Copilot Studio: Build Real-World Agents-Beginner to Pro

Explore the basics of Copilot Studio and learn how to build real-world agents from beginner to pro.

Build your own Copilot agents with Copilot Studio, a Microsoft no-code tool that creates, customizes, and manages AI assistants using natural language and enterprise data.

Trace the evolution of Copilot Studio from Power Virtual Agents to autonomous agents, pro code options, and multi-agent orchestration, culminating in building your own AI agents.

Access Copilot Studio via a web portal with no-code and low-code drag-and-drop design. Use natural language building, API access, and admin controls across Microsoft 365 and Power Platform.

Explore how the Microsoft Copilot Studio helps you build, customize, and publish AI agents that complete tasks across Slack, Teams, and Telegram, with security, governance, analytics, and low-code options.

Copilot Studio pricing offers pay-as-you-go and commitment plans, with 25,000 messages per month and packs, where one pack equals two 25,000 messages, billed by message consumption.

Develop your first agent using Copilot Studio and progress from beginner to pro, building real-world agents with practical steps and hands-on guidance.

Navigate Copilot Studio to create and manage agents using built-in templates, flows, and tools. Connect to external data via REST APIs and the model context protocol.

Explore Copilot studio settings, including general options, session details, and the power platform admin center for governance. Adjust themes, language, privacy, and password, then prepare to create your first agent.

Create a copilot studio ai agent using new agent or describe and configure to tailor a university course advisor for uk students.

Explore the overview of the first agent in Copilot Studio, including its name, icon, and description, plus the core tabs: overview, knowledge, tools, agents, topics, activity analytics, and channels.

Test your agent by prompting it and reviewing greeting-triggered topics and responses. Inspect the activity map to see its step-by-step rationale and sources from trained data and public websites.

Explore the second agent's overview and settings, compare copilot and configure methods, and test the university agent with queries like UK universities for AI, noting JNI orchestration and GPT four.

Explore how to configure your agent’s settings, including generative AI orchestration, connected agents, model selection, formatting, moderation, knowledge sources, and file processing.

Explore knowledge sources to power real-world agents in Copilot Studio, guiding learners from beginner to pro.

Review how Copilot Studio powers a university agent and chatbot, contrasting generated data with public sources, and plan to add UCAS as a knowledge source for accurate answers.

Explore generative answers and how they draw from knowledge sources to deliver intelligent, automatic responses. Understand grounding with trusted data sources and how topics, MCP, and fallback functions guide Copilot.

Explore how knowledge sources power intelligent agents: public websites, documents, SharePoint, and Dataverse, with internal versus external authentication and generative mode.

Disable general knowledge and web grounding, then add a public website such as ux.com as the agent's knowledge source. Prepare to test queries in the next lecture.

Demonstrate how adding a knowledge source alters agent responses and grounding, verified via overview and knowledge tab. The demo shows the activity map grounding to the UCAS site with references.

Explore topics and nodes to assemble real-world agents with Copilot Studio, guiding beginners toward professional capabilities.

Define topics to create intelligent conversation paths in Copilot Studio, organizing complex dialogues into modular units. Intents recognize natural language to trigger topics and guide structured, testable conversations.

System topics are pre-built conversations by Microsoft handling greetings, farewells, and escalations. Custom topics create business-specific conversations that pull data from your sources and can be customized, but not removed.

Explore system topics and custom topics in Copilot Studio with a demo of the university course advisor agent. Learn editing, triggers, and how topics shape conversations.

Explore Copilot Studio nodes as building blocks of conversations, including message, question, adaptive card, condition, variable management, topic management, and tool calls to a rest API to design real-world agents.

Understand how topics and nodes form a cohesive conversation architecture, linking what is discussed to how the dialogue unfolds, including triggers and variables in a pizza order.

Test the copilot agent by triggering the course comparison topic and watching the activity map respond. Validate the flow as you choose duration or fee, noting Cambridge's international student fee.

Create a manual topic from blank with the Copilot method, guiding users through course selection and interests. Define conditional flows and save the topic as a recommendation engine.

Demo teaches testing Copilot Studio agent topics by switching between course comparison and a recommendation engine, showcasing workflow-driven topic activation and building custom topics with the Copilot method.

Explore tools in Copilot Studio, the building blocks that let your agent interact with external systems, via APIs and connectors, and perform actions beyond conversation.

