LocDiscussion

Localization Has Always Been A Context Problem...Who Owns Context In Your Process? Ian Cowley, Technical Writer at Salto feels context is even more relevant today than it was a few years ago. While we can now translate a string in seconds, we still need to first understand what that string means, where it appears, who will read it, and what the user is trying to accomplish. "AI needs that context as well. So it's even more important in this day and age." The role of technical writers is evolving. They're no longer just documenting products. They are increasingly responsible for building and preserving the context that everyone else depends on, from developers and designers to translators and AI systems. Where does context live in your organization today? According to Ian, technical writers are becoming "more of a context engineer" and I think that's an accurate description of where many teams are heading. That also explains why he feels so strongly about having a single source of truth for strings. "There always should be one unique source of truth... Having multiple versions of strings is unmanageable." When context lives in five different places, neither people nor AI know which version to trust. How many sources of truth does your team have today? It’s amazing how little attention we sometimes give to the plumbing behind localization. Ian walked through the GitHub integration his team uses with Crowdin. A developer adds new strings, they flow automatically into Crowdin, translators are notified, translations return through a pull request, and the team can review everything before merging. What piece of localization "plumbing" has had the biggest impact in your organization? It sounds simple, but it fundamentally changes the workflow. The repetitive work disappears, feedback stays attached to the strings, and everyone is working from the same source. That kind of automation has probably done more to improve day-to-day localization than many of the AI announcements we've seen over the past two years. Successful AI adoption in localization depends less on choosing the right model and more on creating the right context around it. If the context is poor, AI simply gets the wrong answer faster. If the context is rich, structured, and connected to the source of truth, both humans and AI produce better work. Isn't that one of the most important lessons for localization teams right now?

In Localization, Most of What We Still Call Marketing Doesn't Work Anymore... That may sound harsh, but just look at the amount of spam in your inbox: We publish. We promote. We announce. We send newsletters. A lot of it is nothing more than ads... Then we wonder why nothing moves. The problem isn't the amount of content we're creating. It's that we're competing for something much harder to earn than clicks. We're competing for attention. If nobody stops, nobody reads. If nobody reads, nobody engages. If nobody engages, your message goes nowhere. The uncomfortable truth is that good content is no longer enough. Publishing isn't the same as being seen. Brand messaging alone doesn't create trust. Attention comes first... That's the idea I'll be exploring in my Full Contact Marketing session at the Direct Client Summit. We'll challenge some long-held assumptions about B2B marketing in localization and look at why visibility, trust, and conversations have become the real drivers of growth. If you're responsible for business development, marketing, customer success, or thought leadership, I hope you'll join me. If nobody engages with your content, nothing else you do matters. And that doesn't just apply to social media. It applies to every form of marketing we invest in. I'd love to hear your take before the session. What marketing tactics you think our industry needs to leave behind? It’s a true honor for me to present at the Direct Client Summit this week together with an incredible selection of talented speakers. Looking forward to learning from you: Adrian Probst, Ana Sofia Correia, Dr. Bernard Song 宋鹏博士, Corinne McKay, Diana Gordaliza, Diego Cresceri, Gabriela Kouahla, Gosia Wheeler, CPACC, ADS, Joachim Lépine, M. Ed., C. Tr., Josh Goldsmith, Judy Jenner, MCI, Julia Poger Business, Conference and Diplomatic Interpreter, Laurie Ross, Macarena Troscé ✔️ Mindful Translatoré, Magali Karee, M.A. 🇩🇪, Marta Pagans, Martina Russo, Michael Schubert, CT, Milijana Trobradovic, Nora Díaz, Rane Souza, Renato Beninatto, Sabrina Sbaccanti, Sara Silva, Sherif Abuzid, Dr. Verónica Pérez Guarnieri, William Cassemiro  

