Microsoft has recently unveiled the introduction of its new Copilot Studio platform, aiming to empower Microsoft 365 organisations to build their own generative artificial intelligence (AI) ‘copilot’ assistants.
Announced at Microsoft Ignite 2023 this month, Copilot Studio was officially announced alongside a plethora of other updates (and new Copilots), as part of their ongoing efforts to enhance and expand artificial intelligence capabilities across their core product offerings.
With Copilot Studio, users will be able to efficiently create, test, and deploy independent copilots and custom GPTs, providing them with a seamless experience in harnessing the potential of AI technology.
The announcement of Microsoft Copilot Studio follows the roll-out of a similar DIY copilot creation platform from OpenAI last week called GPT Store. Copilot essentially replaces Power Virtual Agents and is available to try now.
What does Microsoft Copilot Studio do?
Microsoft Copilot Studio leverages low-code, no code and provides a web-based platform that completely transforms the way Microsoft 365 customers users approach application development and extension. Of course, one of the key advantages of Copilot Studio is its seamless integration with the wider Microsoft 365 ecosystem and the Microsoft Graph – enabling developers to easily leverage existing data and tools within their organisation, streamlining the development process and maximising productivity. Copilot Studio, like Power Platform is designed for professional developers or users new to AI application development – providing a comprehensive set of tools and resources to support your journey.
Copilot Studio makes it easy to design and prototype AI applications and offers a wide range of templates and pre-built components, empowering even non-technical users to create innovative solutions.
Copilot Studio also brings robust testing capabilities, including integrated debugging tools and simulation environments to help developers easily identify and address any issues before deploying their applications.
Once the AI app is built and tested, Copilot Studio provides a simple deployment process. It provides seamless integration with Microsoft 365 services that allow IT to publish the application to your Microsoft 365 cloud environment or share it with specific users or teams.
How to build your first Copilot in Copilot Studio (follow along).
Once you have signed up for a trial of copilot (it’s in preview), getting started is easy. To login, simply go to Microsoft Copilot Studio
Step 1: Build my Copilot. The easiest way to start is to follow the wizard and point Copilot Studio at a website. For this example, I have used my children’s school.
There are advanced options which include the ability to dive straight into existing data sets and experiences that you have created in Power Platform, but for simplicity, I am following the wizard.
Step 2: Explore the interface: From here you can then see the options expand. You can go straight to the Generative AI Chatbot and start to test it based on the website you have linked it too, or start looking at the options below which include the ability to:
- Create Custom Topics for your Copilot.
- Create Plugins and Actions to connect to data that is outside the website.
- Extend the Microsoft 365 Copilot with data connectors and plugins to your other data.
Step 3: Customise the Generative AI settings: I’m going to click on the “Go to Generative AI” and customise the “sensitivity” of Copilot (I have set mine to medium). I can also further customisations such as adding additional websites for indexing or even upload specific files. The Generative AI model uses the Azure Open AI service. Here I have added another website to index.
Step 4: Testing my Copilot. I can now quickly test my Copilot out by asking a question. In this example I am going to ask it about the schools in the Academy Trust and then dig a little more into the school he attends.
Step 5: Adding Custom Topics: By default, you don’t have much control over the data that comes with the generative answers. This is where topics come into play. These allow you to add in specific topics that you want the bots to be able to speak or answer that have not come automatically from the data, websites, databases or documents you have connected in.
Step 6: Publishing your Copilot: To Publish your Copilot, you have a few options. You can publish to a website (either your own or one can be created for you), to Teams or other locations. To do this, you first need to navigate to Settings –> Channels and choose where you wish to publish your Copilot. For this example, I have chosen a Demo Website.
Once you’ve done that (you can pick multiple), head on over Publish menu on the left and publish your Copilot. You then get a link to test your Copilot. Here is mine below in a simple demo website.
There we go – a very simple first look at the potential of Microsoft Copilot Studio. Next time we will explore a little more. Have a play and look at the potential and possibilities.
How can I try Copilot Studio?
Copilot Studio is available to try now and will be available to all Microsoft 365 customers who have a Copilot for Microsoft 365 license.
Copilot Studio should bring the same ease of use to Gen AI apps as Power Platform did for apps, virtual agents and flow. It’s central hub, intuitive design tools, robust testing capabilities, and simple yet powerful deployment options aim to help accelerate and simplify the design, build, testing and deployment of custom AI applications.
By providing users with the tools to develop their own copilots, Microsoft is fostering a culture of innovation and collaboration. Copilot Studio opens up countless possibilities for businesses, researchers, and developers alike. Whether it’s creating intelligent chatbots, generating realistic text, or automating repetitive tasks, the potential applications of GPTs are limitless.
Similar to its counterpart GPT Store, Copilot Studio puts the power of AI into the hands of those who seek to harness its capabilities. It offers a user-friendly interface that streamlines the process of building custom GPT models and applications. This accessibility ensures that individuals with various levels of technical expertise can participate in this exciting field, expanding the boundaries of what is possible with language-based AI.
The demo build example is a simple first step into building Gen AI led chatbots and copilots using websites and other data sources. This is simple to test and try.