AI Maturity in Localization: Where Does Your Company Stand?

Episode 267 April 01, 2025 00:37:56
AI Maturity in Localization: Where Does Your Company Stand?
Localization Today
AI Maturity in Localization: Where Does Your Company Stand?

Apr 01 2025 | 00:37:56

/

Hosted By

Eddie Arrieta

Show Notes

What does it really mean to be “AI mature” in the localization industry?

In this episode of Localization Today, Eddie Arrieta sits down with Gabriel Karandyšovský, industry researcher and consultant at Gabe’s Lab and Argos Multilingual, to explore the AI Maturity Model—a framework designed to help localization teams understand where they stand in their AI journey. From AI-Aware to AI-Transformed, Gabe walks us through the five stages, the reasoning behind the model, and how companies can assess their current stage in less than 20 minutes.

We also discuss how different teams adopt AI at different speeds, the evolving roles within localization, and the real-world value of mapping progress.

View Full Transcript

Episode Transcript

[00:00:02] Speaker A: The following is our conversation with Gabriel Karandyšovský. He is an industry researcher, content creator and consultant at the Gabes Lab and also Argos Multilingual. Today we are talking about the AI maturity model that he has worked on and the AI Maturity assessment. It's a great conversation. We start talking about what the AI stages are for companies looking to use gen AI in their localization systems. And surprisingly enough there is a very varied mix of companies that are implementing artificial intelligence and there doesn't seem to be a clear winner in the terms of stages. We will look into five stages of that AI maturity model and how you as a company can do the test in less than 20 minutes. My name is Eddie Arrieta. I'm the CEO here at Multilingual Media. Please enjoy this amazing conversation with my friend Gabriel Karandiszewski. Thank you so much for joining us again here in localization today. [00:01:14] Speaker B: Thank you for having me. Always, always a blast. Always a pleasure. Thanks Eddie. [00:01:18] Speaker A: Of course. And you know, today for those that don't know Gabriel, he is an industry researcher, content creator and consultant at Gabe's lab and he's right now collaborating with Argos Multilingual and other ventures. If you're interested. I'm assuming you know where to find him on LinkedIn. Different, different places. But today we're going to be talking about a very exciting topic, the AI Maturity Model assessment and the AI Maturity model. I should say first the AI Maturity model and then the AI Maturity Assessment. Gabe, welcome once again. Can you tell us a little bit about how this is happening? The AI Maturity model, what is it? And let's just get to it at your own pace please sir. [00:02:05] Speaker B: Thank you. Thanks Eddie. So, yeah, I suppose let's start by explaining what this is about. Right. So the AI maturity model is the fruit of a collaboration going on with Argos Multilingual where over the, say eight past 18 months or thereabouts, we have been working with buyer side localization decision makers or localization leaders and director level folks about how they've been living the transformational journey with generative AI. And so there's multiple things that came out of it in terms of research and one sort of persistent remark or observation or something we felt would be useful to consolidate in a piece of content and an actual framework is the idea that different companies will be at different stages of the journey within AI. What's interesting though is it's not unlike with localization. So you know, if you, if you think about it, different companies, you know, you have the startups then you have the well established ones. Everyone is at the different stage with their localization process from starting up to, you know, on the other side of the spectrum, people who have mastered it all. And so it's, it's very similar with Genai. Some people are still just starting up out and some are well advanced. And so the interesting bit though was compared to localization because where localization has been going on for a while now with Genai, it all sort of started at the end of 2022 and no one really had very much room to prepare for it. And so everyone was starting from the same place. So you see these sort of almost two tracks going in parallel where on one end you had the localization, globalization part of the works and then Genai kicked in and you were well advanced on one and back or at the beginning with the other. So yes, it's a journey and that journey is unique for every single company. But also there are commonalities that emerge with companies in sort of similar stages and experiencing similar things. And so all of that led us to, yeah, thinking maybe there is a, maybe there's a piece of knowledge, piece of resource that we can create to help different companies navigate this journey. And so, yeah, that's the long winded answer to how the AI maturity model came about. Now, if you ask me for a definition, so we do have a definition which is more textbook, like maybe I should have started with that. But the way we define AI maturity within the context of global language operations is that it's a measure of the team's awareness and readiness to integrate AI capabilities successfully. So yeah, how ready, how aware you are of everything that AI can do and where you can implement it. So