Episode Transcript
[00:00:03] Speaker A: Hello and welcome to Localization Today. My name is Eddie Arrieta. I'm the CEO here at Multilingual Magazine. Today we have a conversation with Susan Morgan, a seasoned leader in frontline sales and account management within localization and language services industry, currently serving as the Vice President of Sales AI at Lionbridge. In this role, she drives consistent growth and develops strategic solutions that meet and exceeds client expectations. Her impressive tenure at Lionbridge includes roles as Director of Enterprise Sales, Global Account Director and Strategic Account Director, which she leveraged her deep industry knowledge to support clients across the tech industry, pharmaceuticals and med devices, and enterprise accounts.
Susan, welcome to Localization Today.
[00:01:01] Speaker B: Thank you, Eddie. I'm so happy to be here, excited.
[00:01:03] Speaker A: To chat with you and we are happy to have you with us. And we met@log world52.
How was it for you? How was it for Lionbridge?
[00:01:14] Speaker B: We did meet at Lok World 52 and it was a. For me, it was a really great Lok World, in fact, actually the first LOK World I've ever had the pleasure of attending. So it was really impressive to me to see just how many people were in attendance, how diverse it was from the perspective of linguists being there, localization suppliers being there, companies being there, and just a lot of really interesting conversations around how AI is really shifting the way we think about localization as well. So it was kind of an interesting time for it to be my first attendance, but boy was it fine.
[00:01:57] Speaker A: And that is really, really great to know and I'm super happy to have you with us because this year has been really interesting in the industry. We have so many reports of companies adopting artificial intelligence and yet so many others not doing it. LSPs struggling to show the value of perhaps using more technology to some companies. And then some companies just not even wanted to understand how they can integrate AI into localization tools. So, of course you have been around the adoption of artificial intelligence within the localization industry. How has Lionbridge been a part of all these changes from your perspective?
[00:02:40] Speaker B: I think it's a great question and I think it was really showcased too in Local 52 and some of the talks that I attended. But I think you kind of hit the nail on the head, right? Some customers, buyers are extremely cautious about adopting AI. Some are really super excited about it, and most, I think, are kind of somewhere in between.
I would say that the adoption drivers at this point, really speed is the primary driver for adoption. With embedding AI in some of our localization workflows, we can absolutely reduce the time to market. And I think A lot of our customers are starting to recognize that. So I think that's probably the primary driver. But we also don't want to overlook that quality is still critically important. And we have to be really transparent with customers as localization service providers when we're using AI and what impact that may or may not have on quality and how we sort of mitigate any kind of risk to quality with the use of the AI. And so I think building trust with our customers in this time of shift is really super important at Lionbridge. I would say how Lionbridge has been a part of these changes. I mean, we are absolutely seeing customers asking more and more often to be able to deploy a deeper engagement of AI in their localization workflows. And by that I kind of mean the introduction of automated and assisted post editing solutions. And so Lionbridge has developed those solutions.
We've spent about two years now iterating on and doing continuous, ongoing development on our large language model that's doing the automated and assisted post editing functionality. So we continue to add new features, new capabilities in the models, and these are leading to different kinds of enhancements and upgrades that our customers can then take advantage of. And we're really sort of selecting which features and which enhancements we want to add into it based on market demand and the deep partnerships that we have with our clients. So I would say the other area where Lionbridge is seeing or seeing and also being a part of these big changes is that it's really touching all of our clients. Right. It's not just like our enterprise clients or our big tech clients. We're also seeing adoption of these technologies in the games industry, in the life sciences industry, which I think is especially surprising because sometimes they're a little slower to embrace new technology.
But we're also seeing it in opi, in any kind of localization, and also in AI data training services, which is a new vertical that Lionbridge has brought to bear given the advent of the AI and generative AI and large language models.
[00:05:37] Speaker A: Well, thank you, thank you for sharing that perspective. And of course, it's very different when you have an AI first culture that shapes innovation within an lsp.
What specific strategies have you implemented to help employees and team members really embrace the AI? And I know you have hinted a little bit about this already.
[00:06:00] Speaker B: So I think the AI first culture is really critical to any company who is looking to utilize generative AI or the large language models. And by creating a culture within the company that embraces AI, it helps make sure that everybody is willing to adopt it and they start getting behind the use of AI. And so here at Lionbridge, some of the specific strategies that we have implemented is that every single lion, as we like to call ourselves, has an AI goal and that's part of their annual review process. So each and every one of us has developed something specific to the work that we do that is a goal for us to utilize AI to improve something. The workflow, the outcome, the customer experience, whatever it might be. We've also made available to all of our employees an AI sandbox. So this is like an encrypted location for our large language model where we are all encouraged to go in and try things out.
