LangOps UVU Brouhaha
- Author
- Devin Gilbert
- Published
- Last Updated
In 2024, I published this blog about GILT. I had written it for my students who were taking my Translation Technology course (we talk about the language industry a lot in this course; it’s not just about CAT tools and MT), but then I published it on LinkedIn, and someone I had never met named Stefan Huyghe commented on my post telling me I should be looking into something called “LangOps.” I was immediately worried about this mystery concept I had never heard of before, and I spent some time looking into it, but day-to-day work got in the way, and I honestly never dove deep into the topic.
A year later, in 2025, while teaching the same course, I’d had some more time to research, and I assigned several readings about LangOps, hoping to delve into the topic with my students. My approach was to learn together with my students: I’d found readings I felt were worthwhile, and I was excited to disucss them further with my class. However, I truly wasn’t prepared for my students’ reactions, which turned out to be very anti.
My students' initial contrarian reactions were quite uniform:
- they were confused,
- they didn’t feel the LangOps prophets adequately articulated what LangOps is,
- they didn’t see how it was different from internationalization and localization.
This reaction genuinely surprised me, and I found myself playing devil’s advocate, arguing against my students and trying to convince them that LangOps was something new and distinct. In the end, I feel like I came to understand what LangOps is, but I’m also wondering what the main LangOps proponents (pioneers) would say to the eventual conceptualization/definition I arrived at (read till the end 🤓). Shoot me an email or a message/comment on LinkedIn to let me know what you think.
What They Studied
Here’s what I had my students read and study so you have the same context and background they were coming from. Remember that students had, prior to being introduced to LangOps, studied how internationalization and localization fit into business-minded globalization strategies (see my aforementioned blog post on GILT from 2024).
- The Evolution of Localization: Why LangOps Is the Next Frontier by João Graça, Co-Founder and CTO of Unbabel
- The LangOps Manifesto, including the case studies on the same website
- LangOps: The Vision and the Reality by Renato Beninatto, Chairman and Co-Founder of Nimdzi and CEO of LocWorld
- On the Origin of LangOps: The Evolution of the Localization Roadmap by Andrew Warner, assistant editor and staff writer at MultiLingual
- We are Living in The Era of the AI Idiot by Dion Wiggins, CTO of Omniscien Technologies
- This last reading was actually something my students had read well before we started examining and debating LangOps, when we were in the thick of studying MT and the new generative-AI future that was upon us with the post-2022, meteoric advent of modern LLMs. I’m including it in this list because it’s possible (or likely) it influenced the critical perspective they ended up taking.
Summary of Our Back and Forth
Remember that the core question is what is LangOps? Below I’ll include my student detractors’ (or “haters’ ”) stance along with my defense of the concept.
For the Haters
Unsatisfied by its proponents’ definitions and explanations (they cited lack of clarity and/or concision), to the haters, LangOps simply seems like
- a repackaging of internationalization and localization,
- consultant-speak used to enhance the peddling of their own monetizable idea sharing and thought leadership,
- localization with a heavy focus on AI,
- Many students criticized what they saw as lip-service to human control over AI-optimized processes while simultaneously minimizing or eliminating human participation as much as possible
The haters keep hearing that LangOps is completely new, that it will/should change everything, but they aren‘t capturing the vision of what it really is or how it’s actually different from i18n and l10n.
My Understanding of LangOps
Based on the above readings, I think there are three key elements to LangOps:
- Unified Organizational Strategic GILT
- True, it’s the same ol' strategic GILT that any organization would have always been better off having before the LangOps era. However, LangOps advocates a newfound emphasis on unified, top-down organization-wide strategy:
- Short cheeky summary: Give the language people power to do smart stuff.
- The next two main elements would be consequences of this first element
- Linguistic/Content Data Stewardship
- Having a well-thought-out, consistent strategy for how to effectively store, manage, and leverage your organization’s linguistic data and content data:
- corpora (your organization’s textual content)
- parallel corpora (TMs, bitexts, less-structured parallel texts)
- monolingual corpora
- termbases
- other types of content (images, videos, other assets)
- Automation, in all its many forms
- Efficient automation of content lifecycles
- How do you make everything that happens between
- authoring content or creating a new product and
- a customer consuming that content or product in a variety of different markets and locales
- be easier, more cost-efficient, and suck less?
- Automatic (and human-in-the-loop) translation of content (much of this has been a part of “old-fashioned” (traditional, i.e., pre-2022) l10n workflows for differing amounts of time now)
- Don’t just take whatever MT engines that big tech feeds you, but customize your own MT engines using all the data you are gathering nice and carefully.
- Feed your data as context to LLM-enabled agentic translation/localization workflows to achieve higher-quality automated translation/localization.
- Maintain well-managed terminology resources for human and machine use
- Use automatic quality estimation to route content through workflows with different levels of human involvement depending on content use cases.
- Additionally, other applications for linguistic/content data, including some that are *“new-fashioned” (i.e., more viable in the post-2022 era)
-
- sentiment analysis of customer reviews
- *content generation in other languages (using generative AI)
- *customer-facing or internal AI chat bots and agents
In one sentence:
LangOps is the strategic leveraging of linguistic/content data and technical expertise to improve automation-centric workflows in order to pursue a business’s globalization goals.
Feedback?
Once again, shoot me an email or a message/comment on LinkedIn to let me know what you think.
I’m also curious to hear what people think about the current status of the term “LangOps.” Has the term achieved the purchase many thought it would? Is it still a relevant term? I don’t seem to see it being thrown around as much as I used to, and this begs the question: Are the concepts that are core to LangOps still ever as relevant in today’s language industry, but has the term “LangOps” itself fallen into a bit of disuse?