The Unsexy Way to Build a Customer Support Unicorn

The Pragmatist's Bet Against AI Maximalism

While venture capital pours into generative AI customer service startups promising to revolutionize every interaction, a two-decade-old company in Athens is making a different argument: sometimes you need a knife, not a bazooka.

Omilia, which has been automating voice-based customer support since 2002, just closed a $67 million Series B round led by Expedition Growth Capital. The funding arrives as the company reports annual recurring revenue of $60 million, representing a tenfold increase since its previous raise in 2020. That earlier round brought in $20 million from Grafton Capital, making this only the second time the company has sought external investment in more than two decades of operation.

The contrast with the current crop of AI-first competitors is stark. Where startups like Sierra and Decagon position themselves as generative AI platforms first and foremost, Omilia CEO Dimitris Vassos frames his company's approach as tool-agnostic. The philosophy centers on a simple observation: a substantial portion of customer inquiries involve routine information retrieval, like checking account balances or confirming order status, tasks that do not require the computational expense or latency of large language models.

"You may have a bazooka, but if your enemy is near you, you need a knife," Vassos said, describing the reality of contact center operations. The company positions itself as deploying whatever technology fits the specific interaction, rather than forcing every query through the same AI pipeline.

Unit Economics as Competitive Moat

At DailyTechWire, we have tracked the enterprise AI funding wave across Asia and North America, and Omilia's trajectory stands out for its capital efficiency. While competitors have raised hundreds of millions to scale go-to-market operations and subsidize early customer acquisition, Omilia has grown to $60 million in ARR on just $20 million in prior funding. The company now employs roughly 500 people and expects to reach 600 by year-end.

Vassos frames this as a deliberate strategy rooted in what he calls "great unit economics" for both Omilia and its customers. The implication is that customers see faster return on investment when automation targets the right use cases with the right technology, rather than applying expensive generative models universally. For Omilia, this translates to higher gross margins and more sustainable growth, even if it means less visibility on LinkedIn compared to the latest AI darlings.

"We don't mind that we're not as sexy as ElevenLabs and Sierra on LinkedIn right now," Vassos said. "We care about growing steadily and building the foundations for a billion-dollar revenue company in the next three years."

That timeline is ambitious but rooted in existing traction. Omilia's customer base includes Capital One, Discover, RBC, the UK's Department for Work and Pensions, and utility PSEG. The company has also made inroads into quick-service restaurants, a sector that has become a significant focus. Taco Bell has deployed Omilia's voice ordering technology across more than 1,000 locations, and Vassos indicated the company is in discussions with two additional restaurant chains in the United States.

The Quick-Service Restaurant Wedge

Voice ordering in drive-throughs represents a particularly demanding test case for conversational AI. The environment is noisy, accents and speech patterns vary widely, menu customization is complex, and latency must be minimal to avoid frustrating customers and slowing throughput. Traditional rule-based systems struggle with edge cases, while pure LLM approaches can introduce unpredictable responses or unacceptable delays.

Omilia's hybrid approach, which layers self-learning agents on top of deterministic logic for high-confidence tasks, appears well-suited to this environment. The company's deployment at Taco Bell has not been without controversy, however. A widely circulated report last year claimed a customer managed to order 18,000 cups of water through the voice system, raising questions about guardrails and validation logic. Vassos disputed the incident, stating that Omilia's internal logs show no such order was placed. We have reached out to Taco Bell for confirmation, and will update if the company responds.

Whether or not that particular incident occurred, it highlights a broader challenge facing voice AI in production environments: the need to balance flexibility with constraints. A system that can handle complex, multi-item orders with substitutions and special requests must also prevent absurd or impossible transactions. The tension between these requirements is where unit economics become critical. Over-engineering guardrails with expensive models can erode margins; under-engineering them can lead to operational chaos and customer dissatisfaction.

The U.S. Expansion Play

Omilia plans to deploy a significant portion of the new capital toward expanding its U.S. presence, a market that already accounts for a large share of revenue. The company is opening a new office stateside and hiring for three senior go-to-market roles: chief revenue officer, chief marketing officer, and vice president of revenue operations. These hires signal a shift from product-led growth and word-of-mouth to a more structured sales and marketing engine, a transition that many European enterprise software companies undertake as they scale in North America.

The timing is strategic. Contact centers across the United States are under pressure to reduce costs while handling increasing query volumes, particularly as labor markets remain tight and wage expectations rise. At the same time, enterprises are growing wary of AI projects that promise transformation but fail to deliver measurable ROI. Omilia's pitch, grounded in existing deployments at household-name financial institutions and restaurant chains, offers a lower-risk entry point than bleeding-edge generative models that remain difficult to audit and control.

The company's Athens headquarters also positions it well to tap into Greece's growing software engineering talent pool, which has become a quiet hub for European tech companies seeking cost-effective development capacity without sacrificing quality. While the ecosystem lacks the venture density of London or Berlin, it has produced a handful of successful B2B software exits over the past decade, and Omilia's longevity and profitability make it an anchor employer in the local market.

The ROI Reckoning Ahead

Vassos's prediction that "the next few years will see the companies that can offer real ROI prevail" reflects a broader shift we have observed across enterprise AI. Early adopters are moving past proof-of-concept pilots and demanding clear metrics on cost savings, accuracy, and customer satisfaction. Vendors that cannot demonstrate these outcomes are finding it harder to renew contracts, let alone expand within accounts.

This environment favors incumbents with established customer bases and proven deployment models. Omilia's 22-year operating history and focus on voice, a narrower surface area than omnichannel platforms, give it credibility that pure-play generative AI startups lack. At the same time, the company faces competition from larger players like Google Cloud's Contact Center AI and Amazon Connect, which bundle conversational capabilities into broader cloud offerings and can afford to subsidize customer acquisition through platform lock-in.

The challenge for Omilia will be maintaining its unit economics advantage as it scales go-to-market and enters more competitive deals. Hiring a C-suite focused on revenue growth suggests the company is preparing to invest more heavily in sales, which will pressure margins in the near term. The question is whether the underlying product efficiency and customer ROI are strong enough to support a more aggressive growth posture without sacrificing the profitability that has defined its trajectory so far.

Choosing Weapons Wisely

The broader lesson from Omilia's approach is that enterprise AI is entering a phase where technology selection matters as much as technology capability. The companies that win will be those that match tools to tasks, rather than applying a single hammer to every nail. For routine queries, deterministic logic and lightweight models deliver faster responses at lower cost. For complex, open-ended interactions, generative models add value despite their expense. The trick is knowing which is which, and building systems that route intelligently between them.

Omilia's metaphor of knives and bazookas captures this pragmatism. In a market still enamored with the promise of generative AI, the company is betting that customers will ultimately reward results over rhetoric. With $60 million in ARR and blue-chip customers already deployed at scale, that bet is starting to look prescient. Whether it can reach Vassos's billion-dollar revenue target in three years will depend on execution in the U.S. market and the ability to maintain cost discipline as competition intensifies. For now, the unsexy path appears to be working.