Dun & Bradstreet Reinvents Commercial Graph for AI Agents

Dun & Bradstreet (D&B), with a legacy spanning over 180 years, has recently transformed its Commercial Graph, which now includes an impressive 642 million businesses. This strategic shift responds to the increasing need for AI agents to access and interact with data more efficiently than traditional human analysts. The challenge emerged as D&B's clients began incorporating AI into workflows across credit, procurement, and supply chain sectors, revealing the limitations of a system originally designed for human use.

The Challenge of Fragmented Data

Historically, D&B's Commercial Graph consisted of a patchwork of separate systems developed over decades for various markets and use cases. While this architecture effectively supported human analysts using SQL queries, it became cumbersome for AI agents that require rapid, real-time data access. The database's growth—nearly doubling from 300 million to over 642 million records in just five years—exacerbated the challenges, as agents needed sub-second query responses from a fragmented infrastructure.

Moreover, the relationships tracked by the database were static. Legacy systems linked a CEO to a company without accounting for dynamic changes like leadership transitions or ownership shifts. AI agents conducting tasks such as credit assessments need a more fluid understanding of relationships, as they cannot afford the delays that come with human-driven analysis. Gary Kotovets, D&B's Chief Data and Analytics Officer, emphasized, "We need to think about agents as our new consumer category, evolving from our standard credit analysts or sales and marketing professionals to also now catering to these customers' agents."

A New Foundation for AI

To address these challenges, D&B began a complete overhaul of its data architecture, migrating to a cloud infrastructure that enabled the consolidation of fragmented databases. This transition resulted in the creation of a unified knowledge graph capable of continuously updating and enriching itself through AI-driven data processing while meeting regional compliance requirements. The enhanced graph now tracks billions of relationships across its extensive business records.

Along with the knowledge graph, D&B established a structured access layer designed for AI agents. The traditional SQL access proved inadequate for the volume and latency demands of agent queries. Instead, D&B implemented a suite of tools through its MCP (Managed Cloud Platform) that packages data with context, allowing agents to efficiently access the necessary records. A match and entity resolution engine further improves this process, ensuring that agents receive verified and specific identity information for their queries.

Addressing Identity Verification

An equally important aspect of the upgrade involved rethinking agent authentication. The legacy model focused on human users was incompatible with machine identities. To address this, D&B introduced a new registration system for agents that requires mapping to a verified IP address and registering an individual access key. This approach resembles the 'Know Your Customer' model used in traditional finance. Kotovets elaborated, "We actually have a concept of Know Your Agent, similar to know your customer, that does those additional verifications."

This dual strategy of verifying agent identities and ensuring data consistency across workflows tackles what Kotovets describes as the outbound problem, where multiple agents may lose track of the entity being analyzed in collaborative workflows. To mitigate this, D&B's business verification agent can be integrated into various workflows, serving as a persistent reference point to ensure all agents are discussing the same entity—similar to a digital handshake between systems.

Lessons for Enterprises

D&B's extensive overhaul presents several key considerations for enterprises aiming to implement AI agents effectively. First, a solid data foundation is essential; many CDOs and CIOs have noted that AI initiatives can falter without clean, normalized, and consolidated data. Second, systems must accommodate dynamic relationships instead of static connections, allowing agents to adapt to changing business environments.

Third, incorporating entity consistency checks into multi-agent workflows is vital to prevent discrepancies when different agents interact with the same data. Finally, establishing a clear lineage for data insights from the beginning is crucial for maintaining accuracy and accountability, especially in sectors where mistakes can have serious consequences.

D&B's transformation of its Commercial Graph not only addresses the immediate challenges posed by AI integrations but also sets a standard for how companies can adapt their data architectures to meet the demands of an evolving digital landscape. As the need for AI-driven solutions grows, the insights gained from D&B's experience will resonate across the industry, influencing the future of enterprise data management and AI.