Enhancing Customer Engagement with Chatbots and Azure AI Studio
In today’s digital age, businesses are constantly seeking innovative ways to enhance their online presence and improve customer engagement. At Ana’s Cloud, we embarked on an exciting project to develop a chatbot that leverages the power of Azure AI Studio and other cutting-edge technologies. Here’s a glimpse of our journey and the solutions we implemented for our client, The Conference Board (TCB).
Project Overview
Our goal was to create a chatbot powered by a large language model (LLM) that provides both public and member access to TCB’s website and exclusive reports. Based on a user’s query and TCB’s member-exclusive content, the chatbot would generate responses that would include a link, redirecting the user back to the content on their website. For non-members, the chatbot would encourage users to sign up for more content and resources, including TCB’s conference and center memberships.
Indexing the Site
Before settling on a product to use, we needed a way for our LLM to access TCB’s documents. We customized a NodeJS script, originally provided by Microsoft, to parse through and vectorize Office365 and PDF documents into a JSON format and upload it to a Microsoft Azure Search Index. The script was set up to run regularly so the LLM is up to date with their daily documents.
Initial Challenges
We initially tried using the off-the-shelf SaaS product from Microsoft called Copilot Studio. While it was good proof of concept, the product did not meet our needs or the clients. It lacked proper vector search capabilities, and UI and UX customization. However, it excelled in having its own web search capability, with the LLM having access to TCB’s website in real time.
Customizing the Chatbot
We ended up moving on from Copilot Studio and found an open source chatbot on GitHub based on the previous product. The new chatbot was also maintained by Microsoft and allowed us to make the modifications we needed to achieve our goals and the clients. The project utilized the Azure AI projects and properly interfaced with our Azure Search Index for vector search. We were also able to add integration with Bing to the LLM, so we can capture the web search that Copilot Studio had. We were also able to utilize prompt engineering to prioritize Bing index results, ensuring that documents always had the most up-to-date information, such as the Consumer Confidence Index.
To further enhance user experience, we implemented a feature that prompts users after a certain number of questions to see if they need help from a human if they are a TCB member. If they are, the user could fill out a form that would then be sent to TCB’s customer service to then follow up with the user. Additionally, the chatbot was capable of database connection, so we enabled an Azure Cosmos DB interface to log conversations and tagged login users in search logs.
Exiting Thoughts
Our journey in developing a chatbot with Azure AI and other technologies has been both rewarding and challenging. By combining these tools, we’ve built a robust solution that boosts customer engagement and delivers meaningful insights into TCB’s data. We’re excited to keep innovating and strengthening our digital presence. This isn’t the first project we’ve completed for TCB, and we’re grateful for their continued trust in us to bring such an exciting project to life.