yeah, that's the AI maturity model. [00:05:36] Speaker A: And of course, as commonalities started standing out, I presume the challenge was how do you buckle them and when do you know a stage has ended and when do you know another stage stage has started? And I assume there were many other hurdles along the way. Could you speak a little bit about that process and how it developed? [00:06:01] Speaker B: Right, very good question. And I think it's also opportunity to mention here that us coming up with a framework isn't something that doesn't have precedent. Right. And it's important to give a shout out to, to Common sense advisory who have come up with the localization and globalization maturity models that have been an influence in this and are by now well known in the industry. So that had an effect in terms of how you display how you portray that journey. And so for Those who know this will sound familiar, but for those who don't know and the way we, we wanted to start with a visual representation of the maturity and the journey that goes with AI. So essentially you have five steps to it. You have from left to right or from top to bottom. You're starting with something that we named AI Aware. And then you have five steps in total and you arrive at AI Transform. So that's the ultimate stage of, of Gen AI adoption or AI adoption. And in a way the exercise in itself was fun and maybe sort of answering your question here, as in you kind of see where different companies are. And what was also interesting was trying to put them on this map of sorts from left to right and trying to pinpoint their location. And from there, sort of naturally it emerged what the different stages could look like, what their characteristics, characteristics would be, as you can imagine. And we'll maybe get to the detail a little bit further. But every stage has its characteristics and sort of also the marker, you know, as a milestone, as in, okay, at this point you're jumping to the next stage or you're, you're able to say that, okay, I'm no longer just AI Aware, which is the first step, but I'm also AI active. So yes, it's been a combination of elements, essentially drawing inspiration from something that is, that exists, but also seeing that and wanting something very visual, visual aid that can help companies situate themselves, but also understand where they are and what are the steps they need to accomplish or tick off in a way to move to the next stage. [00:08:35] Speaker A: And that is great. And of course before we move on to the next stages and just outline the five stages in the AI Maturity model and those that are listening and say, okay, I'm five minutes in, maybe the question is a little too late, but who is this model good for and who is the assessment intended for? And then we can get into the details so that they know what they'll find. [00:08:57] Speaker B: Right, right, Very good. So it's a couple of things right, as we're mentioning keywords and or names here. So the AI Maturity Model is a, is a theoretical framework, so theory visual with text and that anyone can access and sort of try to incorporate. And then the AI Maturity Model assessment is the second step, which is more of a hands on. Okay, let me try to really assess and pinpoint our location here. So two things that are sequential in a way, theoretical and practical. And as far as the audience goes, it's really targeted at, I would say practically any company of any size that has a formalized localization or globalization function, whether they are just now starting out and they are just now consolidating processes or around language related activity. Although you would think that, you know, it's 2025. Are there still companies that don't have, you know, a localization process or translation process? Probably there are still some, but also the companies that are, that are more advanced, more mature as it were. Because for them also it's a way of sanity checking or validating whether they have thought about all the different, you know, facets of AI in localization. So very broad audience. Everyone will take something different out of it with these things. It's a resource you can tap into to draw inspiration from, try and understand where you are and how far you still have to go. [00:10:51] Speaker A: And that's fantastic because then every company can start looking into what it is that they are doing or not doing. You'd be surprised what I found. In many, many ecosystems that I've been into. In most countries there are companies doing localization that they don't know they are part of a localization industry. They are like the corpus sapiens. They are corpus sapiens. They are not corpus sapiens sapiens. They don't know who they are. They are part of a much larger, much larger corpus of companies. It's very interesting those that perhaps are looking to do localization and are not yet aware of, of how it's actually done, or they're doing it internally just because they don't know any better. [00:11:35] Speaker B: In the. Yeah, the classic. Anyone you know, send it to your colleague next door. You can translate it. He speaks FRENCH, right? [00:11:42] Speaker A: Yeah. [00:11:43] Speaker B: So there are still many of those, I would think. [00:11:46] Speaker A: Yeah, a bit like that. And I like all these, all these standards and assessments that allow to look into these levels and stages, in this case of AI maturity. So let's get into