#269 Milestone Localization Founder on Automated Glossaries, LSI Leadership, AI Fatigue

0:00In the last 16 months, there's been so much new technology. There's so many new tools available. There's a lot of noise and a lot of hype, of course, especially around AI. And there's been a lot of changes in client expectations. There's a lot of pricing pressure.

0:21Hey, everyone, and welcome to another episode of SlatorPod. Today on the podcast, we welcome Nikita Agarwal. She is the founder of Language Solutions Integrator Milestone Localization. Milestone also recently launched a new AI-enabled localization product called Kavya AI. So hi, Nikita, and thanks so much for joining today. Hi, Florian. Really excited to be here. So where in the world are you recording this from? We always ask this question at the beginning of the podcast. Yes, I'm in Bangalore in the south of India. Bangalore in the south of India. All right. So tell us more about first, let's get started with kind of the professional background, entrepreneurial journey.

1:04So Milestone, you know, isn't a 20-year-old, 30-year-old kind of old school localization company, but it's relatively new. So maybe just tell us how you got into the space, how you started Milestone. Yes. So I don't have a very long professional journey to talk about. I started Milestone Localization in 2020. It happened to happen in the middle of when the pandemic started. So before this, I used to work in global sales. And I used to travel a lot in Europe and sometimes in China. So when we were traveling there, we used to use translation services. So we used to use interpreters for some of our meetings, and we used translation services for the websites and brochures.

1:48And that's when I first came across this industry. Before that, I didn't know it existed. And the person who used to do some of the languages for us used to call themselves a localization professional. And that's when I first heard of the word localization. At this point, I was very actively looking for something of my own to start. and when I came across this industry, I was like, okay, this seems super interesting. It's global and I was kind of looking for something which didn't have a huge capital expenditure. So when you're doing services, the barrier to entry is really low. So I started doing some research and it sounded like something fun.

2:27Honestly, I was 25 when I started this. So yeah, that's how I started. I'm not a linguist like most people in this industry. I just found the industry interesting and it happened to find me at a time when I was looking for something new to do. Very cool. Like, I like what you mentioned with the kind of low barrier to entry because that is certainly true. Like, you can start with a ton of capital and go in and build something really crazy and big, but you can also start with very little and just kind of start getting your first clients, right? Yes. So, and relatively a low, yeah, initial investment, I guess.

3:04So, how did you get started? What were some of the first clients you got, some of the first accounts you landed? Because if you're in sales, maybe that helped getting some of the first clients? So what I started with is luckily where I am, there's a lot of small translation companies. And as you know, in India, a lot of the big companies have their offices where they do fulfillment with thousands of employees. So I started with three people and one of them was sales and one of them was projects and one of them was a bit more technical. The very first project we got was one of their old clients who moved to a new company and they needed translation.

3:43So they got in touch with my project manager. And the person I hired was from sales in this industry. So he was able to contact a lot of previous clients and get things started. So I knew I wanted to, instead of figuring out everything myself, get people from the industry and then start. Very cool. I want to talk about sales a little bit later on as well. I think it's one of those industries where if you're really excelling at sales, I think you can get really, really far. So, you know, let's speak about that in a second. But then, so we spoke about kind of what the idea originally, you know, kind of where you got the idea from.

4:24But now it's five years later. So how is it to run an LSI in late 2025? Let's give us kind of the broad overview, because so much changed since 2020, obviously. Yeah, I think I was deflecting on, you know, if it was a good decision, if it was the right industry, because this year has felt incredibly overwhelming. I think not just this year, the last 16 months. And it feels more overwhelming than the first two years when I started this. For me, when I started, it was a completely new industry. but in the last 16 months there's been so much new technology there's so many new tools available there's a lot of noise and a lot of hype of course especially around AI and there's been a lot of changes in client expectations there's a lot of pricing pressure this year more than ever felt like running to stay in the same place and a lot of difficult client conversations and what I've been seeing is that some specific industries are way more affected, for example, software localization, website translations, legal.