Another fun thing that we do here at Lionbridge is we actually, excuse me, run a monthly AI contest. So we ask all of our employees to use AI to optimize the work they're doing daily, as I mentioned. And then we have a showcase once a month where people can come in and present the idea that they had, how they solve for it and what the outcomes were and the result. The results of this are honestly nothing short of amazing. Like we've seen some incredible impacts to our recruitment, to operational efficiencies, to client based solutions, and then beyond that, actually it's having a ripple effect within the organization and we're seeing workforce engagement at an all time high. People are not only excited, but they're really passionate about the work that they're doing. So to give you some numbers around this, our AI showcase has had 103 presentations done so far this year. We have had 290 individuals apply into our AI contest and we have seen over 5,000 unique team members entering into the sandbox and doing something, writing some kind of prompt. In fact, almost 200,000 prompts have been generated by LIONS at Lionbridge. And we're telling the AI what to do. Right. It's really exciting. So I think these are some real achievements that Lionbridge is very proud of and I think we're very proud of the AI first culture that we have developed here at Lionbridge.
[00:08:41] Speaker A: Oh, this is, this is incredible. And I was very surprised to hear the numbers on because it's all use cases and different evolutions of it and everyone learns from one another in that context. It's really exciting.
[00:08:54] Speaker B: Absolutely. It's super fun.
[00:08:57] Speaker A: Yeah, it must be really fun every week and you don't know what to expect, right?
[00:09:02] Speaker B: No. Right. Everybody kind of has a different, a different way that they look at the work and everybody has a different job that they perform and so it's really super interesting and like absolutely fascinating problems are getting solved and interesting new solutions are coming out of it that are helping people do their job better every day. Not to mention that everybody's just so super excited about something new. Right. So it's, I mean, the ripple effect has been really incredible.
[00:09:31] Speaker A: That's really refreshing to hear because otherwise you'll have an approach to the whole technology that's kind of like dull and painful and you don't want anything like that around, but integrated into. And I love to just get your perspective on this. How is AI integrated into localization tools and processes? And what direct impact does it have on creating ready to market multilingual content? How is AI really helping get there?
[00:10:06] Speaker B: Yeah, I think one of the most important things that Lionbridge has done in Recent times since AIs come on board, is that we actually have completely and totally revamped our platform that we use every day to do the work. And we now have what we're calling a global content platform platform in Aurora AI. And this is a platform that allows for workflow orchestration with our customers. And it's really meant to encompass the entire global content lifecycle, right? From content development all the way through content transformation, translation quality, all of the things that are important to us in the localization industry live inside of our Aurora AI global content platform.
And the way that we're seeing some direct impacts on this is because of the workflow orchestration. We are actually shaving time off of delivering projects throughout the project lifecycle. So imagine for a moment that tasks such as opening a zip drive, right, you have to double click to open the zip drive, then you have to pull out all of the content and you have to close the file, zip it back up. And these are all things that people had to do previously. And now we've been able to automate all of those tasks inside of the workflow. And so what that ultimately does is it, it makes it so that Lionbridge is faster at delivering a project to our customers. But what that actually does for the customer, which is the most important piece, is that it reduces their time to market. So we're starting to enable our clients to reach their customers faster in a more engaged way. And this is, this is partly because of the workflow orchestration, right? But there's a whole other factor to this that we need to consider too. We're also able to localize more rapidly with being able to use large language models to help post, edit content and reduce the amount of work that has to go to the human, we're able to actually even localize faster. So you can kind of see these little places where we gain little bits of time ultimately add up to a significant amount of time savings when we're working on these projects. And then the other piece that I think is kind of interesting to point out is that we also are utilizing large language models inside of our Aurora AI platform to help with content creation. Right. And we're predicting that content creation is going to be an element of client engagement with Lionbridge. We've been predicting that for a while now, but we're actually starting to see it come to fruition. And we're seeing clients engaging with us, wanting to leverage our LLMs for their content creation, their source creation. Right. Not just the localization component. So I think there's direct impacts in many different ways. We're basically building out all of our tools based on client demand and what we're seeing in the market so that clients can then create multilingual content all simultaneously.
[00:13:21] Speaker A: And that's really, really impressive. I'm super happy to continue hearing these. And it just seems that if we are to talk like a week later, you're going to be sharing much more advancement. So I'm really, really happy to hear how the technologies are progressing, the use cases are progressing, and how human stays at the center. Right.
So interesting how from what you're mentioning, like, humans become really critical in this conversation.
