ABM.dev Gives AI Agents Per-Field Confidence Scores to Prevent Acting on Bad Data

A new B2B data enrichment platform called abm.dev is designed to prevent AI agents from acting on unverified or inaccurate information by attaching a source, confidence score, and verification status to every individual data field it returns. The platform aggregates data from ten providers — including LinkedIn, Companies House, and Hunter — under a single API call and unified schema, covering 89 fields across person and company records. When sources disagree on a value, the platform retains both results and flags the conflict rather than silently choosing one. This allows AI agents to make rules-based decisions, such as acting on high-confidence data, holding on low-confidence data, or escalating uncertain records to a human reviewer. The developers acknowledge the tool has limited utility for companies with little public presence, but say it will explicitly signal uncertainty rather than return a confident but fabricated answer.