Variance, a San Francisco-based company building AI agents for risk and compliance, has closed a $21.5 million Series A round. Ten Eleven Ventures led the investment, joined by 645 Ventures, Y Combinator, Urban Innovation Fund, and Okta Ventures. The company has now raised $26 million in total.

The new capital will go toward expanding the infrastructure behind Variance’s agentic AI platform and deepening its work with financial institutions.

Fraud Is Getting Faster. Investigations Haven’t.

Fraudsters are adopting generative AI to fabricate synthetic identities at a scale that traditional detection systems were never built to handle. At the same time, the evidence needed for a proper investigation sits scattered across disconnected systems, often slow to access and harder to interpret.

Variance is designed to close both gaps at once. Its agents reason over fragmented, complex data and return fully auditable decisions in minutes rather than days.

“AI is the biggest leap we’ve ever had for both facilitating and stopping crime. We intend to use it to its fullest, to make sure it’s on the side of the people trying to stop it.”

Karine Mellata, CEO and co-founder of Variance

From Signal to Decision, Without a Queue

The funds will support investment in the infrastructure underpinning Variance’s investigative agents and expand capacity to serve more financial institutions. Mellata has been direct about the goal: “This fresh round of funding will allow us to get our technology into the hands of every compliance team that’s still doing this work the hard way.”

Beyond scaling the platform, the company is focused on broadening the workflows its agents can handle and extending reach into new enterprise environments where compliance backlogs remain a persistent problem.g reach into new enterprise environments where compliance backlogs remain a persistent problem.

How the Agents Actually Work

When a new customer triggers a KYB review, a Variance agent pulls entity registration documents, cross-references beneficial ownership records, traces ultimate beneficial ownership structures, checks sanctions lists and adverse media, and returns a decision, all within minutes, in accordance with the client’s existing compliance playbook.

A second agent monitors data continuously, identifying suspicious patterns and entity relationships that fall outside predefined rules, and surfaces risks before they are formally recognized as such.

The difference between a classifier and an investigator matters here. A classifier returns a probability score. A Variance agent pulls business registrations, traces ownership to sanctioned entities, surfaces adverse media, and delivers a complete evidence pack with a recommended action, following the customer’s own compliance procedures.

Three components make this possible. A context engine maps each customer’s full data universe, including entities, events, and relationships, into a single ontology the agent reasons over, enabling multi-hop investigations at machine speed. An SOP enforcement layer encodes compliance procedures in plain English, so agents execute them consistently with a full audit trail every time. A data access layer connects over 150 global business registries, sanctions lists, court dockets, adverse media sources, and identity verification platforms, including sources across the surface, deep, and dark web.

Results at Enterprise Scale

At one of the world’s largest retailers, Variance agents traced ultimate beneficial ownership chains, uncovered coordinated fraud rings, and cross-referenced sanctions and adverse media, completing in minutes what previously required analyst teams and weeks of manual work. At a leading Fortune 500 company, the platform achieved human-level precision on Know-Your-Customer investigations across hundreds of cases, fully replacing manual review queues.

The platform processes more than 70 million context signals per day and executes approximately 300,000 automated enforcement actions across client environments. For institutions with strict data sovereignty requirements, Variance supports fully on-premise and air-gapped deployments with customer-managed encryption keys, field-level data controls, and confidential computing. Customer data never leaves the institution’s own infrastructure and remains inaccessible even to Variance itself.

Founded by Engineers Who Built at Apple’s Scale

Variance was founded in 2023 by Karine Mellata and Michael Lin. The two met while working together on Apple’s fraud engineering and algorithmic risk team, where they built systems that had to operate reliably at enormous scale. That background shapes how Variance approaches the problem: not as a compliance product layered on top of existing tools, but as a purpose-built investigative infrastructure.

The company was previously known as Intrinsic. It is headquartered in San Francisco and serves large enterprises running mission-critical compliance workflows across KYC, KYB, AML, transaction monitoring, and fraud detection.

The Investors Behind the Round

Ten Eleven Ventures led the Series A. The firm has focused exclusively on cybersecurity since its founding and has raised over $1 billion across more than 60 investments worldwide, with a portfolio that includes KnowBe4, Darktrace, Cylance, and Ping Identity.

“The Variance team not only deeply understands the pressure compliance teams are under, but also the massive opportunity these teams have at this moment to add a new level of precision and efficiency to their workflows with new agentic processes. This is a new approach using deep contextual reasoning, automation, and auditability, giving enterprises a powerful way to achieve their risk reduction, brand trust, and marketplace integrity goals.”

Megan Dubofsky, Partner at Ten Eleven Ventures

Origianl Creator: Ekaterina Pisareva
Original Link: https://justainews.com/companies/funding-news/variance-raises-21-5m-series-a-to-automate-financial-crime-investigations/
Originally Posted: Tue, 31 Mar 2026 13:49:14 +0000