How Large Language Models are Transforming Bot Building and Making Them More Useful for Everyone

The latest wave of innovation around large language models (LLMs), such as ChatGPT and GPT-4, is rapidly transforming the world of bot building. Natural language processing (NLP) and machine learning have long formed the foundation for building intelligent bots, but the capabilities of these LLMs are paving the way for exciting new possibilities in this space. One example of how this technology is changing the experience for both users and authors of chatbots is Power Virtual Agents, specifically the new Copilot feature and the Boost Conversations capability. You can watch an overview here.

Power Virtual Agents (PVA), a low-code approach to building chatbots with an intuitive visual drag-and-drop interface, has already transformed the bot-building process by making it easier and accessible to non-technical bot builders. However, with the introduction of Copilot, bot building has become even simpler. The new feature uses the latest generative AI capabilities to allow authors to create entire topics from a simple description, including relevant trigger phrases (used for NLU), questions, messages, and conditional logic. Copilot allows anyone to create topics in minutes, democratizing conversational AI, and broadening the potential audience further than ever before. Additionally, the ability to author with natural language is not limited to just creation, with PVA innovating further, with Copilot also allowing you to continually iterate with follow-up prompts.

Building a bot topic with Copilot in Power Virtual Agents

Boost Conversations in PVA is another exciting feature powered by OpenAI. Traditionally, bots are limited in their ability to respond to users, by the capabilities explicitly added by the author. Whilst this has been improved over recent years with the ability to search across curated QnA knowledgebases, getting a bot to the point of adding value for end users still takes time and they are often faced with the bot simply not being able to provide an appropriate response. Now though, using Boost Conversations, organizations can point a bot to useful data sources, like external or internal websites, and the bot can immediately start to use that data to construct a response, searching across the data and using GPT to summarize the response. This means that out of the box, the bot is ready to answer user questions based on the information on your site — all without authoring a single topic or spending additional development cycles. This feature is particularly useful for organizations that may have been reluctant to create bots due to the expected build time and cost.

Configuring Power Virtual Agents with Conversation Boosters using your website URL

While LLMs clearly have significant potential in the world of chatbot building, it is important to acknowledge the challenges associated with their usage, such as their ability to hallucinate. This is something that Microsoft has worked to address, by creating responsible AI by design. The work is guided by a core set of principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. For example, within Boost Conversations, content moderation, and filtering has been included, with settings available to fine-tune the level of moderation applied to bot responses, to try and ensure that the information provided is accurate and relevant to the user.

The latest wave of innovation around large language models is clearly, rapidly, revolutionizing the chatbot experience, making bot-building simpler and more accessible than ever before, and ensuring bots are useful immediately. Ultimately, this is making bots an even more valuable tool for organizations looking to improve customer outcomes and increase resolution rates. Given the speed at which these innovations are being made, the pace of positive change in this area is only likely to increase.

Additional Resources:

• Find inspiration with relevant content ideas using Copilot | Microsoft Learn
• Overview — Power Virtual Agents | Microsoft Learn
• How to build a basic chatbot — Training | Microsoft Learn
• Describe complementary Microsoft Power Platform solutions — Training | Microsoft Learn

About the Author:

Gary Pretty — Principal Product Manager, Microsoft

Gary is a Principal Product Manager for Power Virtual Agents and Conversational AI at Microsoft, focused on delivering exceptional conversational authoring and development capabilities. Formerly, Gary was a Microsoft AI MVP.

Originally posted on OpenDataScience.com

Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform. Subscribe to our fast-growing Medium Publication too, the ODSC Journal, and inquire about becoming a writer.