5:34Those are where we've been having the most difficult conversations, where they want to use complete AI, or I would say even up to like 95 % AI with only some few pages having human translations. Luckily for us, which I'll get into more later, is we do a lot of work in the regulated space. So we've been a bit insulated. It's also like if you watch all the different webinars and you see things on LinkedIn, everyone is feeling a lot of pressure. So I would say this year has been stressful. But I also want to say that it's been really fun. I've completely had to rethink which sectors we want to work on, where we want to focus our resources, and how to make our processes more efficient.

6:18So I've made more decisions in the last one and a half years on where to go with the business than I did in the last five years. So yeah, I think it's been challenging, but it's helped us create new strategies, which we'll have to see how they pan out. So you need to balance kind of that service, solution, technology, AI, like, and you said you needed to take, you know, so many decisions over the past, let's say 12, 16 months. So how do you do that? How do you balance that? Do you have somebody like some sparring partner to work on some of these decisions? You make these decisions on your own?

6:53How do you do that? And then how do you balance it? Do you say, okay, at the core of a service org or a solution integrator or you're saying, well, actually long-term, we want to go towards tech only? Yeah, just walk me through that decision-making process a little bit. So for me, it depends on, I think, for any company, which service they're in. If they're working with an industry which requires a lot of automation, then this decision becomes more on how can we implement tech very fast. Luckily for us, most of our work is in life sciences, and that's a very heavily regulated industry where they're not looking for too much automation.

7:32It's still more focused on process and accuracy. But keeping that in mind and thinking of the next five years, which is difficult to know what will happen. What I've been trying to do is keep our services steady. That's where we get our money. That's what I've built out in the last five years. So I have a sales team which handles the service sales and we have really good processes. So what I'm trying to do is spend, let's say, 50 to 60 percent of my time on that and set up a good system where it can be managed. And the other 30 to 40 percent. I'm trying to see what new technologies we can introduce, whether we can offer something new to our clients.

8:14That's how Kavya came about. And also, what can we do in our processes to make them more efficient so we're more competitive down the line? So it's been more about balancing my time. I do have other translation integrator, LSI company owners I talk to and then other people who are in tech who I discuss these with, but I mostly make these decisions by myself with inputs from other people. I want to talk about life sciences, but just before, if you're originally from a sales function, how do you resist the temptation to just sell, sell, sell, always hop on calls yourself and just, you know, keep feeding the funnel?

8:51I don't resist it because I think me being on initial calls makes a huge difference, right? And what I do is I do I still do a lot of sales myself. I would say 30 to 40 % of the sales are still done by me. But for my sales team, I'm always on the first call. And then if it's a very big client, they expect you to be involved. And it's a very long sales process. But if it's not, then I let my sales team take it forward. But 90 % of the calls my sales team does, I still attend the first call. It's very helpful also to understand, you know, obviously the customers, what the pain points are, etc. so let's talk about life sciences so what what areas in life sciences are you kind of strong at in particular more the clinical medical maybe or more horizontal like content like you know marketing things like that and and is there any particular language focus you know being being very strong obviously in the in in india even though we're in india like less than five percent of our work is indian languages and for me the reason for that is very simple because i think Indian languages are the cheapest.

9:58So you need to do 10 times the volume to do the same revenue you would for, let's say, Scandinavian languages or Japanese. We have a very specific focus within life sciences. We do regulatory work. And even within life sciences, majority of our work is medical devices. It happened completely by chance in 2021. So more than four years ago, we got one big company which contacted us because of the new EU MDR regulation, which requires you to work with an ISO 17100 certified company. And there's only a handful of those in India. So they contacted us to work with us. And from there, it just took off.

10:40Now we have three ISO certifications. And I have one salesperson who's only dedicated to selling services to medical device companies. so we do all the regulatory work like for regulatory submissions mostly in europe so most of the work we do 90 is european languages so okay medical advice and then you you were saying that you have a dedicated salesperson just for that i think that's just super essential i think that's what so many kind of maybe new entrants to the space even on the ai side or tech side underestimate that like if you if you hop on a call um and you don't know about these regulations, about EU, MDR, et cetera?

