Omilia secures $67 million in Series B funding to accelerate global expansion of its Agentic Self-Learning CX platform for large enterprises

The $67 million Series B that Omilia closed this week is not just another capital event in the conversational AI category. It is a signal that the enterprise contact center market is finally separating technology that works at scale from technology that demos well.

Expedition Growth Capital led the round, a transatlantic firm that specializes in software and AI infrastructure plays. Their conviction rests on a simple thesis: the winners in enterprise voice AI will be the companies that own the full stack — models, runtime, telephony integration, compliance posture — rather than those stitching together third-party LLMs, TTS engines, and SIP trunks and calling it a platform.

Omilia's numbers make the case. Annual recurring revenue has crossed $60 million, a tenfold increase since its Series A, achieved without interim equity raises. The customer roster reads like a regulated-industry roll call: Capital One, Discover, RBC, PSEG, the U.S. Department for Work and Pensions. Taco Bell, the outlier in consumer brand terms, now runs Omilia voice agents across more than 1,000 drive-thrus in 38 states. The Forrester Wave named Omilia a Leader in Conversational AI Platforms for Customer Service in Q2 2026.

The architecture beneath the metrics

What distinguishes Omilia is not the logo slide but the architecture that makes those logos possible. The company's voice-first platform is built on proprietary agentic Voice AI — models it trains, owns, and serves from its own infrastructure. That ownership chain is the lever for three enterprise non-negotiables: security certification depth, latency predictability, and cost determinism.

Omilia holds FedRAMP, PCI-DSS, SOC 2, HIPAA, and GDPR certifications simultaneously. Those are not checkbox exercises; they are the admission tickets for Tier 1 banks, insurers, and healthcare systems. A platform that routes audio through a frontier LLM provider's API, a separate TTS vendor, and a CPaaS layer inherits the compliance posture of its weakest link. Omilia's single-stack approach means the audit boundary stops at its own perimeter.

Latency tells the same story. The company cites 50,000 concurrent voice interactions for a single client at sub-second response times. That is not a benchmark; it is a production reality at a top-tier U.S. bank handling well over one million calls per day. A second deployment, a real-time agent-assist program for a multinational U.S. insurer, processes 600,000 calls daily. These volumes are not achievable when every utterance triggers a chain of external API hops, each with its own queue, retry logic, and SLA.

Cost predictability as a product feature

The pricing model is where architecture becomes commercial strategy. Omilia charges zero token pass-through. Enterprise buyers can calculate ROI from day one because the cost curve is a function of Omilia's own compute economics, not a third-party LLM vendor's token meter. In a market where CFOs have been burned by variable inference bills that scale unpredictably with conversation length and complexity, that predictability is a competitive moat.

The Agentic Self-Learning CX platform, launched after the Series A, compresses time-to-production for new AI agents. The "self-learning" label refers to the platform's ability to ingest unstructured call data — transcripts, recordings, outcomes — and automatically refine agent behavior without manual intent mapping or flow diagramming. For enterprises sitting on years of recorded interactions, that capability turns historical exhaust into training signal.

Market context: the great consolidation

The funding arrives at an inflection point for the CCaaS and CX automation categories. The first wave of LLM enthusiasm produced a thicket of startups wrapping GPT-class models in prompt layers and calling them voice agents. Most have discovered that enterprise voice is a different discipline: barge-in handling, silence detection, DTMF interplay, compliance logging, and the brutal acoustics of drive-thru lanes and call center floors. The orchestrators are now confronting the hard problems that Omilia solved by building downward into the media plane.

Incumbents — Genesys, NICE, Five9, Talkdesk — are acquiring or building their own model layers. The pure-play LLM orchestration vendors are scrambling to add telephony, compliance, and deterministic pricing. Omilia's Series B, sized for global expansion rather than survival, positions it as a consolidator rather than a consolidatee.

What the capital buys

Expedition's capital funds three vectors: geographic expansion into EMEA and APAC regulatory regimes, deepening the self-learning loop with the newly launched Lexis generative TTS model, and extending the agent-assist modality that now handles 600,000 daily calls for the insurer deployment. Lexis is notable because it keeps TTS inside the proprietary boundary — another piece of the stack brought inward.

The investor syndicate also matters. Expedition's transatlantic mandate aligns with Omilia's dual Athens-New York footprint and its customer base that spans North American banking, European public sector, and global QSR. The firm's software-and-AI specialization suggests they underwrote the technical diligence on model ownership, not just the commercial traction.

The view from CRM Today

Omilia's trajectory illustrates a hardening truth in enterprise AI: the application layer is not where value accrues when the workload is voice, the buyer is regulated, and the scale is millions of sessions per day. The value accrues to the infrastructure layer — the team that trained the acoustic model, wrote the media server, certified the compliance envelope, and priced the compute on its own terms.

The $67 million is not a milestone. It is the war chest for the next phase of a category shakeout that has been overdue since 2023. Enterprises have stopped buying demos. They are buying architectures that survive audit, scale without surprise bills, and learn from their own exhaust. Omilia built that architecture before the market asked for it. The Series B is the market catching up.