#284 Market Reality Check, RWS-Cohere, Data-for-AI, GlobalComix

0:00Florian:The central bottleneck in AI has shifted from building the base capabilities, kind of those foundation models, training the base capability of the model to now ensuring deployment readiness.

0:14Florian:Hey, we're back, all shiny and new. Hi Esther. Hi Florian. We are reinvented. We're new. We're all shiny, new, new and improved. If you don't know what we're talking about, head to the website slater.com. yeah so that was a busy few weeks wasn't it yeah few weeks few months few months but the sprint the final sprint and the bug hunting on the website kept us busy so but yeah now we're back can do a podcast we'll walk you through the new website and all the good stuff that we've launched in the meantime since we spoke I don't know you and I maybe like 8 weeks ago was the last news podcast We'll walk you through at the end of this podcast because there's a bunch of other things to go through.

1:02Florian:And yeah, it's not that we haven't been busy on the podcasting front. I mean, we had a number of great guest episodes there with Audioshake, where I learned a ton, with RWS, where Ben Fass walked us through the Cahere partnership. And then last week with WeLo Global, the new WeLocalize, WeLo Global. Also Recon. Also Reborn, also have now kind of a holding page for the company. Highly recommend you go back and listen to those. Very, very interesting insights there. With Paul Carr from WeLo Global, I think one segment that I learned a lot was on the data for AI, where we also published a report which we're going to go through.

1:51Florian:So it was very interesting to hear his take on the business and how suitable it is for LSI. So again, highly recommend you go back. Yeah. So again, give you a slightly longer walkthrough, including those on YouTube. Quick visual screen share walkthrough at the end of the podcast on the new website. Really proud. Went really well. Now SlaterCon coming up in, what about four and a half weeks, I think. We're heading to London. Just four and a half weeks. Yes. I guess I'd say five weeks. Five weeks? Tell me it's five weeks and not four and a half. It must be. It must be. Okay, five weeks. Yeah. Five weeks.

2:31Thank you. Thank you. We'll call it five weeks.

2:34Florian:But lineups almost complete. I think there's like a single slot that we're still working on, but the rest is definitely confirmed. Got a great presentation. RWS is coming and explaining the LanguageWeaver Cohere integration. We got the Startup Accelerator, Dubform and Tuner. Startup Accelerator is going to be in a separate stage, so kind of a new format we have here. Then we have an incredible panel on voice AI with Gradium, PolyAI and AudioStack. Super, super interesting. Spoke to the AudioStack CTO, Peter Coyle, when we had our LanguageAI meetup in London than a few weeks ago. Yeah, that's right.

3:18Florian:Yeah. Yeah. Very, very interesting. So I think a lot of us are going to learn a ton. Netflix is going to speak, Alison Anders, Research Accelerator, again, happening in the separate stage, Buy versus Build, Illusion and AI Localization with Phrase and TripAdvisor, and I think I'm going to moderate this panel. Am I? Very nice. And then Data for AI. So confirmed speakers are Sergio Bruccoleri from Appen and Artem Grigorev, CTO of Toluca, a big data for AI firm and I think we're confirming somebody else there as well. Still last second, working on this. Startup accelerator in the afternoon and then we have Akulat giving a presentation as well on how AI is rewriting global content supply chain.

4:10Florian:We got the investor panel with Araya Ventures, Soluna Capital and TwinPath Ventures, which you're going to moderate. Yes. And then you're going to close. And then we're going to hear from John Tinsley, translated VP of AI Solutions at the networking. So we're all set. Tickets are for sale, looking good. Make sure to join us in London. One question. So you mentioned there's a separate stage. Does that mean people have to choose between the main stage and the separate stage? They do not. They do not. So it's during the break. It's during the break. So we keep that one stage formula that we have, right?

4:46Florian:So we have the main stage with that incredible backdrop, that giant screen, the biggest screen ever. And then there is a separate room that seats about 40 to 50 people where this is happening while people are networking. So you can choose to go there during the networking break. Chris, there's some people that may not be the extroverts that may want to go and get more content. So that's what we're doing there. Cool. So we didn't just launch in the mean, since we spoke last time Esther on this in February, we didn't just launch a new website and a data for AI report, but we also just launched the Slater Index.