The third and final day of the GenAI in Localization Conference started by giving the microphone to the audience... Mark Jones reminded us that curiosity may be the most valuable skill in localization today and Gary Hess gave us a behind-the-scenes look at the Hackathon, showing how AI coding assistants have compressed months of development into days and are changing what domain experts can build. 💡Dave Ruane and Jourik Ciesielski continued to guide us through a second round of Innovation Challenge demos: Ligia Sobral Fragano showcased why subtitle segmentation isn't just preprocessing. It's infrastructure that influences everything that follows. Kincaid Day challenged us start building localization stacks that manage quality, cost, governance, and risk together. Maksym Hamii showed how LLMs become production-ready when they're wrapped in workflows that preserve terminology, context, security, and existing CAT tool integrations. Diego Cresceri turned translation memories and customer documents into AI-powered terminology assets, proving that organizations often already own the knowledge they're looking for. Maria Martinez-Illescas Goldaracena explored Purpose over Perfection: Quality in the Age of AI. Yana Kolesnikova and Ksenia Lykina (Naberezhnova) demonstrated that translation was never the biggest bottleneck. AI-powered context enrichment can dramatically reduce the time localization teams spend chasing information before translation even begins. Maja N. showed how AI can completely rethink translation memory maintenance, transforming projects that once took weeks into workflows measured in hours. The Custom.MT MAD Hackaton then delivered a set of amazing prototypes tackling website QA, context-aware localization, collaborative translation, market readiness, AI model evaluation, model selection, and multilingual website auditing. Thank you for your excellent efforts: Gary Hess, Gabriela Janiszewska, Willian Magalhães, Vuk lazovic, Elena Murgolo, Matteo Cehovin and Christèle Forget! Congratulations Willian Magalhães on a well deserved win. A huge thank you to the entire GenAI in Localization team, Konstantin Dranch, Jan Hinrichs 🌍, Maria Braddick, Jillian Maxwell, Andrew Hickson, Daria Stepanova, and everyone working behind the scenes for another outstanding conference. What were your takeaways: Amanda Downing, Alla Stöckli, Aneta Sapeta, Beshar Bahjat, Lieven Lannoo, Lena Razumnik, Núria Calafell Cardenal, Stephen Healy, Yevgeniya Sergeyeva, Noëmie Baron, Igor Juríček, Katerina Gasova, Nektaria Notaridou, Sandra Najar, Zhanina Zhelyazkova I've pulled together the biggest ideas from every presentation in my latest AI in Loc newsletter. ⬇️

The second day of the GenAI in Localization Conference actively moved into building with AI... Maria Braddick shared her journey from translator to Interpretation Manager for the Milano Cortina 2026 Olympic Games, reminding us that language expertise opens doors far beyond traditional localization careers. 💡Dave Ruane and Jourik Ciesielski once again kept the Innovation Challenge moving at an incredible pace, asking the questions every buyer wanted answered and giving each innovator the opportunity to defend their ideas under pressure. Philippe Cao showed why multilingual content strategy is shifting from SEO to AI search, and why Answer Engine Optimization may soon become every global marketer's concern. Ana Dumančić ACAR challenged the idea that every LSP should simply adapt to existing software. Sometimes the fastest path forward is building technology around your own workflows. Arianne Farah demonstrated that not every automation project requires a software engineer. Giving linguists the tools to automate repetitive work may prove just as valuable as building bigger AI models. Adam Bittlingmayer argued that quality estimation should stop producing scores and start making decisions, allowing human reviewers to focus only where they add value. Artur Nowakowski reminded us that AI should not replace twenty years of linguistic knowledge. Translation memories, terminology and reviewer feedback remain strategic assets. Viveta Gene, PhD showed how knowledge graphs can transform AI translation from fluent to compliant, especially for highly regulated industries where correctness must be provable. Priscila Mottola demonstrated how millions of multilingual user ratings become actionable business intelligence when AI helps teams find the patterns hidden inside the data. Pavel Trysmakou made the case that localization should be measured by business outcomes. His Kraken platform showed what an AI-native localization ecosystem can look like. I teamed up with Sebastian Dzięcielski and Gergana Toleva to demonstrate how n8n allows localization teams to connect existing systems and automate complex workflows without building everything from scratch. We hope we inspired you to build the next N8N solution: Aneta Sapeta, Charline Lacoste, Beshar Bahjat, Andrew Henderson, Anastasia Taymanova, Amanda Downing, Alla Stöckli, Maja N., Dmitry Kurbakov, Denis Zhilko, Diego Cresceri, Diana (Dee) Savage, Erin Harchuck, Laszlo K Varga, Mark Jones, Maren Dietzel, Lieven Lannoo, Lena Razumnik, Núria Calafell Cardenal, Stephen Healy, Yevgeniya Sergeyeva, Noëmie Baron, A big thank you to the whole GenAI in Localization conference team: Konstantin Dranch, Jan Hinrichs 🌍, Jillian Maxwell, Andrew Hickson and Daria Stepanova who made everything run smoothly as always! I've pulled together the biggest ideas from every presentation here ⬇️🔥

