Editorial illustration for D&B rebuilds 642 million‑business database after AI agents hit limits
D&B rebuilds 642 million‑business database after AI...
Why did D&B have to start from scratch? The answer lies in a data architecture that was never meant for autonomous agents. The Commercial Graph, a patchwork of separate systems, let human analysts stitch together SQL queries or use pre‑built interfaces; an AI agent could not.
Over five years the repository swelled from just over 300 million to 642 million business records, each with roughly 11,000 fields. D&B now runs about 100 billion data‑quality checks each month as those records flow through the network.
But scale wasn’t the only hurdle. Legacy tables captured only static links—a CEO tied to a firm, a subsidiary’s ownership frozen in time. Credit‑risk or third‑party‑risk assessments need dynamic connections: when a CEO jumps ship, where does their track record go?
When a subsidiary changes hands, how does that ripple through the hierarchy? Those questions once required custom analyst work, something an agent can’t wait for.
The problem isn’t unique to D&B. According to Kotovets, dozens of CDOs and CIOs have told him their AI ambitions stall because their data foundations aren’t standardized, normalized, or agent‑queryable. D&B had that foundation—until it broke.
Agents could not.
The scale of the underlying data compounded the problem. The database had nearly doubled in five years, expanding from more than 300 million to more than 642 million business records, with 11,000 fields per record, according to D&B. The firm now runs approximately 100 billion data quality checks per month as records move through its systems. Querying that at the sub-second latency agents require, against a fragmented architecture, was not workable.
The relationships the graph tracked were also the wrong kind. Agents working on credit assessments or third-party risk need dynamic relationships: when that CEO leaves for a new company, which organization does their track record follow? When a subsidiary changes ownership, how does that propagate across a corporate hierarchy? Agents cannot wait for custom analyst work.
The broader problem is not unique to D&B. Kotovets said he has spoken with hundreds of CDOs and CIOs over the past six months and consistently heard the same constraint: they could not build what they wanted in AI because their data foundations were not standardized, normalized or agent-queryable. D&B had that foundation, built over decades to serve human analysts.
Why this matters
We see D&B’s decision to rebuild a 642‑million‑record database as a reminder that legacy data stacks often assume human users. The Commercial Graph was a patchwork of systems, held together by custom integrations that analysts could stitch together with SQL or UI tools. Agents could not.
As the record count swelled from 300 million to 642 million, with 11,000 fields per entry, the fragmentation grew harder to mask. D&B now runs roughly 100 billion queries on the new platform, a volume that suggests a deliberate shift toward an architecture optimized for agent‑friendly access and higher throughput. For developers, this underscores the need to expose clean, unified APIs when scaling data products.
Founders should ask whether their own data pipelines can survive a surge of autonomous queries. Researchers may find a more predictable source for large‑scale business analytics, yet it remains unclear whether the rebuilt system will handle future growth without similar limits. In short, building for machines from the start appears less optional than before.
Further Reading
- The year companies stop building AI agents and start running them - IBM
- Enterprise AI Agents Are Failing — And Your Data Layer Is to Blame - Airbyte / YouTube
- AI Agents Are No Longer Experimental -They're Becoming Your Workforce - T2C Online
- Dun & Bradstreet Signals New Era for Enterprise AI with Launch of D&B.AI Suite of Capabilities - PR Newswire
- Databricks: Only 19% of Organizations Have Deployed AI Agents but They're Already Creating 97% of Databases - SaaStr
Related Reading
Open Source SDOF Adds Two Defensive Layers via Intent Router and StateAwareDisp
Open Source Cohere launches Command A+ 218B sparse MoE, runs on two H100 GPUs using QAD
Open Source Trump cracks down on Anthropic after Amazon tip; staff largely foreign
From Archives Experimental MLX Delegate Enables PyTorch Models on Apple Silicon GPUs
From Archives PyTorch Docathon 2026 merges 150+ PRs, adds API and ExecuTorch docs
Common Questions Answered
Why did Dun & Bradstreet need to rebuild their Commercial Graph database?
D&B's original Commercial Graph was a patchwork of separate systems designed for human analysts using SQL queries and pre-built interfaces, but AI agents could not navigate this fragmented architecture. As the database grew from 300 million to 642 million business records with 11,000 fields per entry, the complexity became impossible for autonomous agents to work with effectively.
What limitations did AI agents encounter with D&B's legacy data architecture?
AI agents could not stitch together the custom integrations and SQL queries that human analysts used to access data across the patchwork of separate systems in the Commercial Graph. The fragmented structure that worked for human users proved incompatible with autonomous agent workflows, forcing D&B to completely rebuild their data infrastructure.
How many business records does D&B's new database platform handle?
D&B's rebuilt database now manages 642 million business records with 11,000 fields per entry, processing roughly 100 billion queries on the new platform. This represents significant growth from the original 300 million records that existed over five years prior to the rebuild.
What does D&B's database rebuild reveal about legacy data stacks?
D&B's experience demonstrates that legacy data stacks are typically built with human users in mind, relying on custom integrations and manual query construction that autonomous agents cannot replicate. As organizations scale their data infrastructure and adopt AI agents, they must fundamentally rearchitect systems that were never designed for autonomous access patterns.
Further Reading
- The year companies stop building AI agents and start running them — IBM
- Enterprise AI Agents Are Failing — And Your Data Layer Is to Blame — Airbyte / YouTube
- AI Agents Are No Longer Experimental -They're Becoming Your Workforce — T2C Online
- Dun & Bradstreet Signals New Era for Enterprise AI with Launch of D&B.AI Suite of Capabilities — PR Newswire
- Databricks: Only 19% of Organizations Have Deployed AI Agents but They're Already Creating 97% of Databases — SaaStr
