With Pavel Riazanov, Senior Localization Manager at HiBob

Below is an automated transcript of this episode

Stephanie Harris-Yee (Host): 0:04

Hi, my name is Stephanie Harris-Yee, and I will be your host today for this episode of Global Ambitions. And I am joined by Pavel Riazanov, who is the senior localization manager from the product team at HiBob. And we are going to be talking about what he has worked with their existing localization program this year and their workflows to adapt and change them to take advantage of new technologies and systems. So thank you for coming on the show, Pavel.

Pavel Riazanov (Guest): 0:45

Thanks so much for having me. Yeah, I’m really excited to talk about one of my favorite topics is the implementation of change in localization programs, especially this year where AI has been evolving so fast and it has reshaped nearly every part of how we operate.

Stephanie Harris-Yee: 1:01

I can imagine. It sounds like it’s been quite the year for you. So as you’re going into this change management and working through all these different systems, what has been one of the biggest challenges that you’ve kind of come up against this year?

Pavel Riazanov: 1:15

This year it’s it’s really been about navigating change on multiple levels. AI has been a big catalyst, of course, but the bigger challenge was how I combine AI adoption with the shift toward like a more agile incremental approach to localization. So, you know, historically, like I don’t know how how other companies operate, but a lot of localization work, especially when you launch a new language to a product, it usually happens in a monolithic way, like when you take a large monolithic batch, you translate it huge volume, you review a huge volume, and only then you achieve the final result. And it can be slow, it contains a lot of risks, and it’s really hard to scale.

So this year I finally arrived at the stage where I felt that I am able to break this process and to change it, you know, to break it into smaller controlled batches. And instead of doing one huge, enormous translation drop, we decided it would split the content into iterative chunks, and that allowed the translation and review, you know, go in parallel, save us a lot of time, give us a lot of visibility, and make quality far more predictable. And when we when we combined that with the artificial intelligence and the workflows based on on AI, so the result was a system that was much faster, much more scalable, which is most important, it did not sacrifice quality. So it helped us you know ship more efficiently, and we we’ve had far fewer surprises on the way.

Stephanie Harris-Yee: 2:53

Oh, that’s really exciting. It sounds like you’re in a good place then going into next year. Just for our folks here listening to try to give them some practical examples, maybe. Can you go into anything specifically that you did? Like this is one exact use case that we did that showed these kinds of benefits, anything like that?

Pavel Riazanov: 3:11

Oh, yeah, yeah, sure, sure. I have a great example. For a few years already, we’ve had Latin American Spanish in our product, but as we expanded into European markets, so we got to Spain, the Spanish customers were not really satisfied with that. They said that it doesn’t really fit their reality. So at some point we just had to take this decision and to to start adding the European Spanish, Castilian Spanish. And instead of launching a brand new locale from scratch, so we just uh decided that we take our existing Es Latan foundation and we make it much more targeted and relevant by taking the glossary, updating it, changing the terminology, updating the style guide, and adjusting the translation memory.

So once we refresh the core linguistic assets, so we used the updated TM to pre-populate the majority of the content. If I remember the numbers correctly, it was about like 70%. It was pre-filled from the older translation memory that had been adapted in advance. And that meant that only the segments that didn’t match, or you know, some brand new content, the delta that was still being created, so that it needed additional translation. So that’s where we used AI, but always with human review, you know, to ensure consistency and quality. And if we compare it to traditional new language rollout, it usually involves translating everything from zero. This approach was dramatically faster and it was significantly more cost effective. So it was a perfect example of how we combined strong linguistic groundwork with targeted AI usage, and no doubt it accelerates delivery and doesn’t compromise quality, as I mentioned earlier.

Stephanie Harris-Yee: 4:43

Yeah. As a follow-up question, how did you update the TM? Was that all on the internal manual side of things, or did you utilize AI as well to just scan and update some things there? How did that process work?

Pavel Riazanov: 4:56

Oh, that’s interesting. Yes, I did. Actually, I performed surface linguistic research on the differences between the dialects, and then I collected some data from open resources, and of course, I also consulted some uh native speakers, and and then I just uh used AI to do the transformation to implement the relevant changes to the existing TMX file. Then I had that reviewed as well, and only that I uploaded that to LMS and launched the pre-population. So it worked perfectly well. Wow. Maybe with minor inaccuracies that we could tweak later, but yeah, it actually was super fast, and as I said, it did like 70% of the of the work.

Stephanie Harris-Yee: 5:36

That’s amazing. That’s very, very cool. Okay, so if someone is in the same place that you were at, so they’re looking to change, update their workflows, etc., what is one thing, or or two things if you have them? One thing that you would recommend someone in the same situation definitely do.

Pavel Riazanov: 5:54

I would say do not start with tools. Okay, so first of all, start with clarity. Okay, before you just rush and introduce AI or reinvent a workflow, you have to get very clear on what problem you’re solving, who the audience is, and what kind of language you’re working with. That’s super important because understanding the language itself is crucial. Because some languages are highly context-dependent or morphologically complex. And while AI has been producing impressive results, like in recent months, I’m not even saying years, months. You have to be careful because for certain linguistic structures, you can’t just go all in. You need to test it in small batches, validate quality, and only then you can consider scaling.

Finally, invest in your linguists and reviewers. Okay, AI can accelerate processes, but it cannot replace cultural intuition and nuance. Okay, so for years I’ve been hearing that companies that are developing AI for localization have been intimidating the linguist that your time has come, it’s it’s time for you to go home because AI is coming to replace you. That’s not the case yet. And I can say that as a linguist myself.

Stephanie Harris-Yee: 7:02

Nice. So, what’s next for you? You started implementing AI in these types of areas, you have a good workflow. What’s in the future for you this coming year?

Pavel Riazanov: 7:11

Actually, I’m in the middle of restructuring the entire workflow of how my department is working. We’re planning on a lot of new languages next year, like a lot of new languages next year. And we’re entering a really exciting phase in localization. AI isn’t here to replace us, it’s here to remove the repetitive parts of the job so that we can focus on the strategic ones, how we shape the experiences, how we elevate quality, how we represent users more authentically. And one final thought is that the future belongs to localization teams that are adaptable and collaborative, which makes change not an a one-time event anymore. Okay, it’s not an event, it’s a skill. And the more comfortable we get with it, the more impactful our our work becomes.

Stephanie Harris-Yee: 8:02

Amazing. Okay, well, thank you, Pavel. This has been very, very interesting, and I’m so glad we had a chance to chat.

Pavel Riazanov: 8:09

Yeah, likewise, I really enjoyed that.

Pavel Riazanov

Senior Localization Manager

Linkedin