Amir Haramaty, Co-Founder and President of aiOla, joins SlatorPod to talk about how spoken, multilingual data can transform enterprise workflows and unlock real ROI.

The Co-Founder introduces himself not as a serial entrepreneur but as a serial problem solver, focused on one core challenge: most enterprise data remains uncaptured, unstructured, and unused.

Amir emphasizes that traditional speech tech fails in real-world conditions, where accents, noise, and hyper-specific jargon dominate. He illustrates how he tackles this challenge by building workflow-specific language models that extract only the data relevant to a process.

Amir says aiOla converts speech not into text but into structured, schema-ready data, allowing organizations to automate workflows, improve compliance, and identify trends long before humans can. He explains that the company focuses on narrow processes rather than general conversation, enabling precision in niche environments.

Amir shares how aiOla routinely cuts multi-hour procedures down to minutes, drives efficiency across frontline roles, and creates previously unavailable datasets that feed enterprise intelligence. He highlights ROI examples from supermarkets, airlines, manufacturing, and automotive industries.

Amir explains that after proving aiOla’s value, he realized the fastest way to scale was through firms already embedded in enterprise digital transformation. He notes that aiOla now partners with UST, Accenture, Salesforce, and Nvidia, creating a distribution engine capable of replicating wins across thousands of clients. 

He calls this channel strategy a force multiplier that shortens sales cycles and embeds aiOla inside broader modernization initiatives. Amir adds that these partners not only bring scale but also domain expertise aiOla deliberately chose not to build in-house. 

Amir outlines future priorities, including product-led growth, speech-based coding, and speech-prompted AI agents. He predicts that agentic systems will rely heavily on high-quality spoken data, making aiOla’s role even more central.

Tom Kocmi, Researcher at Cohere, and Alon Lavie, Distinguished Career Professor at Carnegie Mellon University, join Florian and Slator language AI Research Analyst, Maria Stasimioti, on SlatorPod to talk about the state-of-the-art in AI translation and what the latest WMT25 results reveal about progress and remaining challenges.

Tom outlines how the WMT conference has become a crucial annual benchmark for assessing AI translation quality and ensuring systems are tested on fresh, demanding datasets. He notes that systems now face literary text, social-media language, ASR-noisy speech transcripts, and data selected through a difficulty-sampling algorithm. He stresses that these harder inputs expose far more system weaknesses than in previous years.

He adds that human translators also struggle as they face fatigue, time pressure, and constraints such as not being allowed to post-edit. He emphasizes that human parity claims are unreliable and highlights the need for improved human evaluation design.

Alon underscores that harder test data also challenges evaluators. He explains that segment-level scoring is now more difficult, and even human evaluators miss different subsets of errors. He highlights that automated metrics built on earlier-era training data underperformed, particularly COMET, because they absorbed their own biases.

He reports that the strongest performers in the evaluation task were reasoning-capable large language models (LLMs), either lightly prompted or submitted with elaborate evaluation-specific prompting. He notes that while these LLM-as-judge setups outperformed traditional neural metrics overall, their segment-level performance varied.

Tom points out that the translation task also revealed notable progress from smaller academic models around 9B parameters, some ranking near trillion-parameter frontier models. He sees this as a sign that competitive research is still widely accessible.

The duo concludes that they must carefully choose evaluation methods, avoid assessing models with the same metric used during training, and adopt LLM-based judging for more reliable assessments.

Nikita Agarwal, Founder of Milestone Localization, joins SlatorPod to talk about her journey founding a language solutions integrator (LSI) and launching Cavya.ai, a platform designed to streamline translation project preparation.

Nikita began Milestone Localization in 2020 after discovering the language industry while working in international sales. She was drawn to the field’s global scope and low barrier to entry. She emphasizes that sales experience played a crucial role in landing early clients and understanding the value of hiring people from within the industry. 

The founder reflects on the past 16 months as a period of intense change marked by AI disruption, client pressure on pricing, and shifting expectations. She highlights how regulated sectors like life sciences have helped stabilize the company amid volatility. She details how the LSI specializes in medical device translations and regulatory submissions across Europe.

Nikita explains that her new platform, Cavya.ai, emerged from internal needs to improve project preparation. She says the tool automates glossaries, style guides, and document analysis, reducing time and boosting consistency for small and mid-sized projects.

The founder shares her observations on India’s evolving language technology landscape, noting significant progress in AI for major Indian languages. She says increased internet access and AI-driven localization are expanding education and job opportunities across the country.

Nikita concludes that she sees the future in expanding life sciences work, refining Cavya, and developing an AI-powered QA tool. She notes that some clients are showing “AI fatigue” and returning to human-led workflows.

