Every compliance team is being handed agents to work with, but few are handed verified data to run these agents.
KYB has always been the slow part of onboarding. Analysts pull a name, walk it through state filings, sanctions lists, address verification, entity variations. This takes hours per case. The friction is the point of the process. Humans notice gaps when the data doesn’t add up. They slow down and take a closer look when something seems wrong.
But agents don’t slow down. They return an answer with the same confidence whether the business is real, dissolved, synthetic, or matched to the wrong entity. The data quality problem analysts have always managed by hand becomes a systemic risk when you work at agent speed.
Enigma’s internal benchmark data shows the cost of skipping verification. The current generation of models from leading AI companies, paired with web search, had success rates between 37% and 51%. When paired with Enigma’s identity graph, even small models like Gemini Flash and Claude Haiku topped 70% accuracy on entity resolution, at one-tenth the inference cost.
The performance gap is about data quality, not model sophistication. A pizza franchise files as an LLC in Delaware, operates in Ohio, trades under three DBAs, and appears in merchant records under a fourth name. An agent without resolution sees four businesses. Or four errors. Enigma’s business identity graph knows it’s one company.
This verified identity graph is what Enigma’s GCP agent runs on. A compliance officer submits a business. The agent, powered by Gemini Flash, calls the Enigma KYB API, checks the graph, applies the team’s policy across name, address, state registration, watchlist screening, and TIN verification, and returns a verdict with check-by-check reasoning. Single business via chat. Batch via spreadsheet. Verification history persists across sessions, and organizations can register multiple agents tuned to different risk contexts.
The agent is built on Google’s Agent Development Kit and runs entirely on Google Cloud: Vertex AI Agent Engine, Gemini Flash, Cloud Run, Cloud Storage. It offers managed infrastructure, built-in authentication, observability through Cloud Trace, and it’s available through committed spend on the Google Cloud Marketplace, with no separate procurement. The agent also supports A2A, so it can plug into broader multi-agent compliance workflows.
The Enigma KYB Agent is built for every compliance team being asked to ship agents this year. The question that decides whether those agents work in production rests on the data the agent uses. Enigma has spent fifteen years building the verified data layer that makes the answer something a compliance team can act on.