SlatorPod

Gabriel Fairman, CEO of language technology platform (LTP) wxrks, joins SlatorPod to discuss advances in AI translation quality, the changing role of human expertise, and how LTPs must adapt to an increasingly AI-driven industry.
Fairman explains how wxrks has evolved beyond the conventional translation management system (TMS) model, incorporating AI capabilities directly into its workflows. He argues that as AI makes software development easier, enterprise software's value increasingly lies in its security, support, maintenance, and ability to handle complex business requirements.
Discussing wxrks’ latest translation quality research, Fairman points to significant improvements in context-sensitive AI translation, with the LTP's 2026 study showing lower average edit distances than traditional neural machine translation. However, he stresses that achieving high-quality results still requires careful configuration, continuous testing, and human oversight.
As fewer edits become necessary, Fairman sees growing value in capturing the reasoning behind human corrections and using that feedback to improve AI translation workflows. He introduces wxrks' translation reasoning feature, which provides greater transparency into AI-generated translation suggestions, and Free Flow, a segmentless editing environment designed to support context-aware rewriting, subtitling, and multimedia workflows.
The CEO remains skeptical that agentic workflows will replace traditional enterprise software or that MCP integrations will become the primary way users interact with it. However, he sees significant potential in using MCP to connect and analyze data across systems.
Turning to pricing, he predicts that traditional per-word pricing will become increasingly difficult to justify as AI reduces the amount of text requiring human intervention. Instead, he sees value shifting toward human judgment, quality validation, and frameworks that identify where expert intervention is needed.
Looking ahead, Fairman identifies organizational transformation as a major challenge. As AI reshapes established roles and workflows, he believes companies that can navigate structural change, build trust, and help employees redefine where their expertise adds value will gain a competitive advantage.

Gato Plugins Founder Leonardo Losoviz joins SlatorPod to discuss how advances in AI translation are making multilingual websites more accessible, more affordable, and easier to manage.
Leonardo explains how his background in software development and experience with multilingual Asian markets inspired him to create Gato AI Translations, a WordPress plugin that automatically translates website content upon publication, using AI models such as ChatGPT, Claude, and Gemini. 
He highlights the growing opportunity for website owners to reach international audiences as AI significantly reduces the cost and complexity of translation. While he has yet to observe a major shift toward multilingual websites in Southeast Asia, he expects adoption to increase.
He also discusses Agent Bridge, another extension that allows users to translate content using existing AI subscription quotas rather than paying separately for API usage, offering cost savings in exchange for longer processing times.
On website visibility, Leonardo discusses declining Google traffic and questions the conversion potential of generative search, emphasizing that traditional SEO remains essential.
Leonardo also emphasizes the importance of contextual prompting, translation memory, and human review for high-stakes content. He explains how customizable prompts can function as glossaries, helping prevent recurring translation errors.
Looking ahead, Losoviz is developing one-click translation for entire websites and exploring additional integrations. His broader ambition is to make AI-powered translation work across the WordPress ecosystem, regardless of how a website is built.

Artem Grigorev, CTO of AI data company Toloka, joins SlatorPod to discuss the growing importance of human expertise in AI development, the shift toward smaller specialized models, and the challenges of evaluating AI performance.
Grigorev explains that while computing power and model capabilities attract significant attention, high-quality training data remains fundamental to AI progress. As models become more sophisticated, data creation has shifted from simple annotation tasks to complex, expert-driven workflows. He highlights growing demand for domain specialists, including PhD-level experts, as large language models increasingly automate simpler annotation tasks.
Grigorev also discusses Toloka's work on post-training smaller models for specific enterprise applications. By fine-tuning models on targeted datasets, businesses can potentially achieve better performance, lower costs, and faster response times than they would relying on large frontier models.
Turning to multilingual AI, he emphasizes the need to evaluate model performance across languages, rather than relying predominantly on English-language benchmarks.
Toloka Arena, launched earlier this year, is the company's private benchmarking initiative designed to address concerns around contaminated public benchmarks and assess AI agents against realistic business workflows. Grigorev explains how Arena tests agents across multiple attempts to assess consistency rather than a single successful run.
Grigorev also discusses the growing demand for robotics training data, including first-person videos of humans performing physical tasks, and the challenges of collecting such data at scale.
Looking ahead, he argues that human judgment will remain essential, even as AI becomes increasingly capable, particularly in determining whether AI-generated outputs meet human expectations and preferences.

Adam Bittlingmayer, CEO and co-founder of ModelFront, joins SlatorPod to discuss the state of the art in enterprise AI translation and the language technology platform’s (LTP) launch of an AI-native open translation management system (TMS).
Bittlingmayer says translation can appear solved when an AI model is given enough context and the right prompt, but providing that context consistently within complex enterprise localization workflows remains difficult. Enterprise teams must account for different products, content types, terminology, quality requirements, languages, and existing systems.
The CEO revisits ModelFront’s core value proposition of using AI to check and fix AI, allowing enterprises to automatically approve translations that would otherwise require human review. Focused on high-stakes localization workflows, the LTP takes responsibility for translation quality and the engineering complexity behind it. He also highlights its outcome-based pricing model, under which customers pay only for translations that the LTP successfully approves.
Bittlingmayer also talks about why neural machine translation models remain widely used despite advances in LLMs. He says newer models can benefit from much richer context, but existing enterprise workflows and software often make providing that context difficult. He sees verification as increasingly important as AI-generated content and translation volumes continue to expand.
Bittlingmayer then introduces ModelFront’s AI-native open TMS, OpenLoc, which the company is making available for free, with hosting offered at cost. He argues that traditional TMSs were not designed for today’s AI workflows, creating friction that can encourage teams to bypass centralized localization processes altogether.
He says the new platform is designed around openness, flexible AI use, prompt building, and MCP connectivity, allowing users to interact with localization workflows through enterprise AI assistants as well as the traditional interface. Bittlingmayer sees this as part of a broader goal to make AI more accessible while improving enterprise AI outcomes.
ModelFront is launching OpenLoc through an early access program, with general availability planned to follow.

On SlatorPod, AudioStack Co-Founder and CTO Peadar Coyle discusses what it takes to deploy voice AI at enterprise scale, from localization and domain expertise to human oversight.
Peadar explains how AudioStack enables brands, governments, and media companies to produce localized audio across accents, regions, languages, and individual products. Rather than betting on a single voice model, the company works with multiple providers, treating models as building blocks and selecting them according to language, accent, quality, and use case.
The CTO shares how AudioStack moved away from developing its own text-to-speech models after deciding its competitive advantage lay in orchestrating technologies, delivering high-quality audio, incorporating brand guidelines, and applying domain-specific knowledge.
Peadar also argues that humans will remain part of AI audio workflows as automated systems can increasingly identify potential quality problems, but brands still want editorial control over pronunciation, tone, and reputation-sensitive content.
On synthetic voices, he highlights consent, vendor due diligence, transparency, and regulation, with AudioStack introducing metadata identifying AI-generated content.
Looking ahead, AudioStack is investing in agentic capabilities, faster and more intelligent workflows, and a combination of proprietary technology and third-party models.