Announcing the OCI Generative AI Agents RAG service

January 23, 2024 | 4 minute read
Barry Mostert
Senior Director, Artificial Intelligence and Analytics
Text Size 100%:

We’re excited to introduce the beta availability of Oracle Cloud Infrastructure (OCI) Generative AI Agents Retrieval-Augmented Generation (RAG) service, your organization’s own customizable solution for conversing with and acting on diverse knowledge bases. The RAG service is the first of a series of AI Agents, with an focus on OpenSearch.  Upcoming releases are expected to support a wider range of large language models (LLMs) and provide access to Oracle Database 23c with AI Vector Search and MySQL HeatWave with Vector Store.

We’ve all heard the buzz around ChatGPT and other LLMs that can understand and respond to natural language requests. Now, Oracle brings similar capabilities purpose-built for enterprises, by combining user-friendly conversational interfaces with seamless real-time access to corporate knowledge bases.

For example, OCI Generative AI Agents RAG service allows employees to ask questions and get insights from siloed sources across your company through natural conversations. The RAG service performs a semantic similarity match between the user’s request and the documents in a centralized embeddings vector store. The embeddings vector store can contain data from a variety of corporate systems, such as enterprise resource planning (ERP), Human Capital Management (HCM), Customer Experience (CX), and data lakes, and query multiple data silos simultaneously. Users aren’t required to understand how or know where that data may be stored. The service will deliver results using the latest up-to-date information – even on dynamic data stores –  not just static reports or outdated indexes.

The technology

At the core of the solution is a large language model –  Meta’s  Llama2 or Cohere’s Command  –  that allows natural back-and-forth conversations with users. Think of it like a helpful colleague you can ask business questions in plain language. This is supplemented by the embeddings model that generates embeddings vectors for the corpus of corporate documents stored in the embeddings vector database. The retrieval-augmented generation (RAG) system is a combination of the vector database, embeddings model, and text generation model to synthesize grounded responses based on the documents retrieved. Instead of simply displaying the list of documents found, the AI service uses the text generation model to produce a natural language response that answers the original question, supported with references and links to original source documents. In the future, users will be able to not only retrieve information, but take immediate action such as sending emails, contacting customers, or managing orders directly from the same interface without needing to switch to other applications.

Key benefits

RAG provides the following benefits:

  • Democratizing data access: The OCI Generative AI Agents RAG service removes barriers to using enterprise data strategically. You don’t have to rely on specialist teams with advanced skills to query different corporate data stores. Users only need write their questions in plain language.
  • Up-to-date: The OCI Generative AI Agents RAG service retrieves real-time data to answer questions on the fly, instead of relying on batched updates or secondary search indexes. Responses are based on the current data available in the vector database, meaning cut-off dates on training data sets become irrelevant.
  • Understandable contextual results: The conversational experience provides information in consumable natural language, not just raw data. Relevant grounded results include context about sources and references to underlying materials so you can validate the results as needed.
  • User productivity gains: The OCI Generative AI Agents RAG service is designed to deliver productivity gains for business users, with fast, helpful answers fast from corporate systems without having to understand how to query them. 

Potential use cases

RAG offers different use cases for different industries and departments, such as the following examples:

  • IT: Technicians can get answers to system troubleshooting issues such as “What steps resolve the account creation error code I’m seeing?” faster by querying knowledge bases conversationally instead of combing through wikis and manuals.  
  • Sales: Account managers can get customer purchase histories and trends by asking natural questions, such as “How many orders has Acme Inc. placed this year?”, instead of running reports.  
  • Telco: Customer support associates can get problem resolution specifics quickly through natural chat questions, such as “What is the typical fix for intermittent internet with Acme fiber router model X483b?”, instead of performing lookups across multiple systems.  
  • Insurance: Claims adjusters can simplify investigating policy and history details by querying “How many claims has this customer filed in the last 3 years?”, rather than pulling data from various policy and account systems.  

Conclusion

The possibilities are vast for how this RAG service could transform knowledge acquisition and data-driven decision making. Whether it’s increasing sales opportunities, optimizing support resources, or simply empowering any employee with fast access to helpful information, the OCI Generative AI Agents RAG service can improve business productivity and customer experience. 

For more information see the following resources:

Barry Mostert

Senior Director, Artificial Intelligence and Analytics

Barry is a senior director for product marketing covering Oracle's AI and Analytics services.


Previous Post

Generative AI made easy in OCI Data Science

Tzvi Keisar | 3 min read

Next Post


Guide to prompt engineering: Translating natural language to SQL with Llama 2

Anshuk Pal Chaudhuri | 8 min read