Bud Ecosystem, FoundationFlow, and Intel built a custom Retrieval-Augmented Generation chatbot for product-catalogue support — running on Intel® Xeon® processors with the BudServe engine. It delivers GPU-comparable performance on CPU, with roughly 60% better catalogue-query handling.
Customer support is increasingly challenged to meet rising expectations for quick, accurate, and personalised responses — particularly with complex product queries. Bud Ecosystem, FoundationFlow, and Intel developed an advanced RAG-based chatbot, powered by Intel® Xeon® processors, to enhance accuracy and efficiency in handling product-catalogue inquiries.
The solution integrates external knowledge sources, provides direct access to product information, and adds improved table-parsing capabilities — with the BudServe engine leveraging Intel® Xeon® processors for superior performance.
Together these limitations hinder the effectiveness and efficiency of traditional RAG chatbots — and open the door to a custom, Intel-optimised approach.
The partners built a custom RAG solution to enhance context understanding, improve response times and accuracy, and create a more efficient support system. Leveraging FoundationFlow’s Co-pilot Builder module and Intel® Xeon® processors, the chatbot integrates seamlessly with external knowledge sources for precise, informative answers.
The system bypasses manual data entry and separate database integrations by directly accessing extensive product catalogues and documents — so the chatbot always has up-to-date specifications and details.
Advanced parsing and interpretation of tables within product catalogues lets the chatbot accurately address complex queries about features and configurations — overcoming a key limitation of traditional LLM systems.
Fine-tuned for Intel® Xeon® processors, the RAG system leverages their high-performance processing and scalability for fast responses and smooth operation — even at high inquiry volume — while keeping infrastructure costs efficient.
Bud Ecosystem’s proprietary inference engine significantly boosts the performance of the overall stack — delivering faster response times and superior performance even under high user loads. It supports less-compute devices while providing performance comparable to GPU solutions.
The architecture is optimised for high concurrency, ensuring smooth operation even during peak usage, and produces an average output of 125 tokens per second — delivering faster responses and a more responsive support experience.
Superior performance versus leading cloud chatbots — with measurable gains in catalogue handling, throughput, and support quality.
The chatbot solution demonstrated superior performance compared to leading cloud-based solutions — including Google Gemini Advanced, ChatGPT Plus, and Perplexity.
A nearly 60% improvement in handling product-catalogue queries, and an average of 124.89 transactions per second — high efficiency and speed across a large volume of transactions.
Together these enhancements streamline customer-support operations — reducing customer frustration and improving the overall experience, with a scalable, cost-efficient approach that runs on infrastructure enterprises already own.