Swiss-based language services provider Supertext has presented its top tips and recommendations for companies to reposition their language services to AI-first models at SlatorCon Remote on March 18, 2025.
The company offered its advice to SlatorCon attendees following the merger between language services provider Supertext and language AI company TextShuttle in April last year.
Samuel Läubli, CEO of Supertext told the SlatorCon audience, “There’s disruption all over the place, not only in the service business but also in technology and AI translation technology in particular, and I’m very excited about this. This is a huge [opportunity] for us players in this industry because if you combine [services and technology] properly, there are multiple opportunities [out there] for us.”
Läubli described how the company repositioned its website following last year’s merger to showcase a self-service AI translation system “front and center” where users can input text and benefit from a free translation service with additional paid features.
“It looks like your average online translator [platform] and there’s strong competition out there, that’s for sure,” he added.
“The bottom line here is clearly whether you’re an LSP or you’re on the buy side of a localization department, it’s certainly a good time to reassess your AI translation providers,” — Samuel Läubli, CEO, Supertext
Referencing the paradigm shift from statistical machine translation to neural machine translation several years ago, Läubli stated, “What we’re seeing in 2025 is another paradigm shift, and with it comes competition.”
The CEO cautioned the audience on comparing machine translation quality across tools and languages as part of this new paradigm shift, and how objective quality scoring is important, taking the example of comparing Supertext’s machine translation output with DeepL’s.
“[Now,] smaller players with less financial power than the big tech companies have actually shown to outperform these bigger players in terms of quality. […] The bottom line here is clearly whether you’re an LSP or you’re on the buy side of a localization department, it’s certainly a good time to reassess your AI translation providers,” he stated.
Demo of Supertext’s AI Translation Tool
Läubli invited the audience members to think about how users interact with AI-translated output. “You sit in front of your screen, you look at your AI [translation] output, and you have different options: you can accept the risk [of using the output as-is], you can use another AI tool to rate your AI output, […] or you can have it checked by a human.”
“What we asked ourselves is, can’t we do better? Can’t we integrate access to human professionals who would actually look at an [AI] translation? Couldn’t we make this available more directly to where users are?”
“I haven’t seen a system or AI agent that would actually guarantee that a translation is correct in the past ten years. And I’m quite sure I won’t see this in the ten years to come,” — Samuel Läubli, CEO, Supertext
The Supertext CEO proceeded to show the SlatorCon audience a demo of the company’s on-demand AI translation tool that enables users to request a professional review of AI output directly in the browser interface at the click of a button.
Läubli continued: “With [our tool], we’re actually closing this loop now. Just think of this being accessible by an API as well. If you have 80,000 product pages, for example, you could translate them with AI all in one go.”
“Then you can programmatically choose the top hundred pages in terms of visitors, revenue, questions in some Q&A section, and have these translations verified. […] The next time you [update] those 80,000 product pages, you would also benefit from this human work going forward, and this is a very interesting loop that gets better over time,” he added.
The language services provider reportedly provides human verification of AI-generated output within ten minutes during Western European office hours.
Läubli underlined the importance of having this human verification step. “I’ve been working in this field for ten years now, but I haven’t seen a system or AI agent that would actually guarantee that a translation is correct in the past ten years. And I’m quite sure I won’t see this in the ten years to come,” he concluded.