5:26Very true. Yes. Very exciting. Very true.

5:29Florian:So it's no longer called the Slater Language Service Provider Index, but the Slater Index. It's tough to get away from, I have to say. I'm very tempted always to refer to it by the former name after sort of seven years, seven iterations or seven editions of that particular one. But I like the new name. It's short and obviously it encompasses way more than just sort of language service provider. Absolutely. And we can't really rename or, I don't know. Rebrand. Rebrand the kind of LSI, LTP and then still keep that index as the language service provider index. So we discussed this a long time and just felt, why don't we just call it Slater Index?

6:12Florian:Because a lot of people were referring to it as Slater Index before already anyway. So when's your index coming up? It's like, okay, yeah, it's language, I was writing. Okay, so it's the Slater Index and it now also features a selection of LTPs, Language Technology Platforms, with some key data points. So the reason it's still a relatively modest list of LTPs is because we're actually making them go through a survey. We're not just featuring them in a list with some random data points, but it's actually that there's some revenue data for some, there's some usage data, some MRR data, some monthly active user data, you know, whatever we could get.

6:54Florian:And so it's a very interesting list. It's, you know, basically you've got the LSIs on the one tab and you've got the LTPs on the other tab. So why don't we dig into some of the results high level? I think the amount of revenue we indexed went up again, marginally, right? But that should not be interpreted as evidence of like an underlying market expansion because I think we indexed probably more companies and also there's a fair amount of M &A activity. And I think we wrote that as participation in index is voluntary, results are skewed towards high performing firms, selection bias. So if you're doing poorly, you may stop participating.

7:43Florian:So yeah, there's growth in the revenue index, but I don't think that really reflects underlying market growth of like the core language industry. Also, there's a bunch of M &A, which continues to inflate aggregate revenues.

7:59Florian:Because I think if you look at, let me just pull this up, it says 57 % of companies report declining revenues. Yeah, if you look at the overall number, it's not looking super pretty there on the LSI front. Why don't we jump to the actual index and maybe just go through some selectively there at the top. We have, yeah, Trends Perfect reporting 8.3 % growth. We must have spoken about this on a prior podcast, right? Yeah, we did because I think we did coverage of when they announced their financial results and the whole fact that they're no longer or less than 50 % language services. Correct. So, yeah, they eked out like 8 % growth there with some of the other services they're doing.

8:48Florian:RWS is down 3.9 % in dollars. That's verified. That's publicly filed. Appropriate jumped. They are up 71%, but that's on the back of a ton of M &A, including Suricom. So that's probably the biggest one. Then WeLocalizer, Wheeloglobal is down 2.6%. Then AcoLite's down 9.8%. That's self-reported. And then on the super agency front, who else do we have here? We have Star Group down 5.1%. So overall on the super agency front, it's basically a couple of outliers with Propra and TransPerfect, but mostly on the back of M &A. Yeah, I mean, those two companies are probably the, in 2025 in particular, have been among or the most acquisitive in the industry, right?

9:41Florian:100%. And then did DALS acquire somebody, another company, DALS, the UK-based interpreting company, they're PE backed? I think they did. I wouldn't know at what point exactly, whether it was 25 or earlier. They're up 20%, quite impressive. Then Elan Languages, I'm sure they acquired somebody that they're down 2.8%. Argos Multilingual didn't acquire anybody. I'm pretty sure they didn't and they're up 11%. Congrats to Veronique. So that's good. LanguageWire down 16%. They didn't acquire anybody, I believe in 2025. That must have been before the French kind of tech doc acquisitions they made. And then...

10:30They were quite small though, I believe anyway, the French tech doc thing, right? That would have been in 25, I think.