11:19I mean, the client just wouldn't take you seriously. Like you only get into these types of big projects if you have salespeople that are extremely knowledgeable in that. I think it took me longer to understand this than I wish it did. Because when I started, we were kind of trying to do everything. And we actually did a really good sales training last year. And from there, we just decided to have each salesperson specialize in one sub-industry, not even one industry. And that's made a huge difference. I think this year, the reason we've seen growth is because of this hyper-focus. And it also makes it way easier to train the salesperson to make templates for your emails, for your cold calls, and also to decide what kind of content to put out.

12:03So whenever we get a salesperson, it's very specifically for one target industry and we just train them on that. It takes courage though to like say, all right, look, let's focus on this and let's maybe pause that. And we're not going to take on anything or everything that's coming into the funnel, right? Takes a bit of courage. I want to talk about pricing, just briefly. Pricing trends, you mentioned it before. What do you observe? I mean, is it just down, down, down on the unit costs? Is there maybe new models that you can apply to clients, maybe our or other kind of more innovative things that we've been talking about for like a decade in this industry now?

12:40I think pricing has been one of the hardest things this year, right? Everyone you talk to, they expect translation services to cost less. And when you tell them, they don't really want to listen to what it takes to get a good quality translation. And sometimes it's not even just the person you're talking to, right? They'll say that, okay, you've explained all of this to me, but management has said that they want to automate processes. And every company is kind of feeling pressure to see efficiencies from AI. And one of the easiest things for them to say is to automate translations. So I think there's a lot of pricing pressure and like I said, more specifically in software and legal and websites, but we haven't seen pricing pressure with regulatory.

13:28We work with RAQA teams and for them quality is extremely important. We have had some people wanting to discuss how we can use AI, but what we tell them And what we actually do is that we use AI to make sure the quality is better by doing more checks, by doing better preparation, by making things faster. But we can't reduce the price of the translation because a good medical translator still has to go through the files. So we haven't experimented with any other types of pricing. We do hourly for some functions like review, which is, again, that's what's been traditionally done. But what I'm seeing and what we're trying to do is even to offer more advisory and offer more holistic language services instead of just translation, because people don't know how to assess AI quality.

14:19They don't know how to implement translation AI. So we are trying to work on offering AI consulting and language consulting instead of just the per word translations. All right. And you've also got into products with what we mentioned before, Kavya.ai. So tell us a bit more about what is it? Why did you develop it? What was kind of the traction so far? And just to add to this long question, are you kind of using it internally? And yeah, just tell us more about Kavya.ai. Let me start with what it is. So it is what we call a project preparation platform. It is meant to solve the first mile of translation projects, which according to me is the most important.

15:06You have to create a good foundation for a successful project. So it's not another tool which translates content. It's built for linguists and project managers. And we have three tools in the platform. So one is a glossary maker. So this helps you extract glossaries for your project. we have a style guide generator which creates a very very comprehensive style guide for your project including target audience tone how to handle gender measurements currencies everything and we have a smart document analyzer so what we've seen is something that would take a project manager or a linguist four hours to do which would be research manual hunting reading entering can be done in less than four minutes with Kavya.

15:54So you just need to upload your documents onto the platform, select which tool you want to use, and you get very, very comprehensive documentation, which you can actually use with just a few minor edits. So this actually didn't start off as something that we wanted to sell. When you're a small LSP, LSI, sorry, when you're a small LSI, what happens is you end up getting a lot of small projects. So a lot of the projects, which we still do, are under$500. So when you get a project, which let's say for an example is$150, you don't want to spend a lot of time creating documentation because that cuts into the margin.

16:34So you need to create a style guide to maintain consistency, or you want to create a glossary just to get the client and your team and the translators on the same page, you might skip this step. And what I've seen is these are the small projects where you skip preparation, which really come back later with issues. And then it's frustrating because you are spending way more than you wanted to on this project, especially for non-recording clients, just one-off projects. So as a company, we really value quality. And what we want to do is give the same kind of preparation to each project. So when AI came in, we tried it in the first year, it was giving awful output.