Explore the tools currently supported in Copilot Studio, including the prompt tool, agent flow, custom connector, and MCP model context protocol, and how they automate actions and connect to data.

Showcases creating and attaching a custom prompt tool to an agent, then testing it to reveal future AI career prospects and next steps like internships and certifications.

Revisit the Copilot Studio architecture and add a rest api tool using the UniDB api to attach third-party data to your university agent.

Connect your Copilot Studio agent to the unidbapi.com API to access university course data using the popularity rank endpoint, and learn to obtain an API key with hands-on testing.

Connect the uni db API to Copilot Studio via a custom connector in the power platform, import the open API file, configure api key authentication, and test before deployment.

Add and configure the uni db API connector to an agent, supply the API key, and test the degree popularity rank tool for the university course advisor.

Test the agent with a rest API connector to fetch biology universities from UCAS and retrieve popularity rank via a no-code, natural-language workflow.

Create an agent flow in Copilot Studio to send an email, connect flows with power automate, publish and rename the flow for testing.

Test the agent flow by triggering the send email Jarvis action, verify email delivery, and explore how agents perform outside world actions using Power Automate and knowledge sources.

Learn to design and implement multi-agent systems that operate in real-world scenarios, guiding beginners toward advanced proficiency with Copilot Studio.

Explore multi-agent support in Copilot Studio, using connected agents and parent-child subagents to create modular, team-based AI workflows that collaborate like humans.

Compare connected agents and child agents in Copilot Studio, where connected agents are external services linked by APIs or plugins, and child agents are internal sub-agents.

Demonstrates creating a child admissions agent inside a university agent, configuring it with an Oxford admissions knowledge source, and wiring collaboration so the university agent hands off queries.

Test the multi-agent setup by querying the admissions sub-agent for Oxford, extracting chemistry admission requirements from a knowledge source PDF, and validating references in Copilot Studio.

Publish real-world agents built with Copilot Studio, mastering deployment and showcasing capabilities from beginner to pro.

publish your agent with one click to reach multiple channels, including microsoft teams, copilot, sharepoint, and custom websites. update changes by republishing to ensure up-to-date ai capabilities across all connected channels.

Explore Copilot Studio channels, including demo websites, Teams, SharePoint, and social apps via Azure Bot Service, to publish across multi-channel environments and hand off to live agents.

Compare no authentication and Microsoft Entra ID for publishing agents, with guidance on when to use each and applying governance policies and access control for sensitive data.

Publish your agent across channels like the demo website, Microsoft Teams, Microsoft 365 Copilot, and SharePoint, then assign per-user or tenant licenses in the admin center for free trials.

Assign licenses in the Microsoft 365 Admin Center for active users, and deselect the Microsoft 365 Copilot studio trial license to avoid conflicts before publishing agents.

Learn to create a dedicated publishing environment in Copilot Studio using the Power Platform admin center, then build, publish, and test agents in a non-prod sandbox.

Publish your agent to the demo website channel, adjust authentication, and verify a public endpoint to test real-world access.

Publish your Copilot Studio agent to Teams by enabling Microsoft authentication and adding the M365/Teams channel, then test the university advisor inside Teams with source-based responses from UCAS.

Learn to build real-world ai agents with Copilot Studio, advancing from beginner to pro.

Develop a study scheduler agent using Copilot Studio, guiding beginners to proficiency in building real-world agents through practical, hands-on strategies.

Learn to build a study scheduler agent with adaptive cards and prompts, converting subjects into a JSON list and generating a schedule with start time, end date, and daily hours.

Create and customize a study scheduler agent that collects subjects, lists them, and generates a study schedule table, with guardrails and settings to limit web access and generative ai.

Build a topic in Copilot Studio, trigger it with new schedule phrases, collect subjects, and create a prompt that converts input into a list of subjects stored as a variable.

Explore adding an adaptive card to the create schedule topic, collecting start date, start time, end date, and time in hours via a JSON payload and a submit schedule button.

Learn to design a prompt that converts JSON to a table schedule, expanding across dates, deriving start times from durations, and randomizing daily subjects for a study plan.

Convert inputs to text using functions, store the output in a final_schedule variable, and print final_schedule.text as the final message for the upcoming agent test.

Test the agent by creating a study schedule for physics, economics, and chemistry, using an adaptive card to generate a randomized plan starting at 7:00 a.m. for 3 hours daily.