Today, we’re expanding the Contextual AI Platform to support additional unstructured modalities alongside structured data sources. With the latest enhancements, enterprises can build specialized RAG agents capable of actively retrieving and reasoning across their entire knowledge base, unlocking higher-value AI use cases that require analyzing unstructured and structured data together.

The platform’s document understanding pipeline can now process and comprehend complex visuals, such as charts and diagrams, in unstructured data sources. This capability allows organizations to capture valuable statistical insights and abstract concepts that complement traditional document text, providing a more complete understanding of media-rich assets like PDFs and HTML.

We’ve also implemented new integrations with major structured data platforms—BigQuery, Snowflake, Redshift, and Postgres. These integrations give agents direct access to mission-critical business data, enabling more accurate responses to queries that combine insights from both structured and unstructured sources.

START FREE TRIAL

Bridging Unstructured and Structured Data

Enterprise knowledge today exists in a fragmented landscape, scattered across various data sources ranging from unstructured documents to structured databases to SaaS applications. For enterprise AI to deliver meaningful value, it must excel at processing and understanding data across all domains. However, traditional RAG systems often only support a subset of data types, failing to accurately answer questions that require analyzing multiple sources.

The Contextual AI Platform solves for these challenges by providing broad support for unstructured and structured data within a single, unified platform. Our comprehensive approach eliminates the need for organizations to manually piece together disparate components for different data types, while also delivering state-of-the-art performance out of the box.

With specialized RAG agents that can accurately reason over both multimodal documents and structured data, organizations can unlock more ambitious and higher-value AI use cases:

1. Financial Research and Analysis: Generate deeper market insights by analyzing charts and qualitative commentary in SEC filings and industry reports alongside quantitative trends in portfolio and securities data.
2. Customer Support: Deliver personalized solutions and faster issue resolution by connecting customer usage metrics with relevant troubleshooting instructions and diagrams from product documentation.
3. Sales Intelligence: Accelerate preparation for customer meetings by accessing a unified view of the customer’s transaction history with insights from past call transcripts, email threads, and proposal documents.

Enhanced Capabilities for Unstructured Data

With upwards of 90% of enterprise data existing in unstructured formats, the ability to effectively parse and understand complex, multimodal documents is a critical requirement for enterprise AI systems. We’re introducing several key capabilities to address this need.

The cornerstone of our latest platform update is the new visual data analysis capability. This advancement enables specialized RAG agents to interpret charts and diagrams with exceptional accuracy, extracting quantitative findings and logical relationships that were previously locked away in visual formats. By combining the insights with standard document text, organizations can now access a richer understanding of their content.

We’ve also introduced several quality-of-life enhancements for supporting unstructured data sources. Agents can now embed reference URLs and hyperlinks retrieved from source documents directly in their responses, allowing end-users to quickly access contextual information beyond the document itself. Additionally, we’ve expanded our document extraction capabilities to support HTML files, complementing our existing PDF support.

Powerful Structured Data Integrations

While unstructured data makes up the majority of enterprise knowledge, structured data sources contain vital information around transactions, customers, products, operations, and other critical components of the business. With new integrations for BigQuery, Snowflake, Redshift, and Postgres, specialized RAG agents built on the Contextual AI Platform can now seamlessly access data from popular data warehouses and databases.

Using advanced text-to-SQL retrieval capabilities, agents can generate precise SQL queries that understand complex schema relationships and combine the output with insights from unstructured documents. This is especially powerful when complementing quantitative analysis (the “what”) with nuanced explanations and commentary around the results (the “why”).

With integrations for popular structured data platforms, specialized RAG agents can perform advanced quantitative analyses and combine them with insights from unstructured documents

Text-to-SQL retrieval also democratizes data access to non-technical users, enabling them to query databases and perform sophisticated quantitative analysis through natural language. By bridging the gap between business users and valuable data, enterprises can accelerate decision-making, generate new insights, and reduce the burden on data science teams.

Looking Ahead: Private Preview Features

Our commitment to expanding the platform’s multimodal capabilities continues with several exciting upcoming features. Currently in private preview, these additions to the Contextual AI Platform will further enhance our coverage of input modalities and provide access to new end-users.

• SaaS integrations: Massive volumes of critical knowledge are shared everyday in popular enterprise SaaS tools like Slack, Google Drive, Sharepoint, GitHub, and more. With seamless access to these applications, specialized RAG agents can provide responses grounded in the most up-to-date work happening in your business.

Retrieve valuable data from popular SaaS applications like Slack, Google Drive, Github, and more

• Support for circuit diagrams: For technology and engineering companies, agents that understand complex technical schematics have the potential to drive significant productivity gains and unlock new use cases. We plan to soon add support for interpreting circuit diagrams in unstructured documents to best serve these customers.
• Multilingual support: Lastly, we’re expanding our platform’s reach with multilingual support for both query input and response output. We plan to start with Mandarin and quickly follow with additional languages to broaden accessibility to new users.

Stay tuned for more updates as we continue to enhance the Contextual AI Platform and help organizations build powerful, versatile, and comprehensive enterprise AI solutions.

START FREE TRIAL