

Foundry 200 path: models, tools, RAG, evaluation, responsible AI.

Weeks 6–8 · ~18 hrs
PHASE 4

Advanced Copilot Studio

Copilot Studio 300: tools, MCP, agent flows, governance, evaluation.

Weeks 9–12 · ~28 hrs
PHASE 5

Multi-agent / A2A

Copilot Studio 400: connected & child agents, Agents SDK, Agent Framework.

Weeks 13–16 · ~28 hrs

Track 1 — Microsoft Copilot Studio (100 → 400)

The low-code agent path. Start by understanding agents, then build and publish, then extend with tools and governance, and finally orchestrate multiple agents together.

Copilot Studio

100 · Foundations

Outcome: Can explain agents, topics, and knowledge; has created a working agent in a trial environment.

Copilot Studio

200 · Build & publish

Build real agents — topics, entities, variables, knowledge sources, actions — test them, and publish to Teams / a website. The 6-module Microsoft Learn path is the spine here.

★ Core learning path: Create and publish agents with Copilot Studio Verified · 6 modules

Open the full learning path → (Intermediate). Portions reflect the "classic" authoring experience; pair with the new-experience docs below.

Copilot Studio

300 · Advanced

Move from conversational bots to capable, autonomous agents — custom tools, connectors, MCP servers, agent flows, prompts / code interpreter — plus the "enterprise" disciplines of governance, security, ALM, and evaluation.

Copilot Studio

400 · Multi-agent & agent-to-agent

Orchestrate agents that call other agents — the headline objective of the advanced track. Learn connected/child agents in Copilot Studio, then bridge to pro-code orchestration with the Microsoft 365 Agents SDK and Microsoft Agent Framework, plus enterprise identity & governance for agents at scale.

The "agents-to-agents" core Confidence: High

Copilot Studio supports composing agents directly: an orchestrating agent can delegate to child agents or connect to other existing agents. Start here, then expand to cross-platform orchestration.

Track 2 — Microsoft Foundry (100 → 200)

The pro-code generative-AI platform. Understand the platform and responsible AI, then build generative AI apps and agents with models, tools, RAG, and evaluation.

Microsoft Foundry

100 · Foundations

Outcome: Can navigate the Foundry portal, deploy a model, and explain core gen-AI & responsible-AI concepts — and is ready for the new AI-901 exam.

Microsoft Foundry

200 · Develop generative AI apps

Build generative AI applications and agents on Foundry — selecting and deploying models, adding tools & RAG, optimizing performance, evaluating, and applying responsible AI. Anchored by the official 6-module path (AI-3016).

★ Core learning path: Develop generative AI apps on Microsoft Foundry Verified · 6 modules

Open the full learning path → (Intermediate; some programming experience helps).

Stretch cert after Foundry 200 Confidence: Medium (new/beta)

AI-103: Azure AI Apps and Agents Developer Associate validates building production-ready agents with multi-agent orchestration in Foundry — the natural convergence of both tracks. Watch the Skills Hub blog for go-live dates and the study guide.

Power Platform → Copilot Studio synergy

If you come from Power Platform, you're not starting from zero conceptually — that expertise is a direct on-ramp. Here's how existing skills transfer.

Power Platform skill you may haveMaps directly to in Copilot Studio
Power Automate cloud flowsAgent flows & actions — same designer DNA; used to give agents deterministic automation.
Connectors & custom connectorsAgent tools via 1000+ connectors — instantly reusable knowledge.
Dataverse & environmentsAgent storage, security roles, and ALM (dev/test/prod solutions).
Power Apps maker mindsetLow-code agent authoring; embedding agents in apps; Agent Builder in Power Apps.
DLP & admin center governanceCopilot Studio data policies, tenant governance, and Agent 365 at scale.

Practical implication: with this background you can likely move through Phases 1–2 faster than a true beginner and should lean into actions / flows / connectors early, since those are immediate strengths.

