D&B's Database of 642 Million Businesses Was Built for Humans, Not AI Agents. So They Rebuilt It.

Why It Matters

Enterprises can no longer rely on human‑optimized data stores; D&B’s shift demonstrates that AI‑ready, dynamic, and secure data foundations are essential for scaling automated decision‑making across risk and supply‑chain functions.

Key Takeaways

• D&B unified 642M records into a cloud‑native knowledge graph.
• New agent‑centric API delivers sub‑second query latency.
• ‘Know Your Agent’ framework adds IP and key verification.
• Dynamic relationship modeling enables real‑time corporate hierarchy updates.

Pulse Analysis

For more than a century, Dun & Bradstreet’s Commercial Graph has been the go‑to source for credit analysts, risk managers and sales teams, aggregating 642 million business entities and 11 000 data fields. The architecture, a patchwork of legacy systems stitched together with custom integrations, was optimized for human‑driven queries that could tolerate latency and manual entity matching. As enterprises began embedding AI agents into credit, procurement and supply‑chain workflows, those agents hit a wall: the fragmented, static data model could not deliver the sub‑second, context‑rich answers required for automated decision‑making.

To overcome the bottleneck, D&B migrated all data to a cloud‑native platform, collapsed the disparate stores into a single knowledge graph, and introduced a data‑fabric layer that normalizes records while respecting regional compliance. On top of this unified graph, the company launched a structured access layer—delivered through its MCP suite—that packages data with context and routes each request through an entity‑resolution engine. A new ‘Know Your Agent’ registration system ties every machine request to a verified IP and access key, ensuring both inbound and outbound identity verification.

The rebuild highlights four prerequisites for any organization looking to scale AI agents. First, a clean, consolidated data foundation must precede agent infrastructure; without it, AI projects stall. Second, systems need to capture dynamic relationships so agents can reason about changing hierarchies and roles. Third, multi‑agent workflows require built‑in entity‑consistency checks to prevent divergent records. Finally, provenance and lineage must be baked in from day one, giving users a transparent audit trail for every automated decision. D&B’s transformation offers a roadmap for enterprises seeking to turn data assets into agent‑ready intelligence.