This market has the potential to offer opportunities not only for Language Technology Platforms (LTPs) but also for Language Solutions Integrators (LSIs) whose “multilingual expertise, linguistic review workflows, and global networks of contributors provide a foundation for producing and evaluating datasets across languages and cultural contexts.”
Despite this growth opportunity, almost half (47.6%) of Slator readers polled say their companies aren’t participating at all in the data-for-AI space. Still, that means just over half are active in the market. A third of respondents (33.4%) said they were “somewhat” active, and one in five (19.0%) described their company as “very active” in the data-for-AI market.
Beating the Headwinds
Super Agency TransPerfect reported 7% year-on-year growth in 2025, as its diversification strategy, which led to 3% growth in 2024, continues to bear fruit. The company cited its proprietary tech and embracing AI tools as major revenue drivers.
A portion of the company’s year-on-year growth can also be attributed to a string of successful acquisitions, including their rival LSI Unbabel and the German dubbing studio SPEEECH Audiolingual Labs.
At the same time, this news comes with a potentially uncomfortable caveat for the language industry: 2025 marked the first year that language services represented less than 50% of the company’s revenue.
Phil Shawe, Co-CEO at TransPerfect, told Slator, “some may interpret this milestone as headwinds for the language industry, but we see it as a true paradigm shift — one that has seen us move upstream with our customers to deliver industry-specific solutions that blend language services, technology, and consulting.”
Whether they share Shawe’s perspective on the situation or not, nearly two-thirds of responding Slater readers (63.9%) also saw the proportion of language services in their business decrease in 2025. About one in five (22.2%) saw no change, and the remaining group (13.9%) actually saw an increase in the proportion of language services in their overall business.
To Buy or to Build? That is the question.
In early March 2026, the CEO of Phrase, Georg Ell, joined Slator’s Florian Faes for the third time on episode #279 of SlatorPod, discussing how the language technology platform (LTP) is responding to the AI boom.
In particular, the rise of vibe-coding tools like Claude Code and Lovable has led some buyers to question whether it’s worth buying services from an LTP when they can potentially build their own solution internally at lower cost.
Ell highlighted that Phrase has found success not by answering the “buy or build” question but rather by adopting a “buy and build” perspective as they embrace integrating other products into their platform. He notes that the company encourages customers to “build on top of” Phrase. At the same time, Ell cautioned buyers to consider the cost, labor, and functional challenges involved before trying to build their own localization tech stack entirely in-house.
Slator’s Florian Faes and Esther Bond also discussed the buy-or-build question with Daniel Sebesta, focusing on Walmart’s major AI translation initiative.
Walmart international CPO, Tim Simmons, stated in a podcast interview that the retailer had been spending USD 25m per year on translation before developing the Walmart Translation Platform (WTP), which now allows the company to “translate millions of catalog items per month across 22 languages” at only 1% of the cost.
Despite cost-saving success stories like Walmart, nearly half of Slator readers (46.9%) think buying localization services is still the way to go. About one in five (21.9%) said it’s better to build your own, and almost a third (31.2%) would recommend trying both to see what works best.
The More the Merrier
In a March 25, 2026, press release, RWS announced the launch of “Language Weaver Pro,” its new AI translation tool, which was developed in partnership with Cohere.
Ben Faes, CEO of RWS, discussed the collaboration on episode #282 of SlatorPod. Faes highlighted the impact of integrating Cohere’s AI tech into the Language Weaver platform, saying, “behind this LLM partnership there’s a fundamental shift in how we think about localization in many instances.”
It seems to be a positive outcome for both companies. RWS will use Cohere’s advanced AI models as the “brain” driving their translation product, and Cohere benefits from the long roster of end clients using their model through the Language Weaver platform.
With companies like RWS and Phrase integrating models developed by partners, end users may benefit from more convenient access to a variety of integrations they would otherwise have to source individually. This begs the question, how many different models are companies actually using in their translation workflows? We asked Slator readers exactly that, and received a range of answers.
The majority of respondents report their company using less than five AI models for translation, with over a quarter (28.0%) saying they use two to four models, and an equal cohort (28.0%) using only one. The group using more than five models is equally split between those using five to nine (16.0%) and those who use ten or more (16.0%). Surprisingly, about one in ten respondents (12%) didn’t know how many AI translation models their company uses.
