CLEARWATER, Fla. — Sept. 3, 2026 — HPCwire reports that Dun & Bradstreet is connecting its commercial graph to Perplexity so enterprise teams can run research, risk, and onboarding inside an answer engine.

“With Dun & Bradstreet and Perplexity, clients can replace lengthy, fragmented manual onboarding processes with one agentic workflow, allowing them to onboard corporate clients in minutes with automated identity verification, risk insights, and audit-ready documentation.”

— HPCwire, quoting the Dun & Bradstreet announcement

Jeff Grimes of Perplexity told the same report that tying the graph to Perplexity Computer gives users “deeper business context to make faster, more confident decisions.” HPCwire also records the scale Dun & Bradstreet claims for that graph: more than 650 million global business entities.

What the deal actually is

This is a paid commercial-data pipe inside an answer product. It is built for banks, insurers, and procurement teams that already buy Dun & Bradstreet. It is not a public page a roofing company can send to ChatGPT.

AI Verified did not commission the story and does not restack Dun & Bradstreet’s entity count. The newsroom point is narrower: retrieval with a vendor graph is still retrieval. The buyer of that graph is the enterprise sitting in Perplexity, not the small firm being described.

Why small firms should care anyway

When an answer engine names a local contractor, it is usually reading Maps, a review site, or a messy website. It is not opening the Dun & Bradstreet terminal. Those two layers get talked about as if they were one product called “AI visibility.” They are not.

A public identity file — legal name, website the owner controls, phone, dated evidence — is the object a model can fetch without a commercial-graph contract. Gold is that file after the owner proves the domain and, when they have them, the government numbers. It is not a seat inside Perplexity’s enterprise workflow.

Read the source: Dun & Bradstreet Brings the D&B Commercial Graph to Perplexity (HPCwire / BigDATAwire, Aug. 27, 2026).