Verifying identity is a hard problem, and it has been the core of our work for fifteen years. With the rise of AI agents, it is only getting harder. Enigma started with one belief: the public record should be a source of truth for what U.S. businesses are actually doing. Today, that work is part of Parallel.ai’s Index, a new platform for helping content and data providers share in the value AI agents create at the moment of inference.
With Index, the contribution our data makes at the moment of inference is measured and surfaced, not assumed. For work as hard to replicate as identity resolution across all U.S. businesses, that distinction matters.
Enigma does two things inside Index. We provide the substantive record of a business: linked legal entities, operating locations, ownership structure, officers, potential conflicts of interest. This tells the agent which business it is actually looking at before it starts. Generating tokens costs time and money. Every token an agent spends figuring out identity is a token that delays or even derails a substantive answer. We run air traffic control: the business is confirmed, the right connections are surfaced, and the false positives are cut before the work begins. That covers every U.S. business, not just the ones that show up in investment databases or news coverage.
A few weeks ago, I sat down with the Parallel team in Palo Alto to show them what this looks like in practice. We ran the same set of U.S. businesses through agentic web search on one side and Enigma’s knowledge graph on the other.
The quality difference came down to data coverage: Enigma sees U.S. business data that isn’t publicly available. In practice, that means fewer tokens spent on figuring out who you’re looking at, and more time doing the work. That’s what landed in the room.
What drove it home was during a live demo of a California-based company with a distinctive name ending in “Co.” In a bakeoff, every generic agentic search returned the same answer, a $15M flooring manufacturer in Minnesota. The California company is a real wholesale distributor, registered with the state, $177K in revenue. It just does not have a website. That is exactly the kind of gap that turns an expensive agent into a wrong answer. Without a verified identity layer, agents will not simply fail by saying “I don’t know.” They will fail by giving false information with persuasive assurance.
Now, when one of Parallel’s research agents shows up to Enigma’s business identity graph, it comes with whatever identifying information it has at hand: a name, an address, a phone number, sometimes only a description of what the business does. The first question is always the same: given what I know, what business is this really? The graph answers that question, and the many that follow.
None of this is figured out at runtime. A brand name and the LLC behind it might never appear together on the open web. Somebody has to do the connecting. What a research agent gets back is that connection, pre-computed in the Enigma business identity graph, sitting there waiting to be asked.
Less time on identity means more time on the actual work. A financial analyst asking Parallel for a deep dive on a private company gets the parent, the subsidiaries, the officers, and how all of that has changed, with citations down to the filing. A sales team’s prospecting run goes deeper on each account without the overhead. Any question you ask Enigma about business identity gets back an answer that took us years to establish, returned in milliseconds.
“Content that is uniquely valuable and hard to replace earns more.” That is how Parallel describes Index. It is also, without qualification, a description of Enigma’s data.