Driving smarter data enrichment: IBM and Tavily partner for Agentic AI solutions

As the first step in a strategic partnership with IBM, Tavily is proud to launch an open-source solution that brings live web insights into IBM watsonx.ai™ Granite models—transforming how enterprises enrich and interact with spreadsheets.

This launch is just the beginning. By combining the intelligence of IBM’s advanced LLMs with Tavily’s real-time data infrastructure, teams can build AI applications that are secure, production-ready, and grounded in current, trusted knowledge—unlocking deeper insight and reasoning at scale.

Enterprises need more than just powerful models—they need outputs grounded in reality, tailored to their context and compliant with evolving governance standards. That’s where this integration shines.

Tavily—a search platform optimized for LLMs and autonomous agents—enables real-time access to web data, closing the gap between static models and the dynamic world they operate in. Its powerful search, extraction and crawl APIs deliver high-fidelity web data in real time, designed from the ground up for seamless LLM integration.

By combining IBM watsonx.ai’s Granite foundation models with Tavily’s real-time web search APIs, organizations can:

1. Ground AI outputs in current and verifiable information, bridging the knowledge cutoff of static LLMs.
2. Enhance auditability and trust through transparent source citations for every enriched output.
3. Accelerate workflows at scale, with high-speed research and enrichment—powering agents and automation with real-time insights in seconds.Adapt to any enterprise environment, thanks to an open, flexible architecture that supports proprietary data, custom search parameters and model-agnostic design.

This integration reflects IBM’s commitment to building responsible, production-ready AI—and Tavily’s mission to make the web truly usable for intelligent systems.

Tavily offers a suite of web data APIs, including real-time web search, web scraping, web crawling and document summarization. Key features include:

• Real-time access and navigation of public web information
• Advanced extraction capabilities to retrieve and structure data
• Customizable search parameters including time-range, web domains, and more
• Privacy-first design with secure, compliant handling of data, zero data retention, and PII redaction
• Enterprise-grade security for detection and mitigation of prompt injection

By merging LLMs with real-time web data, enterprises can unlock numerous innovative use cases, including but not limited to:

1. Market research teams can quickly populate company profiles with live revenue estimates, market expansion news, leadership changes and relevant articles sourced from the web (see example here).
2. Finance teams can track real-time stock updates, competitor news or regulatory filings to stay ahead without switching between tools.
3. Business analysts can enrich CRM data with firmographic insights, enhancing lead qualification and personalization.

These use cases are just the beginning. To empower developers and teams to build and extend these workflows, we’ve made the project open-source and highly customizable.

This open-source project provides a reference architecture for enterprises aiming to implement agentic enrichment workflows at scale. Each column can be designed as a standalone agentic unit, enabling customized processing—from simple lookups to deep, contextual analysis—based on the specific requirements of the task.

The architecture is fully extensible, allowing teams to adapt it to industry-specific use cases, proprietary data sources, or internal compliance needs. We hope IBMers and the broader enterprise community draw inspiration from this work to build powerful, production-ready enrichment solutions across verticals. 

How it works:

1. Input Spreadsheet: Users begin by populating a spreadsheet with entities or prompts requiring enrichment.

2. Real-time Search: Tavily's API fetches fresh, reliable web data.

3. LLM Processing: IBM's Granite model processes, extracts, and structures the information

4. Enrichment and export: The enriched data is populated into the spreadsheet, complete with source citations, and can be exported as a CSV file.

Architecture Diagram:

The partnership between IBM and Tavily represents a significant step forward for intelligent data enrichment workflows. By combining Tavily's precision search APIs with the powerful natural language understanding of IBM's watsonx.ai Granite models, organizations can unlock smarter, real-time insights at scale directly within their daily spreadsheets.

Experience the power of this integration yourself by exploring the open-source project and demo available on GitHub.

Demo Video

Explore the GitHub Project

Tavily Product Overview