Certifications & applied skills

A pragmatic credentialing ladder. Fundamentals first (fast wins), then the agent-developer associate as the capstone. Confidence: High

CredentialLevelWhy it mattersWhenStatus
PL-900 Power Platform FundamentalsBeginnerAnchors Copilot Studio in the Power Platform domain; covers copilots.Phase 1Current
AI-901 Azure AI FundamentalsBeginner / 100Foundry foundations & gen-AI literacy; the Foundry-100 target.End of Phase 3New — replaces AI-900 (retires Jun 30, 2026)
AI-103 Azure AI Apps & Agents Developer AssociateIntermediate / 200+Agentic + multi-agent orchestration in Foundry — unifies both tracks.After Phase 5New (beta 2026) — replaces AI-102
Microsoft Applied Skills (scenario credentials)VariesShort, hands-on, validated builds — check the catalog for current Copilot Studio / AI agent skills.OngoingBrowse catalog
PL-200 Power Platform Functional Consultant (optional)IntermediateDeepens core consulting credential alongside the AI work.OptionalCurrent

Smart sequencing tip

Fundamentals certifications don't expire, so PL-900 and AI-901 are durable wins. Associate-level certs renew annually (free online assessment). Register exams with a personal Microsoft account, not a work/school account, so records survive a job or contract change.

Hands-on labs & capstone projects

Learning sticks when you build. Pair each level with a concrete project.

HR Helpdesk Agent 200

Knowledge-grounded agent over a SharePoint policy library; publish to Teams. Covers topics, knowledge, testing, publishing.

IT Request Automator 300

Agent that creates tickets via a connector/flow tool and calls an MCP server for live data; add evaluation test sets & DLP.

Travel Concierge (A2A) 400

Orchestrator agent delegating to child agents (flights, hotels, expenses). Then re-implement orchestration with Agent Framework.

Foundry RAG Chat App 200

Deploy a model from the catalog, add a tool + retrieval, evaluate quality, and apply content safety in Foundry.

Suggested weekly schedule

A realistic cadence at ~6 hrs/week (e.g., 3× 2-hour blocks). Adjust freely. Confidence: Medium — illustrative pacing

WeeksFocusHrs/wkMilestone
1–2Copilot Studio 100 + Foundry 100 reading & first agents6First agent built; (optional) PL-900 booked
3–5Copilot Studio 200 learning path + new-experience docs6Published agent to Teams; Agent in a Day done
6–8Foundry 200 path + Agent Service quickstarts6Foundry RAG app; AI-901 attempted
9–12Copilot Studio 300: tools, MCP, flows, governance, eval7IT Request Automator with tests + DLP
13–16Copilot Studio 400: child/connected agents, Agents SDK, Agent Framework7Multi-agent capstone; plan for AI-103

Ongoing resources to follow

References & confidence

Primary sources used to build this curriculum (all Microsoft official), retrieved June 30, 2026. Confidence reflects how directly each claim is sourced.

  1. Microsoft Copilot Studio documentation hub — learn.microsoft.com/microsoft-copilot-studio High
  2. "Create and publish agents with Microsoft Copilot Studio" learning path (6 modules) — learn.microsoft.com/training/paths/work-power-virtual-agents High
  3. "Develop generative AI apps on Microsoft Foundry" learning path (6 modules, AI-3016) — learn.microsoft.com/training/paths/develop-generative-ai-apps High
  4. AI-102 certification page (retirement notice, Jun 30 2026) — learn.microsoft.com/credentials/.../ai-102 High
  5. Azure AI Fundamentals certification (AI-900 → AI-901) — learn.microsoft.com/credentials/.../azure-ai-fundamentals High
  6. Skills Hub blog — new 2026 cloud/AI/security certifications & retirements (incl. AI-103) — techcommunity.microsoft.com/.../4494128 Medium (some dates are planned/beta)
  7. Power Platform Fundamentals (PL-900) certification — learn.microsoft.com/credentials/.../power-platform-fundamentals High
  8. Agents hub (agent ecosystem, multi-agent concepts) — learn.microsoft.com/agents High
  9. Microsoft Foundry documentation hub — learn.microsoft.com/azure/ai-foundry High

Confidence methodology: High = directly stated on an official Microsoft Learn/Docs page verified during research. Medium = official but subject to change (beta exams, planned dates, hour estimates). Time estimates are typical ranges, not guarantees.