Architecture Bootcamp - V3
Architecture Bootcamp - V3
Three-day architecture bootcamp rebuilt around the agent harnesses — Copilot, Standard, and GitHub Copilot — their component models, Cowork, workflows, and coding agents. Deep dives on knowledge, tools, instructions, and skills, closing with ALM, analytics, and governance.
Report an issue- 17Modules
- 8Labs
- ~15.8hContent
- 100–200–300 Levels
Agenda
- Day 1 — Framing, On-ramp & Copilot Harness
-
09:00Welcome
-
09:15Introduction to Hands-On Labs
Lab environment, tenant setup, and how the hands-on flow works across the three days.
-
09:30
Opening frame for the bootcamp: what an agent actually is in 2026, the five kinds of agent, the three harnesses that sit on top of Copilot Studio, and the key concepts — knowledge, tools, skills, instructions, descriptions, plugins — that every later module builds on.
Slides -
10:15Coffee break / Q&A
-
10:30
Microsoft 365 Copilot as an agent surface — public-web grounding, Work IQ, response modes, multi-model support, frontier agents, and where it sits against the other surfaces.
Slides -
11:00
Use Microsoft 365 Copilot chat, Work IQ, response modes, multiple AI models, and frontier agents to turn public and organizational inform...
-
11:30
Cowork at bootcamp altitude — what it is, where it fits against the other agent surfaces, and how skills and plugins extend it.
Slides -
12:00Lunch break
-
13:00
Delegate a document-grounded task to Copilot Cowork and have it draft real output in Microsoft 365 on your behalf.
-
13:45
The overview that anchors Days 1-2: how orchestration has changed, what a harness is, and the component model each harness exposes — before the three dedicated deep dives.
Slides -
14:30Coffee break / Q&A
-
14:45
Dedicated pass over the Copilot harness: Agent Builder, pro-code declarative agents, and declarative agents authored in Copilot Studio.
Slides -
15:30
Build declarative agents in the Copilot harness with Agent Builder — grounded in web content and in your own SharePoint, with code interp...
-
16:00Q&A
- Day 2 — Standard & GitHub Harnesses, Creation & Workflows
-
09:00Re-Welcome / Re-cap
-
09:15
Dedicated pass over the Standard harness and the component model it exposes in Copilot Studio.
Slides -
09:45Coffee break / Q&A
-
10:00
Assemble an agent in the Standard harness from its component parts — topics, knowledge, tools and variables — one piece at a time.
-
10:45
The longest dedicated harness pass: the GitHub Copilot harness and its full component model — instructions, skills, knowledge, tools, memory, sandbox, and delegation.
Slides -
11:45Lunch break
-
12:45Lab 5: GitHub Copilot harness — Component Model
Placeholder — lab to be authored. Seeded by mcs-orchestration UC3 (Agentic Reasoning Loop), roughly half the runtime.
-
13:45
Creating agents in the GitHub Copilot harness two ways — from the authoring UX and from natural-language chat.
Slides -
14:30Coffee break / Q&A
-
14:45
Workflows are the next generation of autonomous agents in Copilot Studio — inline agent connections, plus scheduled and event-driven triggers that let an agent act on its own. This module introduces the new capabilities and how they evolve beyond the classic autonomous-agent trigger model.
Slides -
15:15
Build an autonomous agent using Workflows — the next generation of autonomous agents in Copilot Studio — with an event-driven trigger and...
-
16:15
What agent usage actually costs and how to keep it predictable — the credit and message model, licensing paths, capacity planning, and where to watch consumption once agents are live.
Slides -
16:45Social / Networking
- Day 3 — Knowledge, Tools, Skills & Operations
-
09:00Re-Welcome / Re-cap / Quiz
-
09:15
Merged deep dive on grounding and action: knowledge sources and retrieval, then connectors, agent flows, custom prompts, and MCP servers.
Slides -
10:00Coffee break / Q&A
-
10:15
Learn how to extend your agents with tools including connectors, agent flows, MCP servers, and custom prompts in Microsoft Copilot Studio.
-
10:45
How instructions shape behaviour on every turn, and how descriptions drive selection — which tool, knowledge source, or agent the planner reaches for.
Slides -
11:15
See what the generative orchestration planner actually reads when it decides which agent, tool, and input to use - then watch the New Orc...
-
12:00Lunch break
-
13:00
Skills as instructions on demand — packaging behaviour into a reusable unit the orchestrator loads only when it becomes relevant.
Slides -
13:30
Package a repeatable procedure as a Skill, then run that one Skill across a Copilot Studio agent, a workflow, Copilot Cowork, and coding ...
-
14:15
Multi-agent architectures and collaboration patterns in Copilot Studio.
Slides -
14:45
Coding agents against Copilot Studio: the plugin for GitHub Copilot CLI and Claude Code, migrating existing agents, and creating new ones from the terminal.
Slides -
15:15Coffee break / Q&A
-
15:30
One operations module: traditional ALM and component collections, folded together with reporting, analytics, and agent evaluations.
Slides -
16:00
Closing operations module: governance zones, DLP and managed environment controls, brought together with Agent 365.
Slides -
16:30General Q&A
- Continue on your own — optional labs
-
Build a main agent with child agents for product and account information. 30 min.
-
Solutions, environment variables, connection references and deployment pipelines. 45 min.
-
Build, share and version component collections across agents. 30 min.
-
Monitor performance, configure evaluation test sets and review results. 30 min.
-
DLP-restricted agents across green, yellow and red zones. 45 min.