[00:13:51] Speaker B: So critical, so critical. And I know there's a lot of anxiety out there about humans becoming less important or less needed in an AI driven kind of world. And the reality is really just the opposite. I think the humans are more and more important to continue to be in the loop or at the core. There's a lot of different phrases that we use to talk about the human involvement, but I think their ability to stay involved is absolutely critical when we're using AI, because we want to be using it responsibly, we want to be using it ethically, and we want to make sure that AI is delivering accurate, relevant results. And without having a human in there to validate all of these things, we're really doing a disservice. Right. We want, we want to use the AI to enhance processes and make things go faster, but not at the expense of inaccuracy. And we really need the humans to be able to support that. And I think we need to as people, we need to be courageous and we need to embrace AI and start to think about the ways in which we're going to be interacting with it. And linguists may do a less traditional translation than they have historically, and I say may, maybe not. But they are definitely going to be interacting with the large language models and their expertise in understanding semantics and syntax and grammar and all of those kinds of things are incredibly valuable. It's just a little bit of a shift in the way in which they use their linguistic talent to work with the AI rather than using it in the way that we've traditionally seen localization done.
[00:15:36] Speaker A: That's absolutely fantastic.
And they always say this. Any model is as good as its training data. So the humans are critical to community.
So it'd be really interesting to hear from you in which ways you believe LSPs can utilize their community networks to become leaders in providing high quality AI training data. And how is Lionbridge addressing this opportunity?
[00:16:02] Speaker B: That's a brilliant question.
We here at Lionbridge, as I mentioned earlier on, have stood up, if you will, a new vertical here at Lionbridge called AI Data Training Services. And so Lionbridge AI training actually has a really deep crowd. It's not just the linguists. The linguists are a critical component of that, based on what I was just sharing, but we also have managed crowds, open crowds and curated expertise that we will go out and look for in order to support AI data training projects. And we have incredible talent within our community of individuals. We have exceptional expertise, we're able to produce incredibly high quality data sets, and it's absolutely scalable because we have such a vast network of community members. So we are in the process of currently upscaling our existing communities. This, I say this upscaling. We also kind of hear it called upskilling sometimes. But the idea here is to be able to provide our professional linguists, our crowd linguists, our managed communities.
We need to be able to provide them training and access to the AI sandbox, to prompt work to validation work, to AI training work, to diversify their skills and ensure that they continue to have the ability to take on work in this new age of using AI. Right. And so we're using our crowd to help perform different kinds of validation outputs, different kinds of linguistic testing, or upskilling them to do more than they've done before. And I think that part of the way that we have looked at this strategically is to build out a dedicated AI resourcing team. So we've actually got a fully dedicated community resourcing team building out resources, building out mentoring programs, building out training webinars, different ways in which we will help to upskill the community. And in Addition to this, something else to think about when we're trying to take care of our community network is we really need to make sure that the wellbeing of our community is kind of at the forefront. Because a lot of what happens in this AI data training space might be exposure to what we consider to be harmful content. Right. It could be criminal of nature, it could be swear words. There's lots of different types of things that might be considered harmful content. And when our community is exposed to working on content like that, we also want to make sure that we're taking care of their mental health, making sure that they're taking care of their physical health. So we've built out community wellness programs to support the contributors that are working on this kind of harmful and sensitive content.
[00:19:09] Speaker A: And in terms of that sensitivity, right. It's almost like it leads to a conversation about what's also culturally acceptable and not only in, in the corporate language, but also in things like gaming, where we're seeing in multilingual, is that it becomes increasingly important to have that cultural sensitivity. You might not be saying something that it's rude, you might not be using the wrong word, but maybe you are not applying the right context and then things become really bad very quickly. And I assume that culturalization conversation is something that's critical in this conversation of artificial intelligence as well for you as well?
[00:19:53] Speaker B: Yeah, absolutely. I mean, I think that's one of the things that is, is great about a language service provider moving into the AI data training space is that we have that cultural expertise foundationally at our core. Right. We are a diverse linguistically, geographically, we are set up with, with different cultures and different people all around the world. And so I think, you know, one of service offerings that we have in fact in our AI data training business is that cultural sensitivity training and looking at your model and doing some output validation work to find out whether or not the model is producing culturally sensitive results, if it's recognizing a locale specific prompt and providing accurate, relevant results for that culture and being able to provide that feedback to our customers so that they can further fine tune or further develop their models to, to ensure that cultural sensitivity is kept in mind when the large language models are producing their results, whatever those results might be. So absolutely, I think it's really important to have a very thorough understanding of the cultural sensitivities around the globe and being able to help train the AI models in understanding that cultural sensitivity is, is really critical.
[00:21:16] Speaker A: Excellent. Thank you. Thank you so much. And I think that's kind of like what the future looks like. It's just more sophistication, more opportunity to dig deeper, more time to kind of like add understanding to a lot of the things that are done. What do you see or foresee for AI in personalization, localization or personalizing localization solutions? How do you envision is going to impact client relationships and expectations? Just looking into 2025 and beyond.