it. Tell us more about the model. Tell us about the five stages that you saw in there and what they involve conceptually. [00:12:05] Speaker B: Sure, very well. So I mentioned top to bottom, that's the main visual we have for this. We have design for it. So it goes from AI Aware to AI Active, AI Operational is a third step, then AI Systemic and AI transformed. And so we do have a little bit of definition for each of these. So the more succinct version would be so AI Aware, the very first step, it's for teams or individuals that know that AI exists, but they haven't answered the question of whether they should pursue it or not as part of their language related activities. So yeah, you're aware it's out there, but you're not really sure. Okay, am I going after it or am I testing or am I not testing? Then there's the second step, the AI Active. And the sort of progression is straightforward here. That's for teams that are actually getting busy with AI. So there's a period of learning, or the period of learning is actually continuous. In many cases that goes hand in hand with, with hands on experimentation and hands on testing. So that's the AI active stage of the model where you're actually doing stuff with Genai, or at least getting to know it. Then the third step, and this is where we make the distinction of companies that have actually rolled out AI powered processes as part of their workflows in real life production. So this is the stage between, okay, you've done the tests, you've tried out things, now you're ready to roll it out in actual production. So that's the AI operational. The fourth one is AI systemic. And that in a way is you have a system that is enhanced by AI. So AI permeates multiple components of the localization process. It appears in multiple sort of work streams across the board. We know there are dozens and dozens of use cases. So it's really to distinguish companies that have made AI a core component that influences so many, so many levers, essentially. And then the last step, I would say the elusive step, and we're, I think, myself, I'm, I'm, I'm on the hunt for this kind of company. I haven't come across it just yet. And that is the AI Transformed company. And the, the way we define it, the way we see it, is that the, the localization process has been built with AI from the ground up. And why I'm saying it's elusive, that's an assumption of mine. I'm sure there are some out there who have made it, but the assumption is that AI came at a point where a lot of folks have been well advanced with localization. So for them, they're not necessarily transforming, they're adding something, they are adding AI on top. Although you could argue about the word transforming. They are transforming things as they do. So but for those guys who have been around, they have not deleted everything that went before just to plug in AI. Right. So the AI Transformed is really the type of company that is able to flip localization operations on its head. The classical view we have of localization and really just, yeah, use AI from the ground up. So, yes, those are the five stages and how we Sort of see and define them. Of course there are multiple sort of, you know, what are the stage defining criteria? There's a little bit more that goes into it, but that's the stuff that powers the assessment that I mentioned. [00:16:17] Speaker A: All right, and we'll talk a little bit about the assessment in just a few minutes. But I'm very curious about this AI transform concept and I'm hoping we can dig a little bit deeper. And it's because for most people, when they think about stages, they think about goals, they feel that once goals are achieved, the process ends. But AI transform involve a continuous and dynamic approach to the structure itself, because that's how AI it's naturally done. It's a machine learning system that receives inputs and transforms itself, hopefully for the better. So could you tell us a little bit about this ideal company, a little bit more about what an AI transformed company does internally, how does it operate? What are some of the more salient things, more obvious things that you would start seeing when, when they happen? Because what I'm going to assume is there are, there are a, there are companies that are trans AI transforming themselves. They are in, they are in the process. And that's why, of course, you're doing the assessment. But perhaps there are some signs that can tell companies, oh, I am transforming, but I should do the assessment because I'm still in operational. But there are these ideas that can help me understand that I am on the way and this assessment makes sense to me. [00:17:43] Speaker B: Very good question. I love it. And there's a couple of things that come to mind, and one of them being the, okay, what is the sort of a, sort of loose definition or the definition we have of an AI transformed company? Because indeed you're saying it's true, and I agree with that, that going through this journey, you're essentially transforming on the go. So yes, yes to that. Before I sort of maybe unpack a few of those characteristics, there's an element here where what we noticed, and I mentioned this, is all research based, qualitative research based. When talking to those leaders and decision makers on the company side, what we noticed is that on one hand you have the AI maturity model, which is a sort of a rigid structure. You imagine it with steps. And