Yesterday's opening day of the GenAI in Localization Conference covered an extraordinary amount of ground. We are no longer asking whether AI belongs in localization. We're figuring out how to build with it, govern it, and measure it. If you missed the sessions or simply want a single place to revisit the biggest ideas, I've pulled the highlights into one article covering the entire first day. Konstantin Dranch opened the conference with a clear message: AI is turning every professional into a potential software builder. The future belongs to localization teams that combine linguistic expertise with engineering skills and start building their own solutions. In his keynote, Georg Ell, CEO of Phrase, challenged the industry to stop defining itself by translation. He argued that language professionals should become orchestrators of language intelligence, helping businesses optimize outcomes instead of simply reducing cost per word. Amy Grace O'Brien showed how Adobe is using AI to transform terminology management. By extracting multilingual terminology directly from translation memories, her team is moving terminology work upstream while giving linguists more time for higher-value decisions. Lauren Barker demonstrated what happens when localization is treated like software engineering. By moving the entire translation lifecycle into GitLab CI/CD, translations now ship alongside source content in about an hour instead of several days. Michał Junczyk reminded us that building AI is only half the challenge. As autonomous systems become more common, organizations will increasingly depend on language experts to evaluate, calibrate, and continuously improve them. "The more autonomous AI becomes, the more it needs people who define quality." Thanks also to the extended Allegro team for their excellent workshop: Maria Dembska-Czerniawska, Agata Hajduk-Smak, Joanna Marhula The common thread across every presentation was that localization is becoming more technical, more connected to engineering, and more deeply embedded in product development than ever before. I can't wait to see what the 33 participants in the Hackaton will cook up in the vibecoding kitchen... Willian Magalhães, Gary Hess, Monika Murgová, Gergana Toleva A big thank you to my co-moderators and producers, Jan Hinrichs, Maria Braddick, Jillian Maxwell, Andrew Hickson, Daria Stepanova and everyone behind the scenes who made the conference possible. What were your main takeaways Igor Juríček, Christian Lang, Dmitry Kurbakov, Charline Lacoste, David Lakritz, Alexei Kizya, Alla Stöckli, Denis Zhilko, Ksenia Lykina (Naberezhnova), Laszlo K Varga, Lisa Sidorova, Lieven Lannoo, Mark Jones, Nicolas Jadot, Nektaria Notaridou, Serge Gladkoff, Stephen Healy, Sandra Najar, Oussama Abou-Jamous, Vajk Limbacher, Zhanina Zhelyazkova, Yevgeniya Sergeyeva? If you couldn't attend, or if you want to revisit the sessions in more detail, you'll find the full Day 1 recap here Now... on to Day 2.

