The recent drama at OpenAI shed light on Microsoft’s substantial investment in GPT and its derivatives, powering the expanding Copilot family. These generative AI capabilities, central to natural language interfaces, provide language understanding, summaries, and user-centric text outputs. As Copilots proliferate across Microsoft’s platforms, the next question is: How can we integrate them into our code?
Previously, we explored Azure AI Studio for direct access to machine learning models. Now, let’s delve into Copilot Studio, Microsoft’s low-code AI development tool, designed for users beyond experienced developers and data scientists.
For those not versed in traditional development tools, Microsoft’s Power Platform, akin to Visual Basic in its simplicity, offers low-code and no-code solutions. Power Virtual Agents, with recent enhancements using OpenAI’s generative AI models, now introduces Copilot Studio, an AI authoring environment available as a web application and Teams app.
These complementary Copilot Studios cater to different audiences, empowering development teams and IT admins with the web application, while allowing end-users to prototype ideas through the Teams app. This dynamic pairing facilitates seamless collaboration and transition from internal tools to refined, externally shared applications.
To embark on building a generative AI chatbot with Copilot Studio, you need an account. A trial account, accessible with a Microsoft 365 account, lets you explore before committing. The process involves training your bot on topics using documents, with Copilot Studio providing a GPT-powered “boost” to summarize additional data based on user intent.
Central to Copilot Studio is the concept of “topics,” labeled collections of related information used to train the underlying AI model. For instance, a chatbot supporting IT triage might have a topic labeled “Supported applications” with links to self-service update instructions. The conversation flow, rules, and queries are defined by you, guiding users through various self-help tools.
The GPT-based service kicks in when predefined resources don’t yield answers. Copilot Studio taps into external data sources, acting as semantic memories to ground the language model, enhancing the chatbot’s responses. This hybrid approach maximizes efficiency by employing lower-cost, low-compute strategies and resorting to more resource-intensive queries when necessary.
Remarkably, Copilot Studio isn’t just a tool; it’s a copilot in its own right. The “Create With Copilot” option lets you describe your intended user experience, and the system generates an appropriate topic. This AI-guided development approach marks a departure from traditional design surfaces, showcasing a user-intent-driven model, potentially revolutionizing intentional development.
As Copilot Studio evolves, its integration with Power Platform’s connections to line-of-business applications and Dynamics 365 positions it as a simplified gateway to infuse intelligence into businesses, transcending Q&A-driven chatbots. The journey through Copilot Studio unfolds as a user-friendly expedition into the realms of AI-driven chatbot development.