we've tried to come up with characteristics and sort of those milestones I mentioned, but the reality is a lot more nuanced than being able to say, yeah, company X is a active and company Z is AI systemic. Because actually what we see is that the characteristics from each of those steps can be reoccurring at the previous stages. Right. So it's, you're never, it's never black and white, you're never just in one stage and you're done, as you're saying, you're never done with AI anyway. But yeah, you may be AI active, as in only now being in the process of experimentation and testing. You have not yet rolled out AI as part of live operations, but you already may have some of the characteristics of a more advanced AI systemic or AI transformed company. So yeah, that just goes to show that A, everyone's situation is so unique, B, everyone moves at different speeds in terms of being able to learn and absorb and implement AI. And C, as you say, you're never done because there's always something to tinker with. So I just wanted to make that important sort of distinction or add that little bit of nuance here is that what we are seeing in reality is that everyone is at different stages of the maturity track. And sometimes you skip two steps ahead in one area, but in another area you're three steps behind. But back to your point, and I know I'm talking a lot here, but to your point about AI transformed companies, so the ultimate stage that we have, in a way I mentioned, okay, so one of the elements here is that the whole language works, the localization operation is built with AI from the ground up. That is one thing. So AI is very central to things. But there's also another element here that's part of the assessment, for example, is the scope of activities that are AI powered and what many may have thought about before, something I've seen certainly over the past few years doing the research, is that companies tend to start out with transforming content, transforming assets. So the acts of translation, localization, interpreting and so on and trying to heading further upstream or the shift left, that's a different term used to it. And teams being more present at source when with content creation. And so that's been a sort of the holy grail for many of these companies, right? Trying to be more present where they can, where they contribute to actual creation of the assets, not just transforming them. And so that's where for an AI transform company, the way we define it at least is, is they, they are there, they are at the source. Because with Genai, you already can create content in multilingual, you know, in multilingual format in many languages. Not many are doing it yet. It's just one of those many use cases. But we see that as, as sort of one of the, one of the defining use cases. If you're, or if you've mastered transformation, if you master the place of localization in the organization and you're closer to content creation, that's one of those elements of being AI transformed. And maybe a third sort of big component or big characteristic here is that is on the side of the teams, of the manpower you have internally, of the knowledge you have internally influencing stuff like strategy and human resources. But one element here is that in terms of being able to measure stuff like roi, having analytics, being able to model the ROI of AI that is core component. So you have the sort of visibility, you're able to articulate that. But also when it comes to the team. So the teams being very AI fluence and being trained and upskilled to a point where they are the de facto experts. So that's another characteristic here where it's no longer about them being project managers or even program managers, but technical and language experts and having that position within the company. So there's also the AI transformed company. Yeah, there's an evolution of roles that goes with it as well. [00:23:41] Speaker A: This is a really good point because we've been working on a podcast called Lang Talent and one of the things we've noticed is that there's this default thinking of localization industry. Translators, interpreters, linguists. But there are also many other roles. Software developers, designers, janitors, mail people. There's so many people that make this industry thrive. And the higher the awareness we have, I guess the higher the propensity to generate true innovation because then we won't think that the industry is limited to a specific roles. So that's a really great point you are looking into right there. Of course, perfect segue as well to talk about the test. So if I'm to do the test, I know it's a very short test, 10 to 20 minutes it would take me to do. What can I expect when I get into it? [00:24:41] Speaker B: Yes. So all of it sits online on the Argos web pages and then there's a button that goes take the test, take the assessment. And essentially what it is in short is it's a self service assessment. It's online. It takes you between 15 to 20 minutes. People have option also to save progress and get back to it later because we do have a lot of questions. There are like 52 questions if you, if you, there's, you know, stuff like skip logic and such. But if you answer everything. Yeah, there's 50 questions and we go. So the way we position the test is and, and that's the sort of you know, the, the mirror. The opposite to the AI maturity model is that, okay, we are creating the, this model