The company that spent 24 years trying to kill the phone menu

Call your bank tonight and you may never speak to a human. A voice answers, waits, and - here is the odd part - it actually understands the messy sentence you blurt out about a charge you do not recognize. It knows who you are before you finish talking, because it recognized your voice. There is a decent chance that voice was built by a company you have never heard of, in a place you would not guess: Larnaca, on the southern coast of Cyprus.

The company is Omilia - Greek for "speech" - and it has been at this since 2002. That date matters. Omilia was building what the industry now calls conversational AI a full two decades before ChatGPT made the phrase dinner-table conversation. For most of that time it was profitable, private, and almost invisible. On August 6, 2026, it stopped being quite so invisible, announcing a $67 million Series B led by Expedition Growth Capital.

01 / What it actually doesAnswering the calls nobody wants to make

Omilia sells software that runs customer-service conversations without a person on the other end. Not a scripted phone tree - "press 1 for billing" - but an agent that listens, figures out intent, verifies identity, and resolves the request. It works over the phone, in chat windows, and in messaging apps, and it is aimed squarely at the kind of enterprise that fields millions of calls a month and would very much like to spend less doing it.

The product is the Omilia Cloud Platform, a stack of Voice Agents, Chat Agents, and Authentication Agents sold to banks, insurers, telecoms, utilities, and a growing list of retailers. The scale is the headline: more than 200 enterprise deployments, over a billion conversations a year, and, according to the company, a Tier 1 US bank running north of a million calls a day through its systems.

02 / Who is on the other endCapital One, Vodafone, and a lot of Taco Bell

Omilia's customer list reads like a directory of companies with enormous call volumes and low tolerance for mistakes. In banking and financial services: Capital One, Discover, RBC, Piraeus Bank, Sense Bank, Alfa Bank. In insurance and utilities: Allstate, Aon, PSEG, Enel, the UK's Department for Work and Pensions. In telecom and industry: Vodafone, Nissan, Johnson Controls. And in fast food, Taco Bell, whose drive-thru voice AI Omilia has deployed across more than a thousand locations - which is to say the company's software is as likely to take your order for a Crunchwrap as to move money between your accounts.

These are not experiments. One insurance client, the company says, processes over 600,000 agent-assisted interactions a day. That is the tell about who Omilia is for: not startups kicking the tires, but incumbents who need automation that does not fall over at rush hour.

03 / The problemThe most hated invention in modern life

The interactive voice response menu - the IVR - is a small daily cruelty. It makes you say your account number three times, routes you to the wrong department, and then hangs up. Companies hate it too: every call that fails to self-serve gets handed to a human agent, and human agents are expensive. The problem Omilia set out to solve is the gap between what customers actually say and what old systems could understand.

The payoff, when it works, is measurable. Omilia cites deployments where self-service usage rose 40%, call capacity climbed 43%, internal transfers dropped 96%, and handling costs fell around 20%. Those are the numbers that get a project renewed, and they are the reason a contact-center director in Toronto will take a call from a vendor in Cyprus.

04 / The differenceA bazooka and a knife

Here is where Omilia gets contrarian. The fashionable move in 2026 is to point one very large language model at a problem and let it improvise. Omilia's founder thinks that is a mistake for customer service, and he has a line for it.

Translation: a giant generative model is powerful but blunt. Answering a bank's phone reliably takes a set of specialized tools working together - accurate speech recognition for noisy calls, voice biometrics to prove identity, a dialogue engine that tracks context, and fraud detection running underneath. Omilia built all of these in-house over two decades: deepASR for speech recognition, deepVB for voice biometrics, and DiaManT for dialogue management. Generative AI gets layered on where it genuinely helps, not bolted on as the whole system.

That difference has a name in analyst-speak: Gartner has placed Omilia in the Visionary quadrant of its Magic Quadrant for Conversational AI Platforms two years running, and singled out its focus on security and governance as its clearest point of separation. In 2024 it was the only vendor named a Customers' Choice in Gartner's Voice of the Customer report, with 97% of reviewers saying they would recommend it.

05 / ProductsYour voice becomes the password

The part of Omilia that best explains its enterprise appeal is authentication. Its Authentication Agents verify who you are from your voiceprint - no PIN, no password, no "what was the name of your first pet." That removes a source of friction customers hate and a source of fraud banks fear. And because deepfaked voices are now a real attack vector, Omilia built TalkGuard, a product that listens for synthetic voices and spoofing attempts in real time, during the call.

06 / The modelHow it makes money

Omilia is a business-to-business software company. It licenses the Omilia Cloud Platform on recurring subscriptions, typically tied to call and conversation volume, sold both directly and through partners - notably a strategic alliance with the global outsourcer Atento, and a listing of its voice biometrics on NICE's CXexchange marketplace. Annual recurring revenue reached roughly $60 million in 2026, and, tellingly, the company grew for 18 years without venture money before taking its first outside check in 2020.

07 / The expertiseA first job that never ended

Omilia is largely one person's long bet. Founder and CEO Dimitris Vassos began his career at IBM, where he helped deploy voice technology across more than 70 countries. He took that knowledge home to Cyprus and started Omilia in 2002 - first as a contact-center systems integrator stitching together other people's platforms, then, when those tools could not deliver natural conversation, as a builder of its own. He runs the company with founding partners John Nikolaidis, its chief commercial officer, and Pelias Ioannidis, its CFO. The through-line is that Omilia knows this one narrow, unglamorous domain - the enterprise phone call - better than almost anyone, because it has been living in it since before the smartphone.

08 / The marketWhere it sits, and what the $67M is for

The voice-AI market for contact centers is suddenly crowded. Established platforms like PolyAI, Cognigy, Kore.ai, Google's Dialogflow, and Microsoft-owned Nuance compete for the same enterprise budgets, and a wave of generative-first startups - Sierra, Decagon - is raising large rounds on vision. Omilia's pitch against them is unfashionable: it raised on revenue, not on a deck, and it points to unit economics and profitability where others point to potential.

The new money is aimed at the company's obvious weak spot - it is a Cyprus firm selling largely to America. Omilia says the $67 million will open a new US office, build out its go-to-market team, and add a chief revenue officer, a chief marketing officer, and a VP of revenue operations. The bet is that 24 years of quiet product-building can now be matched by a sales engine loud enough to convert it.

Whether that works is the open question. But there is something clarifying about a company that has answered a billion calls and still describes part of its own work as art. "There is a portion that is art behind what we do," Vassos says. Coming from a firm this focused on numbers, it is an unexpectedly human thing to say about a machine that talks.