Omilia Customer Support Platform Raises $67M

Omilia has raised a $67 million Series B led by Expedition Growth Capital to develop self-learning agents for cross-channel customer support. The practical takeaway for AI builders is less about the size of the round than the company’s deployment thesis: use the least expensive tool that reliably solves each support task, then measure the result in operating terms.

Omilia is scaling a multi-channel support platform

Omilia has worked on automating voice calls and customer support since 2002. Its current product direction combines that history with agents designed to operate across contact points such as voice and chat, rather than treating every interaction as a generic large language model problem. The company says its platform serves customers including Capital One, Discover, RBC, DWP, PSEG, and Taco Bell. It also says its technology is deployed across more than 1,000 quick-service restaurant outlets. These client and deployment details come from the company’s account as reported in the funding coverage.

That positioning matters because contact centers contain many repetitive requests. Account information and other structured workflows may be better handled by deterministic systems, integrations, or smaller models than by an expensive general-purpose model. The useful architecture is often a routing layer that chooses between tools, retrieval, business rules, and generative models.

The ROI argument is the real product signal

Omilia’s CEO, Dimitris Vassos, presents unit economics as a core differentiator. The company argues that customer support automation should be judged by measurable ROI, not by how much generative AI it contains. That is a vendor position, not independent validation, but it reflects a real design constraint for production systems: inference cost, latency, failure handling, escalation, and human review all affect the economics of an automated interaction.

For builders, this suggests evaluating an agent by workflow rather than by model label. A system that resolves a narrow request cheaply and hands ambiguous cases to a human can be more valuable than a more capable agent that consumes more tokens and requires heavier supervision.

Funding will push Omilia further into the U.S.

The new capital will support a U.S. office, go-to-market expansion, and hires for a chief revenue officer, chief marketing officer, and vice president of revenue operations. Omilia has about 500 employees and expects to reach 600 by year-end, according to the report. The company is also pursuing two additional U.S. quick-service restaurant contracts. Those plans are expansion targets, not evidence that the contracts have been signed.

Omilia reports $60 million in annual recurring revenue, up tenfold since its $20 million financing round in 2020. Because the figure is company-reported and the supplied material includes no independent financial verification, builders should treat it as a signal of claimed commercial traction rather than a complete performance assessment.

Production reliability still determines the outcome

The Taco Bell example shows why deployment evidence matters. A reported ordering failure allegedly allowed a customer to request 18,000 cups of water. Vassos disputed that the incident occurred and said Omilia’s logs did not show it. The account therefore remains unresolved in the supplied coverage.

That disagreement points to a practical requirement for customer support automation: logs must cover the full transaction path, including prompts, tool calls, state changes, approvals, retries, and the final action taken by an external system. Without that audit trail, teams cannot reliably distinguish a model failure from an integration bug, a reporting error, or an unverified anecdote.

The decision rule is straightforward. Omilia’s approach is relevant to teams building contact center agents when they can define narrow workflows, instrument outcomes, and prove