How We Created an AI-Powered Chatbot for an Insurance Company

3 min read
Mar 20, 2024 11:32:47 AM
Azure OpenAI and Cognitive Search for IA-Powered Chatbots
5:33

In the digital era, a company's ability to effectively manage and process its information has become a cornerstone of its success. Such processes can see significant improvements thanks to Artificial Intelligence. In this success case, we share how the implementation of Azure OpenAI and Cognitive Search, combined with innovative strategies, revolutionized the accessibility and usefulness of information for one of our allies in the insurance sector.

Project Start: The Need of an Intelligent Portfolio System

Our ally has national coverage and, with a focus on the insurance sector, also offers a wide range of products that are subject to frequent updates. 

Our collaboration began with a vision: To develop an intelligent portfolio system tailored for the insurance company's employees, providing precise, real-time insights into their product offerings.

These intelligent portfolios needed to be accessed synchronously, revolutionizing the way the company stored, accessed, retrieved, and processed information.

The goal was to transcend mere data storage, delivering a solution that granted employees intuitive access and comprehensive analysis of the information found in brochures and portfolios, leading to the creation of an AI-powered chatbot.

Thumbnails-GenAI-Blog

Initial Implementation

The initial phase of the project involved uploading essential documents into an Azure Blob Storage container. This approach not only ensures security and scalability but also establishes a foundation for more accessible and controlled data management.

These documents underwent indexing, utilizing form identification tools to categorize and store each division in a designated space.

Furthermore, indexers were developed to automatically index new information stored in the container or any modifications made to existing data.

Model Deployment in Azure OpenAI

With the foundational knowledge in place, the next step entailed carefully selecting and deploying advanced OpenAI models within the Azure OpenAI infrastructure. 

The models we used were calibrated, parameterized, and tested to establish the foundations of intelligent information processing and shape user interaction.

Integration of Vector Search

A pivotal aspect of the project involved integrating a text embedding-ada model.

This integration empowered us to implement vector searches, utilizing Machine Learning to grasp the meaning and context of information and convert it into a numerical representation. 

The model processed each information group stored in the index, embedding it into the generative AI model within the Azure OpenAI instance, and vectorized its data.

This advanced technology enabled us to initiate complex natural language queries, resulting in responses characterized by a high degree of relevance and precision.

Creating a Bot Using Python

The first version of the bot was developed in Python, leveraging the versatility and power of this programming language.

This initial bot showed significant promise, laying the groundwork for a more sophisticated tool tailored to user needs. It underwent initial user interaction tests and response behavior assessments. Although rudimentary in nature, it paved the way for subsequent improvements and marked the starting point toward achieving the ultimate goal.

Improvements in User Experience: A Leap with Azure Chat Solution Accelerator

With the goal of enhancing user experience, a strategic decision was made to transition and expand the existing solution to the Azure Chat Solution Accelerator powered by Azure Open AI Service. This transition signified a substantial enhancement in terms of functionality, aesthetics, and adaptability.

The Azure Chat Solution Accelerator facilitated the integration of advanced features, including contextual conversation management, a more defined corporate identity, and deeper integration with knowledge sourced from archived documents.

Moreover, this tool facilitated the utilization of threads and historical data across unrelated databases. It also enabled users to directly upload their documents from the chat window and engage with the extracted information.

This improvement not only enhanced the user interface but also elevated the quality and precision of interactions, while expanding the range of product applications.

Impact on End Users and Business Benefits

The intuitive and efficient chatbot has revolutionized the user experience, ensuring swift and precise responses to complex queries within an appealing user interface. This grants users immediate access to pertinent, reliable, and comprehensive information.

Furthermore, this advanced intelligent portfolio system has not only enhanced operational efficiency by providing users with insights into the current product characteristics, but has also expanded its scope. It now encompasses a wealth of information beyond mere portfolios, including regulations and policies.

The ability to provide quick and accurate responses to user inquiries has resulted in a new approach to staying updated and informed about the company's products and conditions.

Are you seeking a partner to boost your data analytics or Artificial Intelligence initiatives forward?

Thumbnails-GenAI-Blog

At Pragma, we are certified partners of leading technology firms and stand prepared to boost your projects with top-tier IT talent.

contact us

 

Subscribe to
Pragma Blog

You will receive a monthly selection of our content on Digital Transformation.

Imagen form