[00:21:47] Speaker B: Sure. I mean, I think right now many clients have internal company wide strategies to adopt AI to solve their business problems. Right. Either increasing revenue, developing new product lines, creating efficiencies. There's a lot of companies out there that have these strategies in place and I think we're already seeing that because we're routinely hearing from our customers, help us, help us. We definitely need to adopt AI.
Our executives would like us to do so. How can we do that? We're also seeing a lot of companies building their own systems and fine tuning their own models. And we're seeing this both in the localization space. Right. Companies are fine tuning models to be able to do quality estimations. They're fine tuning models to be able to do translation or post editing. We're definitely seeing that. But we're also seeing clients building out systems to do things like, like manufacturing companies are potentially looking at large language models to do maintenance detection. Right. Automated maintenance detection. Or we're seeing companies that are interested in validating the content in their blog, validating the accuracy of financial content in their blog, for example. Or we're seeing companies who are looking to develop whole new solutions like autonomous driving for their vehicles. Right. So there's lots and lots of different ways in which people are looking to increase revenue or develop product lines outside of solving some of the localization problems problems, which is really super interesting.
I would also say that the market is really leaning in and we as a company, Lionbridge, are really expecting a business boom for AI data training in 2025. I think we've even seen that here in Q4 of 2024. We're starting to see more and more companies start to ask for support in fine tuning their large language models in some way. I think.
[00:24:00] Speaker A: Yes. I'm sorry, let's continue. That's great. That's great.
[00:24:04] Speaker B: I guess the last kind of thing that I would also say as a trend that we expect in 2025 when it comes to localization specifically is we are sort of expecting to see companies do supplier consolidation. And the reason why we expect to see our customers working with less partners is that the volume of content that you're pushing through a large language model, the more volume you push through it, the lower the tokenization cost becomes. And so it starts to make sense to do supplier consolidation because that volume cost reduction becomes a real incentive. Right? And we also think that part of supplier consolidation is going to play out with partners or companies looking for a language service supplier that has a diverse set of offerings. So someone who can kind of support the entire content ecosystem, right? From content development all the way through transformation and even into the space of developing large language models that help support their content ecosystem. And so I think we're going to see, we're going to see a lot of supplier consolidation and it's specifically going to kind of focus around what do the suppliers have to offer and how are the suppliers embracing AI and integrating it into their workflows.
And then the final thing that I would say is definitely on the horizon for 2025, and this is probably a whole entire separate podcast, so I won't get into it too deeply, but the ethics of AI is going to be a major factor in selecting the right kind of suppliers. I will say that we had a really enlightening, if not sort of terrifying, you know, address from David Harris at Locomorld 52 and he had a lot of really great examples of ways in which AI has been utilized in some irresponsible or unethical kinds of ways. And we really have to adhere to some regulations.
If they haven't been established yet in the US they're still working on some regulatory laws and things like that. In Europe, maybe they already have some developed. But we really have to start thinking about having some well established principles for ethical AI. And it's going to be an expectation of customers going forward. And I expect to see that in 2025, customers are going to expect that their suppliers have established guidelines, principles, rules for ethical AI. The AI is going to have to be trained on authentic, human curated content so that we are able to ensure accuracy and relevancy and that things are factual and that validation is going to be a really critical component to the life cycle of data and understanding and making sure that we're not generating any kind of harmful content, that we're not generating hallucinations, and that every everything is as it is meant to be. And so I think that's going to be one of the biggest, the biggest things in 2025 that's going to start to play a major role in companies selecting a supplier.
[00:27:32] Speaker A: That's really great and we can't wait to look at that. Consolidation over the next year or so and next few years as well. And yeah, we are keeping track on regulation. So we have seen exactly what you are saying and we are. I'm also interested in how, you know, Lion Bridge is looking at how those regulations are going to shape up. So you are completely right. A lot of the things that you're mentioning could be entire new podcast episodes that we could do. Susan, thank you so much for joining us today. Is there anything else you'd like to mention that perhaps we could, we could say before, before we go.
[00:28:15] Speaker B: I guess the last thing, the only thing I would like to add is that I think we as an industry should embrace AI. I think we have to, no matter what it's coming, right? So we might as well get excited about it, be empowered by it and be courageous and try new things with AI and really just see where it can take us because I think it can take us quite far in the industry.
[00:28:41] Speaker A: Thank you so much, Susan. And that's our conversation for today. My name is Eddie Arrieta. I'm the CEO here at Multilingual Magazine and with us was Susan Morgan. She's the VP of AI Sales at Lion Bridge. Susan, thank you so much for joining us today.
[00:29:00] Speaker B: Thank you. My pleasure. Have a wonderful day.
[00:29:06] Speaker A: And for everyone listening, thank you so much for being here. Localization today can be found on Spotify and we also share video on YouTube and snippets on LinkedIn. Thank you so much for following us. And until the next time, goodbye.