Dun & Bradstreet Adds Agentic AI to Compliance Workflows

Dun & Bradstreet announced agentic AI capabilities for compliance and third-party risk workflows in a June 22 news release, presenting the capabilities inside its D&B Risk Analytics product and broader D&B.AI offering, the release said. The release states the new features run "fully agentically" via a Model Context Protocol (MCP) server and embed verified D&B data and workflows into AI assistants and custom agents, which the release said speeds onboarding, screening, and due diligence. The release claims these capabilities can reduce processing times by 70-90% and move tasks such as beneficial owner identification from days to seconds (PYMNTS; company release).

Dealroom reports early internal or pilot results of up to 20x increases in review capacity, 70-96% reductions in processing time, and 50-90% reductions in false positives, and notes availability through Claude with planned Microsoft Copilot and Google integrations (Dealroom).

Per prior coverage and company materials, the approach combines Dun & Bradstreet's verified business identifiers and the D&B Commercial Graph/ Data Cloud with agentic orchestration and API-level model connections; Gary Kotovets, chief data and analytics officer, told The AI Innovator the company's data footprint covers more than 600 million public and private companies and underpins the platform's trust claims (The AI Innovator).

Agentic systems typically require three components to move beyond prototypes: high-quality, canonical data; deterministic decisioning or verification steps; and safe, auditable action orchestration. Companies building comparable agentic flows often pair authoritative identifiers and graph links with model gateways like Claude or Copilot to constrain hallucination risk and speed lookups. Embedding provenance and deterministic checks into the agent loop reduces the surface for model-driven errors but raises engineering complexity around latency, caching, and data freshness.

Firms offering compliance automation confront two consistent pressures-rising regulatory complexity that increases screening scope, and constrained compliance headcount-which public reporting frames as a driver for automation (PYMNTS; Dealroom). Observers and vendors in this space emphasize verified, enterprise-grade data as the gating factor for production-grade agentic deployments rather than model capability alone (The AI Innovator).

If you are evaluating agentic tools for KYC/KYB, expect implementation trade-offs between speed and defensibility. Anchoring agents on authoritative registries, unique identifiers, and document verification reduces risk but requires engineering to preserve audit trails and proof of data lineage. Integrations with hosted model endpoints (for example Claude today and Copilot/Google connectors reported by Dealroom) simplify agent orchestration but add vendor dependency considerations.

Public signals to follow include independent validation of the vendor's efficiency and false-positive claims, customer case studies with documented audit trails, and the breadth of model integrations available via the reported MCP server. Also monitor regulatory responses where agentic decisioning touches sanctions screening, beneficial-ownership rules, or automated adverse-action determinations-areas that typically demand explainability and recorded human review.

Sources referenced in reporting and interviews include Dun & Bradstreet's June 22 release covered by PYMNTS, Dealroom's product report, and an interview with Gary Kotovets in The AI Innovator.