based on four things essentially, and its operations. So there's four areas, four key areas of the localization program that we are assessing on the go. It's operations and so how AI fits into operations, people and knowledge, technology and strategy. So four areas of the assessments, each with between, I would say, 8 to 12 questions per per section, again, depending on how you answer and the outcome of that as you go through it. As you finish. The end screen is a principal mailable result which says you're in this category or in this stage of the AI maturity track. So there's a little bit of text there. It sort of explains the score as well, because this has a scoring component whereby you could in theory, in practice, take the test of few months down the road and see if you've sort of scored better because you, you know, because of the evolution, because you've changed things. And there's also sort of customized recommendations. So as you, as people go through the test and select their answers, we're also trying to provide concrete recommendations based on research again, but sort of showcasing, based on their answer, what is the potential for progression? So what are others doing or what are the areas to consider? What else could you do? So in the end, the result page, it's I think two, three pages of a PDF essentially, but also with specific recommendations as to areas to look into more detail if they want to progress. And I think that's another sort of core objective of the maturity model, but also the assessment that goes with it. It's not only about getting a result. You're on stage one of five or three out of five. But what we feel strongly about is that it's also about broadening your horizons. So when you go into the test, you don't know what the questions are going to look like. But also the way that we frame the questions is because we want to push and prod people to broaden their thinking, broaden their horizon and think, oh, okay, I should be thinking about this as well because it has a bearing on our progression and our capacity to do better with AI. So it's not about the destination, it's about the journey. Right. And the test is part of the journey. And so you kind of draw learnings from that as well. [00:28:22] Speaker A: It's great because it of course points to the different instruments that it's going to point to that you should be able to utilize or that you should use right now that others are using. So it gives an idea of what's happening there, I'd say. And I'm going to ask this just because I know it's already addressed within the website. And then I have a second question on that. The first one is, okay, what about my privacy? I'm gonna be giving you all this information and you know, I'm worried because my company has a lot of private information. I don't want you to use it inappropriately. Could you tell me a little bit about privacy? But is there also a follow up by the Argos team? So I gave you my information, I gave you my data, you gave me some pointers and I really need help. Are you gonna leave me now to my luck or is Argos Multilingual gonna follow up with me on some of these things and help me along the way? [00:29:18] Speaker B: Yeah, a couple of things. So first of all, everything is GDPR compliant and complying with most of the regulations that are standard by now. So just to make that absolutely clear, that's one thing. Another thing is that there's a very small team, as in two people with access, so restricting the access as well. The stuff is kept only for a certain duration and so on. And so I would say sort of standard procedures as far as entering information. We don't necessarily ask for identifiers beyond name and email insofar as this is getting mailed to the people. The result is getting mailed to the people. But that's the only thing. Of course from an email you can sort of deduce what the company is unless they use their Gmail. So but yeah, that's sort of where this sits then in terms of follow up. So yes, there is a component here whereby the result gets emailed, but there's also an option here to get contacted in case you want to get a customized assessment. Because that's sort of another, another part of the, the assessment is the ability to do it live with, with a, with the consultants, with an Argos person on, on the other side of the, the, the screen who then does the test with you and provides a more customized assessment so where what you see online can go only into a certain level of depth, then there's a more deep, more nuanced, more recommendation heavy assessment that we can create in a more controlled environment with the person. And we've done that. So yeah, that's another component as far as follow up. [00:31:26] Speaker A: And Gabriel, you've made some comments throughout the conversation about companies that have helped you build the models with their information. From what you've seen so far, you know, companies doing the assessment, what readings have you had? Do you have more aware companies? Do you have companies going into the transform? Do you have companies really interested in taking onto the journey in a more aware manner, in a more conscious manner for their teams? [00:31:55] Speaker B: Good question. I don't think there's just one answer to that, honestly, because definitely what we've seen is, is you have a variety, you have a richness of situations. And so it's not like any type of company is