{"id":2996,"date":"2024-03-19T08:00:31","date_gmt":"2024-03-19T16:00:31","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/microsoft-copilot\/blog\/copilot-studio\/building-your-own-copilot-with-copilot-studio\/"},"modified":"2024-06-05T10:15:49","modified_gmt":"2024-06-05T17:15:49","slug":"building-your-own-copilot-with-copilot-studio","status":"publish","type":"copilot","link":"https:\/\/www.microsoft.com\/en-us\/microsoft-copilot\/blog\/copilot-studio\/building-your-own-copilot-with-copilot-studio\/","title":{"rendered":"Building your own copilot with Copilot Studio"},"content":{"rendered":"\n

Learn how you can use Copilot Studio to build your own copilot, available on multiple channels, designed for your audiences, and tailored specifically to your business processes and workflows.<\/h4>\n\n\n\n

In November, we introduced Microsoft Copilot Studio<\/a>, a conversational AI platform that allows you to customize Copilot for Microsoft 365<\/a> and build your own copilot.<\/p>\n\n\n\n

Copilot for Microsoft 365 offers AI assistance to create, find, and ask for information across Microsoft 365 apps and provides a Copilot chat available across several employee experiences like Microsoft Teams, Windows, and Bing. However, sometimes you may want generative AI assistants that can reach users on multiple channels outside of those, like embedding on your internal websites, mobile apps, and social channels. You may want to design individual AI assistants that can serve employees, customers, and partners. Or you might want a custom copilot that you can tailor to specific roles and functions, where you can control the orchestration of your large language model using the dialog manager for managed topics with specific workflows.<\/p>\n\n\n\n

When you create your own copilot with Copilot Studio, you are building intelligent chat experiences using ready-made large language models, a dialog manager, 1200+ data connectors, and more within a low code SaaS. These custom copilots are AI assistants that help humans with complex cognitive tasks using your organization\u2019s knowledge, connectors to LOB applications, and your own custom topics. For example, a copilot for IT support, a copilot to help your customers choose the right product, or even help your suppliers track the status of orders.<\/p>\n\n\n\n

How Copilot Studio works<\/strong><\/h2>\n\n\n\n

Let\u2019s talk through the steps to building your own copilot. You can start by logging in to Copilot Studio<\/a> which will spin up all of the required services for you so it\u2019s ready for building in a matter of minutes.<\/p>\n\n\n\n

Imagine you wanted to create a custom copilot to serve customers on your website, for example assisting with product knowledge questions and order management.<\/p>\n\n\n\n

First, start by grounding your copilot in your data. Using generative answers<\/a>, you can enable multi-turn chat over your organizations real-time data\u2014from local files to SharePoint sites, from public websites to data in your own custom backends creating the ability to chat over a wide set of topics within minutes. For this scenario, you would use your public website including your product information.<\/p>\n\n\n\n

For critical topics such as account management, which involves handling customer details, you can design the specific conversational flow<\/a> that you want Copilot to follow. You have the option to use visual authoring or natural language to design these prioritized topics before proceeding to generative AI. For instance, consider a topic that collects user details, authenticates the user, and interacts with your order management systems following a specific sequence of steps. Another example of a managed topic is discounting, where responses may involve legal and compliance matters, and therefore, you provide the managed response.<\/p>\n\n\n\n

With generative actions<\/a>, you can now provide a list of plugins to the copilot, selecting from 1200+ data connectors like SAP, Workday, Salesforce etc, and these will be dynamically executed to help the copilot take action. This allows copilot to effortlessly handle complex queries that were unforeseen or not previously constructed.<\/p>\n\n\n\n

For developers that want more custom development, you can use Azure models and services with Copilot Studio. This hybrid approach of low code with custom pro code integration<\/a> allows you to bring custom language models<\/a>, Azure OpenAI on your data<\/a>, knowledge bases<\/a>, image generation with DALL-E, telemetry, and more into your Copilot Studio copilot.<\/p>\n\n\n\n

When you\u2019re ready, you can publish your copilot to multiple channels<\/a> which can be internal or external facing to interact with users across different platforms, such as websites, Microsoft Teams, social apps, mobile apps, Azure Bot Service channels and more. You can also escalate the conversation<\/a> to 1st party Dynamics 365 and 3rd party tools like Genesys, LivePerson, Salesforce and ServiceNow when the copilot requires human assistance.<\/p>\n\n\n\n

Once the copilot is live, be sure to review the copilot performance with the built-in analytics dashboard<\/a> that helps makers monitor key metrics, evaluate performance, and find new copilot topics. Secure and manage your copilot with governance and control features with the central admin center<\/a> to protect copilot data with data loss prevention policies, role-based access control, environment management and more.<\/p>\n\n\n\n

These are just some of the foundational steps to getting started, but there are many other features<\/a> you can include in your copilot.<\/p>\n\n\n\n

How Copilot Studio will evolve in the future<\/b><\/h2>\n\n\n\n

When you build custom copilots using Copilot Studio today, you are designing chat-based experiences to support users with knowledge discovery, such as finding answers to their queries, and task\/business process completion, such as calling APIs and following workflows.<\/p>\n\n\n\n

We are continuously enhancing the capabilities of Copilot Studio in the upcoming months through several key areas of focus:<\/p>\n\n\n\n