17:14But last year, we started using it to create our glossaries and style guides, and it got really, really good. So I initially created this as an internal tool for my project managers to use. And we started using it in January. We did a lot of testing. We created custom templates. We created custom instructions. We tried five or six different AI tools until finalizing the ones that work best. And then like in, I think, May this year, I thought, OK, this my team loves this product. It's made preparation so easy. And my favorite part is it makes preparation the same level for like a project manager who's just joined a few weeks ago versus someone who's been working with us for a few years.

18:00So everyone's able to produce really good documentation. And we're able to do this for every single project. So that's when I decided to kind of make it public and do beta testing. So we just launched it publicly, I think, a few weeks ago. We did an article on Slator and we did a press release. So we're still seeing initial traction, but we have around 200 users right now. And we have a few signups every single day. So that's how it came in. But for me, something very interesting is I thought linguists would be the ones using it. But we're seeing very high usage from project managers and small LSP owners.

18:42Congrats. Great. Great. Great to see that early traction. Also, very smart. Obviously, you're turning basically a cost into a revenue generator, right? So this is a it's very interesting. Now, so didn't your existing kind of language technology platform, TMS system, etc. So they didn't solve that problem or was it just too kind of complex, too, I don't know, too hard to do or didn't offer it at all? So I don't think there's any translation platform which offers generation of style guides. And there isn't one which does glossary extraction. Of course, there are tools which analyze the document and traditionally based on how often a term appears, they put it in the term list.

19:24But what we did is we created glossaries based on the type of document. So when you use Kavya, you can choose your document type. So we have different templates for patents, for agreements, for instructions for use, for a learning document. And the terms are extracted based on the document type. And they're even grouped. So you'll have regulatory terms. You'll have chemical names. You'll have acronyms. You'll have regulatory body names. So we created very specific templates for more than 15 different document types. And that's what actually made a huge difference in generating a good glossary.

20:01So we talked about selling, but now you're going from selling a complex service with a really long sales cycle to selling a product, selling SaaS software as a service. are you selling it like like with salespeople or you're you're trying to kind of the product-led marketing approach where you know you're seeing people come in from various sources from the internet basically so right now this isn't like it doesn't have a subscription it's a pay-as-you-go model and we're giving everyone free credits when you have a new ai tool you want to allow people to try it out and see if it adds value to their workflows.

20:42So right now we're giving everyone 10 free credits and then you can just purchase as many credits as you want. So it's kind of in the experimentation phase and out of the three tools, the glossary is used way, way, way more than style guides. Another thing we've learned is a lot of companies don't make style guides. That's the feedback we've gotten. And when you ask them why they don't, because to me, it's a very important document. They say that no one reads them, which we have seen. Brutal. Yeah. We've seen translators. Yeah, the translators just don't read them and will say, you know, this said formal, but you've done it informal.

21:18And I get it from the translator's perspective because sometimes you're giving them like a small project and giving them so much documentation to read. So right now, the glossary is the standout tool out of the three that we've launched. So we're still very much in experimentation phases and then we'll basically see how they can integrate this with their workflows. So we're not doing sales because it's not a very high value product as of now. So I'm enjoying doing marketing. We're doing a lot of email campaigns. We're trying influencer marketing. And yeah, we're doing LinkedIn and we're doing groups.

21:54So that's how we've been doing it so far. So it's been fun to do this. Got it. Yeah, it's quite the learning curve. I mean, we had something similar with our website called Lock Jobs. I think we ran it for about three to four years. You know, it was a great experience, but now it's gone. Anyway, I hope Kavea will be, I'm sure that Kavea will be much more successful. So let's talk about maybe briefly, I know that you're not super actively selling into the Indian market, but I do want to talk a bit about the Indian market, the domestic Indian market, kind of the opportunity for LSI's that are there, maybe the state of tech and AI adoption, You know, maybe has the Indian market like leapfrogged and just gone full automation because of kind of domestic pressures?