standing out more than the other because their situations are so unique that it's very early, in a way. And this has been out for a couple of months, so it's still in that sort of testing phase as well, to see how we can improve it and what other bells and whistles we can add. So early to say whether anything stands out. However, I would say that yes, when sort of looking back at how this was created and the types of feedback we're getting, if we, you know, if I think about the five stages of the maturity model, everyone is between 1 and 4 in most cases. But as I, as I mentioned a few moments ago, you have people who are very advanced, who tend to be very advanced in one area, but then, you know, a few steps back in another. And this is, this is just a. Underscores the complexity of AI as part of language operations because it influences so many levels. It's not just about, you know, selecting text, seeing how it fits within the tech stack, whether it plays nice with the TMS or whether the TMS has AI functionality. The progress happens on the level of the team. So you can be very advanced from a tech standpoint and have the perfect solution, but your team is not yet fluent enough or upskilled enough in a way to be able to leverage and innovate and imagine new uses. So, yeah, there's all these, you know, in a way, you're, you're playing with AI, you're. You have a chessboard, right? That's the sort of imagery that comes to mind and you're moving chess pieces across different, different parts. Right, so, so, yeah, a lot of differences. Everyone is sort of progressing. I'm almost thinking that the last five, fifth stage is sort of is just out there as something to maybe grab, you know, try to trend toward whether someone will get there as far, you know, within the confines of how we define it. I'm not sure and I don't think anyone will be ever done with AI, as you mentioned and we talked about already. So, you know, let's see. Ask me the question again in two years. [00:34:44] Speaker A: It's a great challenge. It's a great challenge. AI transformed and you know, just thinking about the potential business models that could allow for it and then you can just push it so far, right. Like it's 100% AI built and whatnot. A few days ago I saw an AI bot was tasked with the, with the, with the. Well, was tasked to invest into a pool of companies. So the AI would listen or read also and listen to the pitches and then it would decide who get a $10,000, which is amazing. I think it's great. Especially if you bet all of this before knowing that they are all good. You just have to take a decision. But of course, we're coming to an end to our conversation. I just want to ask if there's anything that you think shouldn't be left on set. You have any closing remarks, any advice for companies thinking about taking the assessment? [00:35:40] Speaker B: Yeah, good. I suppose the closing ideas is that, yeah, embrace that this is a journey and be prepared for it to be a journey that really never ends. But there are. Despite that, you're able to progress and measure the progress. It's going to boil down at looking at all the different facets of where and how you implement AI origin AI. It will also be about clearly defining a vision and the objectives that go with it and knowing when you are fulfilling them. So knowing what are those milestones for you so that you know that you're progressing. The good thing is that as you know, as the months go by, our industry is producing more and more resources to be able to, to progress and this is one of them. So, yeah, I suppose the call to action here is if you're ever in need and sort of want to validate your thinking or see what else is out there and whether you're covering all the angles, the AI maturity model and the assessment are made for you, essentially. [00:37:02] Speaker A: Gabriel, thank you so much for joining us today. [00:37:05] Speaker B: Thank you, Eddie. Thanks for having me. It's been a blast. [00:37:11] Speaker A: And to everyone who is listening to us, remember Gabe's work and Gabe's words. Embrace the journey, be conscious about what your stage and know what your milestones are. Gabrielle Karandovsky from Gabe's lab and Argo's new multilingual. Always a pleasure to have him with us. Remember we just talked about the AI maturity model and assessment. This is an option for companies looking into integrating AI within their system. My name is Eddie Arrieta. I'm the CEO here at Multilingual Media and this was localization today. Thank you so much, everyone, for listening. Goodbye, Gabe. Goodbye. [00:37:50] Speaker B: Thanks, Eddie. Bye.

Other Episodes

Episode 72

July 14, 2023 00:03:37
Episode Cover

The false promise of generative AI detectors

Is the promise of AI detectors a false one? We dive into a Stanford University study revealing surprising biases in generative AI detectors. Join...

Listen

Episode 84

May 02, 2022 00:03:11
Episode Cover

The Great Translation Movement translates Chinese social media, public opinion

A team of anonymous Chinese speakers is attempting to shine light on misinformation and propaganda on Chinese-language social media. The group launched The Great...

Listen

Episode 51

May 01, 2023 00:08:02
Episode Cover

Take the Plunge How to dive into vendor management and enjoy it

Camila Amici reviews Carmen Cisneros’ Take the Plunge — How to Dive into Vendor Management and Enjoy It, and had more than a few...

Listen