Cuckoo Puts a Real-Time AI Translator in Your Zoom Call, and Snowflake Is Already Using It

The Macro: Global Sales Teams Are Still Playing Telephone

The language services market is projected to reach $98 billion by 2028. That number includes everything from human translation and localization to interpretation and AI-powered tools. What’s interesting isn’t the size of the market but the gap between how companies sell globally and how they communicate globally.

Multinational companies have had translation tools for decades. But most of those tools are designed for documents, not conversations. You can translate a marketing deck or a support article and take your time getting it right. Real-time communication is a different problem entirely. When a sales rep in Chicago is on a Zoom call with a prospect in Seoul, the translation needs to happen instantly, with the right technical vocabulary, in a tone that doesn’t sound robotic.

The existing solutions break down into a few tiers. At the low end, you have Zoom’s built-in translation and Google Translate, which work for casual conversations but fall apart on technical or industry-specific terminology. In the middle, you have professional interpretation services, which are expensive and hard to scale. A live human interpreter for a sales call costs $150 to $300 per hour, and scheduling them adds friction to a process that’s already complicated. At the high end, companies like Unbabel and Smartling handle translation at scale but focus more on written content and customer support than live meetings.

The missing piece has been a real-time translation tool that’s good enough for high-stakes business conversations. Not good enough for casual chat. Good enough that a sales engineer can explain a technical architecture across a language barrier without losing precision or confidence.

The Micro: The Founders Built South Korea’s Fastest Ride-Hailing App

Yong Hee Lee is the co-founder and CEO, handling go-to-market. He was the first GTM hire at two Korean startups in online programming education and AI infrastructure, and led international speaker event marketing. He studied Electrical Engineering and Science and Technology Policy at KAIST. Gunwoo Kim is the co-founder and CTO. Before Cuckoo, he built South Korea’s fastest-growing ride-hailing app, reaching one million users in nine months, which was faster than Uber Korea’s growth in the same market. He studied Computer Science at KAIST.

Cuckoo is a real-time AI translator built for global sales and marketing teams. Companies like Snowflake and PagerDuty are already using it in Zoom calls and in-person meetings. The product interprets across 20-plus languages and learns the technical details specific to each customer’s business. That last part is important. A generic translator might render “data lakehouse” into something meaningless in Korean. Cuckoo learns those terms and gets them right.

They came through YC’s Winter 2025 batch as a two-person team. The product works in both virtual meetings and in-person settings, which expands the use case beyond video calls to conferences, trade shows, and on-site customer visits.

What separates Cuckoo from the generic translation layer is the learning component. The system isn’t just translating words. It’s building a vocabulary of company-specific terminology, product names, technical concepts, and industry jargon that improves over time. For a sales team doing repeated calls in the same vertical, that compounding accuracy is the difference between a tool they tolerate and a tool they rely on.

The Verdict

I think the positioning is smart. Going after sales and marketing teams rather than general business communication gives Cuckoo a clear value proposition and a buyer who can measure ROI. If your sales team can close deals in markets they couldn’t serve before because of language barriers, the product pays for itself immediately.

The Snowflake and PagerDuty logos are strong early signals. These are serious enterprise companies that wouldn’t adopt a translation tool unless it actually worked in high-stakes conversations. The question is whether those are deep integrations or surface-level pilots.

The competitive moat here is the learning. Generic real-time translation is becoming a commodity as the underlying models improve. What’s not a commodity is a system that knows your company’s specific terminology, your product names, your pricing language, and the way your sales team talks about features. That kind of institutional knowledge takes time to build and creates real switching costs.

Thirty days, I’d want to see how many calls per week active customers are running through Cuckoo. Sixty days, the question is whether the terminology learning is actually reducing errors over time or plateauing. Ninety days, I’d want to know the expansion motion: are these companies rolling Cuckoo out to more teams and more language pairs, or is usage staying flat after initial adoption? The product makes sense. The early customers make sense. Now it’s about proving that real-time AI translation can be a standalone product category rather than a feature that Zoom builds in-house.