English into many other languages is a common enough request in the world of localization. But for SaaS provider Language I/O, that translation usually needs to take place in real time, with minimal human involvement.
Language I/O uses machine translation (MT) to enable monolingual (often English-speaking) customer support agents to communicate instantaneously with customers. And minimal human involvement is not just for the translation itself — the API is designed to require as little effort as possible from the customer support agents, too, who can use it within their familiar tools.
Now, “riding the wave of the recent AI boom,” the Wyoming-headquartered company has raised USD 8m in a Series A1 round, bringing the total funds raised to date to USD 22.4m. This includes an angel investment that served as an early pre-seed round; USD 5m in a March 2021 Series A; and USD 6.5m in January 2022.
While Language I/O declined to share details of the deal, Founder and CEO Heather Morgan Shoemaker told Slator that the company’s NRR (Net Revenue Retention) “consistently hovers above 120%, and in H1 we exceeded the 100-customer threshold, many of them Fortune 500 Companies.”
The round’s biggest investor, however, was the Wyoming Business Council’s Wyoming Venture Capital Fund. This was the first investment for WYVC, described in a press release as “an equity financing option for Wyoming high-growth companies with an eye toward a future exit.”
Language I/O plans to use the funds from this round to ready its platform for the integration of generative AI technology. A portion will also go toward hiring, particularly for the existing machine learning team, which will help build the platform.
“This will allow us to continue to improve translation quality and ultimately generate language-specific, company-specific LLM-powered chatbots,” Shoemaker said.
Privacy as a Priority
As Shoemaker explained during her March 2021 appearance on SlatorPod, the process begins with a glossary, if a client has one, or creating a glossary for a client.
Once loaded (or generated), Language I/O constantly scans and updates the glossary proactively, drawing on the client’s public content, such as knowledge bases and marketing materials.
Upon receiving user-generated content, the server scans it for personal data, encrypting and tagging anything that might compromise the “zero-trace” platform. Part of pre-processing the UGC is identifying the language and matching it with one of several MT engines for the best possible general translation for that language.
Language I/O uses proprietary technology to substitute the client’s preferred translations for certain terms; this can happen before or after MT. The process also includes a normalization step for misspellings and slang, among other idiosyncratic content that pops up.
On the topic of customers’ evolving standards and expectations, Shoemaker said that in the Language I/O’s early days, “companies were hesitant to trust their UGC translations to a machine translation engine. Now it’s a very accepted practice.”
She added, “We’ve also seen security become a much larger focus over the past several years than it was at the beginning. Early on, companies who were willing to use machine translation technology didn’t pay a lot of attention to whether the vendor was holding onto their content.”
In addition to upgrading its technology, Language I/O continues to expand its staff of 52 FTEs, both in volume and in geographical spread, with plans to add to the machine learning, sales, and support teams.
The vast majority of employees are in the US, but the company has also added team members in Europe and in China.
Similarly, 70% of Language I/O’s clients are based in the United States; 22% are located in Europe; and 8% are spread across Oceania and Asia.