The Challenge
The client is an employee-owned chain of supermarkets in the midwestern and southern United States, with more than 280 locations. The company also publishes a popular magazine (in PDF format) for an online audience of millions of consumers. However, the company did not have an efficient search engine that allowed readers to search and access the contents of 110+ unstructured PDF files – consequently, basic search functionality of their website limited the value of the content library to customers.
The grocery chain needed to develop and deploy an AI-powered chatbot on its website that could enable end users to query the PDF library for details on recipes, ingredients, articles, and other useful information. Each of the hundred-plus PDF files was approximately 100 pages in length, so the total amount of data was substantial. In addition, data types varied – some PDF data was in the form of image files, while others contained rotated, circular or tabular designs. All of this presented a significant technical challenge, unlike searching flat, text-based content.
The Solution
Persistent, leveraging its longstanding relationship with Google Cloud, recommended leveraging the Google Foundation LLMs (Gemini / PaLM2)embedding model and text model to implement a Retrieval Augmented Generation (RAG) algorithm for searching the content in the PDF files. RAG improves the quality of Generative AI (GenAI) by combining retrieval and generation large language models (LLMs). Working together, the dual models can search large volumes of data and then generate readable text responses. With this capability, RAG enables LLM-powered chatbots to provide personalized answers to customers without the need to retrain the models. Persistent also employed its NOVA accelerator to build the solution faster.
Persistent created and deployed a GenAI-based chatbot on the firm’s website to process customer queries and deliver pinpoint responses. The chatbot was intelligent and contextually aware, for example, offering a “memory” feature to retrieve information from the previous conversation. The cloud-based solution leveraged a wide range of Google technologies, including Vertex AI (as a development platform) and Cloud Run, a fully managed platform that enables customers to run their code on top of Google’s scalable infrastructure.
The Outcome
The highly efficient GenAI-based chatbot decreased product search times by ~20%, with faster and more accurate returns to consumers. The capacity of the system to deliver pinpointed sources and images led to enhanced product discovery, increased revenue, and increased customer acquisition for the grocery chain.
Specifically, the firm realized an increased order value per customer of ~25%, quickly recouping the cost of deploying the solution. The chain’s first deployment on the Google Cloud Platform (GCP) has opened the door for future developments – such as analytics, personalization, and targeted marketing – which promise additional gains and success.
Persistent recently announced a Strategic Partnership Agreement with Google Cloud to jointly deliver market-leading solutions that help enterprises maximize the ROI from their cloud investments while modernizing their infrastructure and data stack. Contact us today for more details.