22:37Or, yeah, what are your thoughts on the Indian market in general? Okay, so like everything about India, I think that's a complicated and multilayered question, right? There's so many different languages here to start off with. And second, like I think even in most countries, there's a company with two or three people and then there's some really big companies here. but just let's talk about the languages first so India has more than 20 officially recognized languages and then there's a lot of small languages also so what we've seen with AI is it works really really well with the major languages and one other interesting problem in Indian languages is that we all speak mixed so there's a lot of code switching between English and whichever regional language you speak we have in the past done a lot of language training projects also for different AI companies.

23:30So we know how challenging this can be. So if you see a lot of LSIs in India have launched their own projects, their own products, they've launched, there's a lot of great work being done with AI voices for Indian languages, with engines built specifically for Indian languages. And those are better than the generic ones that are available in the market. I would say a lot of the medium and big sized companies have really adapted AI and they've launched a lot of their own products and specialized services, which is great to see for language access. But I also think that it's really skewed the kind of quality you get, especially for the languages outside of the eight main languages.

24:17I would say in some cases, the quality is so bad that I tell clients, you'd rather just not translate it than offer something like this to people to read. So I think if you even see startups, like when we follow startup news, there's a lot of startups which were not traditionally in the language space, which have launched phone agents, dubbing companies. So there's a lot being done. And the quality of these is much, much, much better than it was a few years ago, especially for like Hindi, Marathi, Canada, Tamil, Telugu. It's remarkable to see how incredibly good the quality of the AI has become.

24:55So I think it's a matter of time. But I think the very low-resource languages maybe might not happen. But I think maybe for 15 to 18 languages, it's going to be excellent in a few years. And more than from a business perspective, it's really great because only 10 % of India speaks English. So it's really, really great to see this kind of progress with language technology for regional languages. So it's going to have like a real, real life impact if all of a sudden you have content that used to be, you know, kind of inaccessible, inaccessible to a vast part of the population now being accessible in their, I mean, it's not even a local language, right?

25:40those are big national languages. So, I mean, maybe a little bit on a tangent, but how does that influence kind of culture or, yeah, in India, like that people have access to this, I guess, outside content in their own language? Just there's been so much development in India in the last few years, especially with internet access, right? So the main change has been that most of the population is now online. And the fact that all the apps that launch now for India are available in 10 to 12 languages. It's incredible, one, for them to have access to information in their language, but also to be able to use these AI tools.

26:19I've seen videos where people are using full AI voiceovers in languages, which I wouldn't even imagine. So it's been really great for content creators. It's really good for education. You can see so many videos on YouTube, which are available in regional languages. And I think it's going to be a big part of helping India develop, especially outside the metro cities in the next few years. So this technology that companies are developing is very, very crucial to make, to improve accessibility for even not just entertainment, but for education. And in India, there's like a huge gig economy also where people are doing like deliveries, or they're doing like small things.

27:00And we've gotten requests where they want to translate all the resources, like how to use this app, how to provide the service in regional languages. So it even improves job opportunities for gig workers. Big, big unlock and kind of the frontier here. You know, sometimes, yeah, it's just something where you really see that these latest AI advancements in the language space are making a big, big difference, maybe even much more than in a place like Europe, I guess at this point would. But all right, I want to close on opportunity 2026. You know, we're about to go into another SlaterCon remote where we'll talk a lot about what's going to be, what we think is going to happen in 2026.

27:43But on your side, where do you see opportunities to grow? Are you going to maybe double down on life sciences, medical, and, you know, maybe work on promoting Kavya or any other initiatives that are on your plate? Okay, so I would say even with everything I'm doing, it's very difficult to say what is opportunities to grow. I think the last one and a half years, things moved so fast and we have to see what happens in the next few months in terms of how AI progresses. But what I see, like I mentioned earlier, is the opportunity to not just offer translations, but offer more holistic language services and language consulting to clients.