Publishing to Microsoft Teams, the web and mobile - SkilBrill

Microsoft Copilot Studio Training Publishing to Microsoft Teams, the web and mobile

Microsoft Copilot Studio Training

0% Completed

M1 — Copilot Studio and the Agent Model

What Copilot Studio produces and where it runs

Standalone agents versus agents surfaced inside Microsoft 365

The authoring canvas: agent, topics, actions and knowledge

Generative mode versus adaptive mode and choosing the dialog behaviour

Environments, solutions and how an agent moves between them

How Copilot Studio relates to Power Virtual Agents heritage

M2 — Topics and Dialog Design

Trigger phrases: what the language model matches against

Designing a topic that recovers when the user goes off-script

Questions, variables and types the agent can hold

Branching, conditions and the dialog tree that stays readable

Redirects, global topics and reusing a sub-dialog

Escalation to a human and handing over the full conversation

M3 — Generative Answers and Knowledge

Grounding an agent on documents, websites and internal content

Generative answers with citations the user can follow

Improving answers by fixing the source rather than the prompt

Custom instructions and shaping tone, scope and refusal

When generative answers should hand off to a topic instead

Measuring answer quality on real questions, not ideal ones

M4 — Actions, Connectors and Power Automate

Custom actions over an API the agent can call

Connectors: authentication, schemas and failure handling

Calling Power Automate flows from inside a dialog

Passing conversation context into an action safely

Validating arguments and refusing an action the model should not take

Returning a result the dialog can present without exposing raw JSON

M5 — Channels, Identity and Publishing

Publishing to Microsoft Teams, the web and mobile

Authentication options and acting for a signed-in user

Single sign-on and the identity boundary in an agent

Handing off to a live agent with the transcript intact

Branding, the standalone page and embedding in a site

Version promotion and rolling a published agent back

M6 — Governance, Security and DLP

Data loss prevention policies for connectors and actions

What a user can and cannot make an agent do

Audit and analytics: conversations, failures and resolutions

Environment separation across development and production

Solution packaging so changes are reviewable

Writing the acceptance test a business owner can sign off

M7 — Applied Skills

Debugging a topic that never fires its trigger phrase

Reducing a dialog tree that grew to forty unreadable branches

Choosing generative answers versus a topic for a given question

Cost and usage: message limits and what drives them

Reading another team's agent and reconstructing its intent

Estimating build effort for a conversational use case

M8 — Capstone

Design an agent for a real internal or customer-facing process

Build topics with triggers, branching and a human escalation path

Ground generative answers on your own source content with citations

Add a custom action through a connector to a real API

Apply authentication plus a DLP policy and publish to two channels

Deliver tests, version history, analytics and a support runbook

Announcements Course Info