Case Study: Conference FAQ- AI Chatbot

Problem: Responding to conference online visitor questions 

An ambitious technology centric conference like IIT2024 Global Conference experiences a streaming volume of online visitor’s queries, often repetitive, on the topics of conference speakers, panels, venue, entertainment, food, and more.

Since it is not practical to have volunteers answer calls 24x7, the quest was to find a best practice for compiling a knowledge base and responding to FAQs in a timely manner.

Solution: Large Language Model (LLM) with Advanced Natural Language Processing (NLP) Capabilities for Human-Like Interaction. Implementing AI FAQ Chatbot 

Avatara AI’s team selected OpenAI’s ChatGPT as the Large Language Model (LLM) foundation. We upped the knowledge base in the LLM with a feed from the conference database of speakers, schedules, topics, and images into an external database. Additionally, potential questions related to the conference were added- venue details of the conference center, nearby attractions, local restaurants, transportation, etc.

Next, the AI engineering team used Retrieval Augmented Generation (RAG) to acquire context-relevant information from the external database and present it to ChatGPT when someone asked a query, to generate a response. Thus keeping the FAQ relevant and tuned-in to the context of the IIT 2024 conference.

We tested the FAQ AI by asking it several questions and observing if the response was accurate. It passed our testing criteria and was deployed on the conference website as a FAQ AI chatbot.

Outcome: Rapid, accurate, visitor self-service for FAQs

Website visitors used the AI chatbot frequently, especially as the days to the conference came closer. 

This was an immense relief to the volunteers handling phone queries, as their visitor enquiries diminished in volume. Many website visitors were able to self-serve via the AI chatbot. 

Since the chatbot had been fed with contextual data, the accuracy of the answers were satisfactory, and the online visitor did not have to follow up with a phone call for clarifications.

Problem: Managing a large retail store for inventory, staff, and customer behavior 

Managing a retail store with a large floor space requires a substantial number of staff. Stores with multiple product classes like hypermarkets and clothing stores have difficulty checking and updating inventory. Manual monitoring of staff by managers is time consuming and prone to error. Understanding custom behaviour and trends across stores becomes impossible with just ERP and POS data

Solution:Computer Vision based Video Analytics

Avatara AI’s team enabled complete analytics and real-time alerts of the entire store using just video data from cameras placed inside the store. Store managers got real-time product updates and inventory update notifications based on monitoring racks. Data was displayed on visual dashboards for continuous feedback.

Outcome: Continuous, accurate monitoring. Trend predictions.

The video analytics solution offered continuous monitoring of inventory and predicted trends in consumption of the inventory. The store managers could now rely on accuracy of inventory assessment and get knowledgeable on trends such as best selling items, slow moving items, etc.