Thordur Arnason, Global AI GTM Lead at Capgemini Invent, joins SlatorPod to talk about how the consulting giant is embracing language AI through BabelSpeak, its new real-time AI speech translation platform.

Thordur explains that the idea emerged from Capgemini’s AI Futures Lab while researching multimodal AI. Inspired by Meta’s launch of the Seamless M4T model, the team set out to tackle the hard problem of live AI speech translation.

He notes that early pilots with DNB Bank, the Norwegian Red Cross, and the Norwegian Police tested BabelSpeak in critical situations — from refugee banking access to emergency communication.

Thordur highlights Capgemini’s partnerships with Nvidia and Telenor, saying Nvidia provides the AI hardware and models, while Telenor’s sovereign AI infrastructure ensures security, GDPR compliance, and data sovereignty.

He emphasizes that BabelSpeak’s reliability comes not just from AI models but from engineering precision, reducing latency from three seconds to under 300 milliseconds.

Thordur discusses Capgemini’s exploration of agentic AI, where autonomous systems perceive, reason, and act independently. He describes how the company built an “Agentic Workbench” to help non-technical users experiment with AI agents safely and sees BabelSpeak as a potential tool within larger agentic systems.

He concludes that Capgemini is expanding BabelSpeak into a broader suite of language tools, combining secure AI infrastructure with advanced multilingual communication for enterprise and government clients.

Florian and Esther discuss the language industry news of the week, with congratulations to Villam Language Services on its sale to InAnyLanguage. Slator served as joint exclusive advisor with Maveria Advisory, representing Villam throughout the end-to-end M&A process.

The duo turns to Perplexity’s Localization Manager job posting, which they found almost identical to OpenAI’s earlier post, down to matching structure, order, and phrasing. They question whether copying such a specific ad shows a lack of seriousness or simply reflects practicality and efficiency.

Esther and Florian talk about RWS's new leadership hires: Stephen Lamb as Chief Financial Officer and Michael Wayne as Head of Media and Entertainment. Esther outlines how the appointments strengthen RWS’s investment strategy in media localization, dubbing, and content adaptation.

Esther next mentions that Visual Data has named Maz Al-Jumaili as SVP of Worldwide Localization, to lead subtitling and dubbing operations and strengthen client partnerships.

The duo wrap up with the UK government’s bizarre energy-efficiency study, claiming AI translation is a thousand times greener than human translation. They review the flawed logic, where the report assigns human translators the entire office energy costs while excluding AI infrastructure.

Tom Bridges, CEO and Founder of CaptionHub, joins SlatorPod to talk about how a small in-house tool evolved into a global AI-powered multimedia localization platform. 

Tom began his career in post-production and visual effects before stumbling into subtitling when a client needed to localize a video into 16 languages overnight. He reveals that the disorganized workflows relying on spreadsheets inspired him to create a more efficient, centralized solution, which became CaptionHub.

Tom explains that CaptionHub has since grown from a subtitling tool into a full multimedia localization platform integrating speech recognition, machine translation, and synthetic voice. He adds that the platform’s strength lies in being AI-agnostic and offering end-to-end workflows that balance automation with human-in-the-loop processes.

Tom describes how CaptionHub’s new product suite, Timbra, enables real-time media localization and has already supported major live events. He says live captioning is technically complex but benefits from the company’s years of research into video-on-demand subtitling quality.

Tom notes that accessibility regulations like the European Accessibility Act are driving demand, while AI and language models are opening new frontiers such as lip-sync and sign-language integration. 

Tom envisions a future where speech-to-speech translation, synthetic dubbing, and real-time localization merge into seamless, scalable experiences. CaptionHub’s mission remains to make multimedia communication universally accessible and efficient through human and AI collaboration.

Florian and Esther discuss the language industry news of the week, with breaking news that DeepL is reportedly exploring an initial public offering (IPO) in the US at a potential USD 5bn valuation. This comes as DeepL now positions itself as a “global AI product and research company”. Florian also notes the launch of DeepL Marketplace and the appointment of Gonçalo Gaiolas as Chief Product Officer.

Florian opens with the first-ever Slator Award at ZHAW Zurich University of Applied Sciences, where Guy Ratnitsky won for his thesis on data security and confidentiality in AI. The program will soon be renamed MA in Multilingual Communication Management to reflect market realities.

The duo turns to Anthropic’s new Economic Index, which shows translators and interpreters make up 0.63% of Claude AI usage, while OpenAI data previously showed translation-related conversations at 4.5%. 

