This marks a nine-point increase from 43% in 2024 and signals a move toward flexible, multi-engine setups that allow LSIs to switch or mix models for different clients, domains, or languages, reducing reliance on any single engine.
The share of LSIs working directly with AI translation providers — such as DeepL, Google, or Microsoft — remained stable at 46%, suggesting that LSIs are not abandoning these partnerships but are increasingly routing access through the TMS layer, where integration, automation, and performance management can be centralized.
In-house development of AI translation engines continues to decline — falling from 16% in 2024 to 11% in 2025. The report attributes this drop to “high costs, technical complexity, and faster innovation from external providers,” which makes in-house development harder to justify — especially for small and mid-sized LSIs.
Rather than maintaining direct connections to AI translation providers or investing in costly in-house development, LSIs are increasingly relying on LTPs and their TMSs as the central hub for AI translation and AI-enabled workflows.
The ALC report concludes that “TMS vendors are becoming more strategically important” within the AI translation ecosystem — a shift that could redefine how LSIs, LTPs, and AI translation providers interact in the years ahead.
