Multilingual AI is entering a new phase. What began as a race to build larger models and better datasets is evolving into a broader transformation of how multilingual content is created, evaluated, adapted, and delivered at enterprise scale.

AI training data represents a meaningful opportunity for language service companies, but it is not a natural extension of existing services. Demand from model builders continues to grow, particularly for multilingual datasets where performance and safety degrade outside English-language contexts.

At the same time, the competitive landscape is broad and evolving. It includes AI-native companies, gig economy platforms, business process outsourcers, and large technology firms. Language service providers bring a meaningful advantage in multilingual capabilities, but that advantage represents only part of what is required to compete effectively.

“AI data training is a growth opportunity for language services companies. But it is a very different business. I mean, when you really kind of get into the actual operating model and the technology requirements, it is quite different. The workflows are different. The talent pools are different.” – Paul Carr, CEO Welo Global

The operating model differs in fundamental ways. AI training data requires new workflows, new quality frameworks, and new technology infrastructure, along with talent that understands how models are trained and evaluated. The buyer relationship also shifts, moving toward direct engagement with AI labs and model builders rather than traditional localization stakeholders.

The opportunity is real, but it requires building distinct capabilities rather than extending existing ones. Treating AI training data as an adjacent service is unlikely to succeed.

This shift is extending beyond model training and into multilingual execution itself. As enterprises move from experimenting with AI to operationalizing it across global content workflows, the limitations of standalone machine translation systems are becoming more apparent. The next generation of platforms is being designed not simply to translate content, but to orchestrate how multilingual work happens across people, models, and enterprise systems.

Why are agentic translation systems like Opal outperforming machine translation? 

* Paul Carr CEO of Welo Global shares why agentic translation system like Opal outperforming machine translation.

“It is a much better version of a machine translation engine. In fact, it’s twice as good in all of the testing we’ve done. It delivers content that is twice as good from a quality standpoint, however you choose to measure it.”
Paul Carr, CEO Welo Global

Machine translation has improved significantly over the past decade, but a new category of systems is now delivering stronger results. Agentic systems like Opal combine machine translation, large language models, and workflow orchestration into a coordinated framework designed for enterprise use.

Opal reflects a shift in architecture rather than incremental improvement. It adapts to enterprise-specific data such as glossaries, style guides, and translation memories, enabling output that aligns with brand and domain requirements. It also operates with broader context, improving fluency and coherence across longer content.

What differentiates Opal is its ability to integrate multiple components, including retrieval mechanisms, structured data layers, and feedback loops that continuously refine output. This creates a system that learns over time rather than remaining static, delivering measurable gains in both quality and efficiency.

The result is not simply faster translation, but a different model for how multilingual work is executed, where AI and human expertise operate together within a single system.

What is open architecture in multilingual AI, and why does it matter? 

“Clients get really upset with being locked in.”
Paul Carr, CEO Welo Global

A defining characteristic of systems like Opal is open architecture. Rather than forcing companies into a single ecosystem, these systems are designed to integrate with existing tools and workflows, allowing organizations to maintain flexibility in how they operate.

This reflects a broader shift in enterprise technology. Companies want interoperability and control over their stack. They want to adopt new capabilities without replacing everything they have already built.

In multilingual AI, this approach enables organizations to incorporate agentic systems while preserving existing investments. It also supports a more flexible ecosystem where different providers can work together rather than competing within closed platforms.

How is AI changing the role of language professionals?  

The impact of AI extends beyond efficiency. It is changing how work is structured within the language industry. As automation takes on more execution tasks, the role of language professionals shifts toward oversight, decision-making, and domain expertise.

This reinforces the importance of specialization. Generalist work becomes less central, while domain-specific expertise becomes more valuable. That is particularly true in industries such as life sciences and intellectual property, where multilingual content must align with regulatory requirements, technical terminology, and structured review processes.

The organizations that succeed will combine technical systems with human expertise, ensuring that AI delivers outcomes aligned with real-world requirements.