Build Your Own AI Agent in Microsoft Copilot in Just 5 Minutes

Vídeo em Inglês, publicado pelo canal Azhar.

About this content: The summary, key points and topics below are generated automatically from the video transcript. They reflect the content, opinions, claims and predictions presented in the video and do not represent a position, validation or fact-check by IA Dots.

Video summary

The video demonstrates a step-by-step process for creating a custom AI agent in Microsoft 365 Copilot without coding, covering configuration, instructions, knowledge sources, suggested prompts, testing, and refinement.

Key points

• Open Microsoft 365 Copilot, navigate to agents, select new agent, and choose skip to enter the configuration screen.
• Assign a descriptive name and a short description to define the agent's purpose.
• Define the agent's role, instructions, and rules, such as organizing meeting notes and avoiding fabricated owners or deadlines.
• Add optional supporting knowledge sources like cloud files, Outlook, Teams, Copilot connectors, and web search.
• Create suggested prompts to give users useful starting points and test the agent using fictional meeting notes to refine its behavior.

Topics

Microsoft 365 CopilotAI AgentsNo-code DevelopmentMeeting PreparationConfigurationKnowledge SourcesTesting and Refinement

Video transcript

Transcript from available captions; it may contain errors.

Let's create your own AI agent in Copilot. No coding needed. Let's build our first AI agent. Open Microsoft 365 Copilot. Go to agents and select new agent. Then choose skip to move straight into the configuration screen. Start with a name that clearly describes what your agent does. For this demonstration, I'm calling mine meeting coach. It's an example of an agent that helps organize meeting preparation and notes. Then add a short description of its purpose. Next come the instructions. This is where you define the agents role, what it should help with, and how it should respond. Our example creates agendas and organizes meeting notes into decisions and action items. We're also adding clear rules. Don't invent owners or deadlines. Separate suggestions from agreed decisions and mark missing details as to confirm. Think of this as writing a clear brief for your assistant. The knowledge section is where you can add supporting resources. This is optional. You could attach a guide, a template, or another document relevant to the job you want your agent to do. These switches control which sources the agent can search, enable the sources relevant to its purpose, and switch off those it doesn't need. Cloud files can provide context from One Drive and SharePoint. For example, project documents, reference guides, or templates you already use. Outlook can add background from email conversations and follow-ups. Here, it's an information source. Enabling it doesn't make the agent send emails. Teams can provide context from chats and meeting information. Selecting specific chats or meetings can help keep the agent focused. Copilot connectors bring in external work content enabled by your organization. Web search lets the agent look for information on public websites. The benefit is more relevant context for your agent. Connected work content follows existing access permissions and available options depend on your license and organization settings. Next, add suggested prompts. These give people useful starting points when they open the agent. Each one has a title and a message. The actual request sent when someone selects it. For this example, we're adding three. Improve a meeting, build an agenda, and extract action items. Choose prompts that demonstrate the main tasks your own agent will handle. When the setup is ready, select create. Once you see the confirmation, your agent has been created. In this example, it's private and available only to me. You can now open it and start a conversation. To make changes, find your agent under agents, open its menu, and select edit. Configure lets you adjust the setup. Switch to preview to test it. Use the message box on the right to talk to your agent. The agent builder conversation on the left is for changing how it's configured. Let's test it using fictional meeting notes. We'll ask it to separate agreed decisions, action items, and open questions, and specifically tell it not to suggest owners or deadlines. Notice how it separates agreed actions from a proposal. It keeps Aisha as the named owner and leaves missing details as to confirm. Check that the response follows both your notes and your instructions. Now let's refine the output. Return to configure and add a rule requesting a status column. We'll define the labels as agreed, proposed, or unclear based on the notes. We'll also clarify that missing an owner or deadline doesn't make an agreed action merely a proposal. Select update to save the changes. After refining instructions, start a fresh test conversation and check whether the response follows your new rule. Now return to Copilot and open your agent from the agents menu. Give it a new task. Your saved instructions are already in place. Here we've asked for a 20inut agenda and three discussion questions. The agent produces a draft we can review and use. That's the practical benefit. You can return to the same specialized assistant for new tasks without repeating its full setup. You've now created a reusable AI agent. Give it a clear purpose. Then adapt this approach to planning, drafting, organizing information, or other recurring tasks. your own AI agent built around your work. What will you create?

Assistir ao vídeo original no YouTube

Assuntos: Agentes de IA, Assistente pessoal, Assistente pessoal. Publicado no YouTube em 2026-10-25.

Seleção automática por assunto e idioma. O vídeo pertence ao canal de origem.