How Do You Know Your Translations Work In the Real World? The most important work for Localization teams often begins when customers start interacting with your content. For Aydın Gür, Senior Localization Lead at Hogarth Worldwide, one of the most valuable sources of quality feedback comes directly from the people using the content every day. In his case, that means having conversations with employees in the Apple Stores he supports. For example, store employees told him that customers weren't using the official Turkish word for email. "They just say mail. They use it in English. So why don't we speak their language?" A style guide, a glossary, or a terminology committee don't drive that kind of decision. It comes from observing how customers actually communicate. On the agile localization podcast by Crowdin, Aydin shared how this kind of practical perspective has shaped the way his team thinks about language. How many of your localization decisions are made based on what people should say rather than what they do say?  And if customers consistently ignore the terminology you've chosen for them, is it still the right terminology? Aydın's view of quality follows the same logic: "Even if some small errors happen here and there, it gets fixed before anyone important sees it." And when deadlines are tight? "If the end user sees it, there is no flexibility. It should be perfect." Many localization discussions focus on what happens before delivery. Glossaries, QA checks, workflows, reviews. Customers never see any of that. They see the result. Who validates your translations? How close are you to the people using your content? How often do you hear directly from your end users? Do you ever talk to the people using your translations?

Are You The Builder In Your Localization Organization? Could You Become One? A year ago, most conversations about AI in localization revolved around understanding the technology. Today, I keep meeting people who are building things. A localization manager builds an internal workflow. A project manager creates an automation that saves hours every week. A linguist connects multiple systems together to solve a problem nobody else was going to solve. Konstantin Dranch, at the helm of the GenAI in Localization conference, confirms: "Every localization team is now an AI team." I suspect many people will debate that statement. That's probably healthy. Yet it is hard to ignore how much time localization teams now spend evaluating models, connecting systems, designing workflows, experimenting with automation, and figuring out how AI fits into their operation. "Model-assisted coding opens up the possibility for a localization specialist to create their own apps." For decades, localization professionals have been experts at identifying inefficiencies, bottlenecks, and opportunities for improvement. They knew exactly what needed to be built. Access to development resources was often the limiting factor. That equation is changing. The distance between identifying a problem and building a solution is getting shorter. Konstantin describes AI-assisted coding as "the single most important skill to acquire this year." It also helps explain why this year's GenAI in Localization Conference is so heavily focused on builders. The agenda includes workshops on AI-assisted coding, workflow automation with n8n, LLM evaluation, an innovation challenge, and a hackathon for those willing to put their ideas to the test: "Gladiators wanted." I have a feeling there are more localization gladiators out there than many people realize. Are you one of them?

The New Superpower Is Knowing What To Connect. Here Is Your Chance to Learn From The Best: 🔥🔥🔥⬇️ Innovation has a funny way of spreading. Did you know that most breakthroughs don't travel in a straight line? One person might solve a problem while another sees the solution and applies the underlying idea somewhere completely different. Later this week, 💡Dave Ruane and Jourik Ciesielski will be leading another Innovation Challenge at the online GenAI in Localization conference. (Thank you POEditor for sponsoring!) Innovation in the language industry has become much of their professional focus. So what are some of the common threads they notice among the solutions featured? According to Dave, someone can look at an innovation and think, "I don't necessarily need that innovation myself, but it's triggered something in my own mind that I can use." That's why I'm increasingly convinced that the most valuable conversations around AI have very little to do with models. "The real success lies in everything that happens before the use case and everything that happens after the use case is executed. We typically call that context and guardrails." Jourik confirms. The industry spent too long debating engines. Today, some of the most interesting work is happening around orchestration, context, workflows, knowledge, governance, and decision-making. "For years it has been about dots and commas, and finally I hear people speak about impact and intent and customer acquisition and sales." That may be one of the most important changes happening in localization right now. Quality is still important. It always will be. The conversation is expanding. More teams are asking whether content achieves its purpose. Whether customers understand it. Whether it influences behavior. Whether it drives results. Dave connects that change to technology in a way I hadn't considered before. He describes the emerging superpower of technologists as understanding which technologies should be connected together to create a business outcome. The winners in the next few years may not be the organizations with the biggest platforms. They may be the ones that know how to assemble the right combination of technologies, context, expertise, and workflows to solve a problem that matters. If learning more about the most original localization technologies and how to build them is of interest to you, don't forget to sign up for the GenAI in Localization conference which will take place on Wednesday, Thursday and Friday June 24th, 25th and 26th! "The best innovation challenge isn't the one that finds innovators. It's the one that creates them."