Florian unpacks comments from German Chancellor Friedrich Merz, who, during a visit to Spain, suggested AI could replace EU interpreters in the medium term. He explains that Spain is pushing for Catalan, Basque, and Galician to become official EU languages, but Merz cited translation workload and complexity.

Florian and Esther then run through live AI speech translation updates: Zoom’s in-house rollout, Apple’s AirPods, Google’s translation features, Microsoft’s API, and Meta’s Ray-Bans.

In Esther’s M&A corner, she reports on Bering Lab’s acquisition of Intersphere in Korea and Iyuno’s partnership with Motion Picture Solutions in the UK for a film localization pipeline. Meanwhile, Testronic secured funding to scale down in some locations while expanding in Manila as a hub for QA testing and localization.

Slator’s Senior Research Analyst Alex Edwards joins Esther and Florian on the pod to discuss ElevenLabs’ move from a pure-play language technology platform (LTP) to becoming a language solutions integrator (LSI) by adding a managed service offering.

He outlines that the LSI will now offer managed services such as dubbing, transcription, and subtitling, hiring in-house linguists and vendor managers, while charging about USD 22 per minute for dubbing.

Florian then turns to YouTube’s rollout of multi-language audio tracks, which allows some creators to upload high-quality audio directly to videos and opens major opportunities for AI dubbing providers. 

The discussion shifts to OpenAI’s research on ChatGPT usage, reporting that translation accounted for 4.5% of more than a million sampled conversations, underscoring massive global demand for AI translation.

Esther highlights Microsoft’s launch of its Live Interpreter API, which promises real-time speech translation with “human interpreter level latency”. Esther also details Mistral’s USD 2bn funding to advance European AI capabilities, allowing them to compete with US and Chinese AI giants. Esther closes by reporting on WIPO’s new Korean-English post-editing tender.

Florian and Esther discuss the language industry news of the past few weeks, beginning with a recap of SlatorCon Silicon Valley 2025, where the duo noted strong localization buyer and user turnout, and tech-focused discussions across presentations and panels.

One key highlight was Cohere’s well-timed launch of Command A Translate, which allowed Kelly Marchisio to share details on building multilingual LLMs. Esther notes that Cohere’s multilingual models focus on high-quality coverage of about 20 languages rather than attempting hundreds.

Florian turns to the Apertus launch in Switzerland, where EPFL, ETH Zurich, and the Swiss Supercomputing Centre released a multilingual model trained on over 15 trillion tokens and covering more than 1,000 languages, including Swiss German and Romansh.

Esther reveals that Middlebury Institute will phase out its graduate translation and interpretation programs by 2027, marking the loss of a key training ground.

Esther reports on TransPerfect’s acquisition of Unbabel, with plans to integrate its AI tools, such as TowerLLM and EuroVLM, into GlobalLink, while CEO Vasco Pedro will stay briefly during the transition. 

Florian outlines Apple’s launch of AirPod Pro 3 with live AI translation and Google’s new Gemini-powered updates for AI live speech translation.

Esther concludes with the Inc. 5000 rankings, highlighting 11 language industry companies. She highlights Propio, Boostlingo, and CQ Fluency as repeat entrants, with Propio topping the list but also announcing job cuts following its acquisition of CyraCom.

Dipak Patel, CEO of GLOBO, joins SlatorPod to talk about his journey into language services and the challenges and opportunities of integrating AI into healthcare communication.

Dipak explains that his career began in consulting and private equity, but a personal experience with his mother’s healthcare highlighted the importance of interpretation services and led him to GLOBO.

The CEO emphasizes that since 2020, GLOBO has doubled down on healthcare, embraced AI and large language models, and addressed the mounting pressures of clinician shortages and aging populations. 

Dipak gives an overview of GLOBO’s platforms: HQ provides backend data and reporting, Connect enables access to interpreters through mobile devices, and KAI is the company’s AI interpreter, which is undergoing pilots across US hospitals.

Dipak cautions that AI cannot replace expert interpreters in all situations as interpreters serve as more than simple conduits; they clarify meaning, act as cultural brokers, and advocate for patients. He believes the near-term role of AI is filling gaps in the patient journey where interpretation currently does not happen.

Dipak details how GLOBO is using AI to monitor interpreter quality in real time, checking professionalism, background noise, and accuracy. He stresses that security, data protection, and careful testing are crucial to AI adoption in healthcare.

Dipak reflects on the growth of GLOBO, attributing it to a strong team and relentless focus on innovation. He concludes that while AI will play a bigger role in the next decades, the key lies in balancing it with human expertise.