Graph-based RAG | WRITER Knowledge Graph
Salesforce case study
How Salesforce scales quality with AI apps
How Salesforce uses Writer to build AI apps that scale quality and deliver results
KNOWLEDGE GRAPH
An innovative approach to knowledge retrieval
Knowledge Graph, our graph-based retrieval-augmented generation (RAG), achieves higher accuracy than traditional RAG approaches that use vector retrieval.
HOW IT WORKS
Richer semantic understanding
Knowledge Graph draws on a specialized LLM that’s trained to process data at scale and build valuable semantic relationships between data points. It stores data in a cost-effective, easy-to-update graph structure.
The alternative approach
By converting data into vector embeddings, traditional RAG can only define similarity by distance between data points but has no context on their semantic relationships. Vector databases are also difficult and costly to maintain and update.
Accurate retrieval methodology
Because graph structures retain semantic relationships, Knowledge Graph accurately retrieves relevant data for each query. Our retrieval-aware compression technique condenses data and indexes it with metadata, which gives it rich context.
The alternative approach
Traditional RAG converts the query into a vector embedding and uses a rough algorithm to find the closest data points to the query, without any understanding of the relationship between the data points. When data is dense, this method fails to return the most relevant data consistently.
State-of-the-art LLMs
To generate a response, Knowledge Graph sends relevant data to our Palmyra LLMs, which are top-ranked and trained with 1 trillion tokens of quality data. We apply advanced techniques to enhance performance and minimize hallucinations.
The alternative approach
The quality of the answer depends on the quality of the retrieval, and the level of hallucination depends on the quality of the underlying LLM and the techniques you employ.
PERFORMANCE
Knowledge Graph achieves unmatched accuracy
In a benchmarking study, WRITER Knowledge Graph achieved top scores on RobustQA, which measures accuracy in open-domain question-answering, outperforming seven popular RAG approaches that use vector retrieval.
USE CASES
Build digital assistants you can trust
Knowledge Graph anchors your generative AI solutions in your company knowledge. Create expert assistants for any use case and be confident that your people are getting the correct information.
Nitin Tandon
CIO of Vanguard
Anna Griffin
Chief Market Officer
Ajay Dhaul
Founder & CEO, Omnea Ventures
DIFFERENTIATORS
Designed to meet enterprise requirements
Excels at advanced tasks
Knowledge Graph supports multi-hop questions, handles complex data formats, and produces fewer hallucinations.
Scales with enterprise data
Unlike traditional RAG, Knowledge Graph excels at retrieval with concentrated data, and updating data is fast, easy, and inexpensive.
Provides explainable AI
Knowledge Graph shows thought process, decomposes broad questions into subquestions, and provides specific source citations.
Supports your file types
Knowledge Graph handles structured and unstructured data, including spreadsheets, docs, charts, presentations, PDFs, and more.
BENEFITS
Deploy with confidence with integrated RAG
Rather than stitch together your own RAG, WRITER Knowledge Graph makes it easy to build high-quality agents.
Maintain efficient costs
We’ve seen 67% lower cost than leading alternative approaches when you use WRITER instead of traditional RAG, with savings increasing as you scale.
Increase your speed
Integrate data sources quickly. Our full-stack platform of LLMs, Knowledge Graph, AI guardrails, and a flexible application layer makes it easy to deploy in days.
Stay secure and compliant
Secure your data and leverage infrastructure, which enables you to meet your compliance obligations. Our full-stack platform maximizes security, and we do not train on your data.