The silent shift that already happened: From downstream production to upstream solution

This article was created in partnership with Kevin O’Donnell and aims to resurface and validate the ideas discussed at the Loc Tech Live conference back in February 2026. Read Kevin’s original article here.

Last February, Kevin O’Donnell, a growth consultant and strategist, organized a successful, hands-on, accessible, and online conference on technology and localization: Loc Tech Live. The premise was simple: what happens when you remove the vendor pitches and let practitioners show what they’re actually building? At the AI Loc Think Tank, we feel strongly aligned with this vision and want to amplify the insights gathered during Loc Tech Live, which was a great success with 324 attendees from over 117 companies. The turnaround revealed that our community is eager to engage in high-quality, targeted events (which was then again validated by the AI ThoughtCon, with over 1,000 registrants).

From the conversations during Loc Tech Live, we can spot three main trends:

• Multilingual content infrastructure is evolving

• Teams’ responsibilities are shifting

• Quality is being redefined

These topics and conversations are not completely new, but they are still important because they consolidate ideas that felt vague before, and now we can see how they are actually implemented in practice. Moving away from hypotheses and jumping right into the experimentation and validation. Let’s dive into each of these ideas.

Multilingual content infrastructure is evolving

If I had a dollar for every time someone said that TMS is dead, I might be rich by now. Jokes aside, it is a recurrent topic that requires our attention. In the past few years, we’ve witnessed how most TMSs have been lagging behind when it comes to AI integrations. And, to be fair to TMS developers, that approach makes a lot of sense from a product development perspective. You don’t want to spend a lot of time and money developing a new feature that will be completely useless after a few months. So while TMS infrastructure has been and still is at the core of localization operations, it has been made clear that something else is needed.

And that something else has been taking shape in startups like Blackbird, BeLazy, or n8n. Each of them brings something completely unique and gives solutions to specific challenges, be it the need for orchestration and content management, deep integrations, or the creation of agentic workflows. This idea was validated during Loc Tech Live, and as Kevin put it: “Presenters demonstrated systems that connect content management platforms, developer pipelines, AI services and quality layers into unified workflows. The translation management system was present, but as one component in a larger architecture”.

TMS is definitely not dead; it’s actually alive and kicking, but it’s also not enough to run multilingual content operations at scale anymore.

Teams’ responsibilities are shifting

The next topic, which very much resonates with the Think Tank members as most of us are undergoing this transformation in our own flesh, is the shift in focus of what localization professionals and programs are supposed to be doing. Traditionally, we focused on making sure that the content was properly TEPed on time by super pro linguists, maintaining the integrity of formatting, and adhering to whatever instructions, legal requirements, brand style guides, and terminology were sent our way. Challenges back then were scalability and consistency, and guess who helped with those? Yes, TMS. And a buddy called NMT. Thank you for your service.

But passed those initial challenges, and with the possibilities now brought by the new tech wave, teams are evolving into something much more complex and strategic. Forced by pressure and necessity, we are now looking at content and language tiering strategies. Do we need to localize everything? And from what we need to localize, do we need to review every single string? Globalization teams that are well-positioned are now being heard and consulted to build strategic multilingual content programs that are optimized and aligned with business goals.

And not only are teams being more strategic, but they are also building their own solutions. As Kevin put it: “Domain experts who understand the localization problem deeply are now prototyping and shipping their own solutions, filling gaps between off-the-shelf products without waiting for vendor roadmaps or engineering resources”. To me, this is very exciting because knowledge about workflows, how languages work, and how users behave was taken for granted and hidden before. No one cared much. But now, people who have been doing this work for long enough are the best positioned to dictate how to design the localization strategy of the future.

Quality is being redefined

Another pet peeve of mine is when we describe as “good enough” one of the possible tiers of language quality. If quality in language is so subjective (sorry, MQM, but even with you, we couldn’t reach a solid agreement), then how can we agree that “good enough” is something that we can actually use as a guiding principle and articulate in a workflow? But rants aside, it is true that the concept of quality has dramatically changed in our industry in the past few years, and we need to pay attention. Treating all content equally doesn’t make sense in the first place, and there is so much budget waste there that could be simply optimized with a strategic approach, sometimes even without the need for AI. And the other part of the equation is: have we been measuring the right metrics? MQM, number of bugs, edit distance, and so on. The right question should be geared towards the consumers: what is important to them? And how can we make a positive impact on the user experience? What’s the role that language plays in all of it? This is a topic still to be explored more in our industry.

Going back to being strategic, we’ve seen how most mature programs are creating different pipelines for different content types and making the most of technology AND human talent. I really liked this bit from Kevin’s article: “For executives, the implication is a direct trade-off between speed, cost and coverage. Companies that adopt this model can move faster into new markets, release more content with fewer resources and scale their international presence without linearly scaling their teams. But it requires something many organisations don’t yet have: a governance framework for AI-driven quality decisions”. Especially, the part about having a governance framework for AI-driven. In my experience, this is the main blocker when it comes to further automation. The foundation should be: 1) amazing language assets in the first place; 2) alignment among teams on what’s important (content and language prioritization and what quality actually means for the different stakeholders); 3) crystal clear workflows that are highly standardized, understood and followed by everyone; and 4) clear governance frameworks that link all the previous points together.

Conclusion

So the main conclusion is that, yeah, things are changing, and there is no coming back to the old ways of doing things. I’ve already gone through my mourning period of what we’ve lost in the way and have chosen to look forward and try to figure out how we can continue creating value and a positive impact on the industry and those in it. The silver lining is that all of us who have been doing this for quite some time have the knowledge to help shape the future of localization operations, and our knowledge will be even more valuable than before.

Disclaimer: This article was fully human-written. Any similarity to AI writing is purely coincidental. The infographic was created with GPT, though. I won’t blame you if you hate it.