10:36Florian:At least one or two of them maybe. Oxford Languages. Time moves so fast. You're like, was that 25 or 24? Been doing this too long. Been doing this too long. Oxford Languages. Our friends there at Oxford Languages, Casper Grothwell, we had them at a remote, I think, once. They're up 27%. He was on the pod as well, a long time ago. You're right. He was on the pod also. Yeah. Yeah. Up 27%. So business is going well for them, you know, on the back of Data4AI. We'll speak about that soon because high quality data is very much in demand. And if you're Oxford language, the one thing you have is high quality data.

11:18Florian:So going further down, I think we have AI media verified there because they're public, they're down 2%. Marginal. Linguava interpreters up 19%. Yeah, the two sort of US-based interpreting, right, still seems at least in pockets to be doing quite well. Hannah interpreting also up 24%, congrats. What else do we have? If you go down, T-Works, the collection of companies in the DAG region down 12%. On the flip side, Terra up 17%. The translation people up 37%. Interpreters Unlimited down 12%. And then closing up with AvantPage who were on the podcast as well, up 8%. Yeah, so head over to the list.

12:12Florian:It's freely accessible if you want to get the entire list as an Excel or a Google sheet or whatever, you can, I think we made it so you can actually buy that one. But yeah, otherwise, I mean, generally freely accessible. Great, congrats to the team putting this out mid-April. I know it's a ton of work to get 300, 400 companies to reply to that survey and getting it out so early. So congrats to the team. Great. Hey, I want to just briefly go and revisit the RWS Cohere partnership. Of course, if you want to hear from the CEO himself, go and check out the podcast with Ben Fass. But I just want to share some of my thoughts, kind of the TLR there.

12:55Florian:I still think this is actually a big win for RWS. and it's interesting how the market just did not react to it at all. I think maybe they're, again, no investment advice, but I mean, I think they're underestimating this a bit. It is a win-win. I mean, Cahir really gets to showcase a major kind of enterprise adoption because LanguageWeaver, right, it's a core, I mean, RWS is a language solutions integrator and LanguageWeaver is a core engine of it with AI translation. And now they're saying, well, we take, basically we have the car around that we keep with LanguageWeaver, but let's switch out the engine and we're putting in the Cohere engine, right?

13:43Florian:Okay. That's a pretty big decision. But if you're the engine manufacturer, that's also a pretty big win because you're getting a lot of distribution through RWS, right? So generally, I feel this is such an interesting partnership and I'm really looking forward to seeing what the kind of the impact will be for RWS, but also for Cahir and also wonder like what this means for D-Bell, right? I mean, this is quite a competitive move by these two companies and D-Bell's obviously the engine and now they're building a car around it. So yeah, to belabor the metaphor further. Anyway, I think it's a super interesting partnership.

14:30Florian:And also, Cohere really is doubling down on that multilingual. And even people that used to be ex-Cohere are doubling down on multilingual. This morning, I learned that a company called Adaption was founded by Sarah Hooker, who was the lead of the multilingual, of part of the multilingual team at Cahere and they started a new company. I don't think we covered that. I think it was a$50 million funding. We're trying to dig into this a bit more. But basically, yeah, there's all this kind of activity around Cahere also happening. And they'll be talking, well, at least RWS will be speaking about the partnership in some detail at SlateCon.

15:10Florian:Absolutely. Yeah, I think we're still waiting for the person that's going to come in from Cahir, but I'm sure they're bringing somebody on together with the head of Language Revers. Very, very exciting there. We want to briefly head to Data for AI. So we did publish that amazing report, touched on it in the past couple of podcasts with Ben and Paul, but you really have to get that. I mean, if you're an LSI and you're looking for growth, if you're already in that space, This is an absolutely, it's kind of a seminal report also for us, right? I mean, we're now starting to cover this segment much more aggressively in addition to voice, data for AI.

15:54Florian:I think it's roughly a$10 billion market. What was the exact figure there? 9.3, I think was the number we put on that. Yeah,$9.3 billion currently, and it's expected to grow very strongly. So to like over$20 billion by 2030. That's obviously an educated guess, but you're talking about a major growth market here that is somewhat open to LSIs. We do, of course, caution that there's a bunch of things that LSIs need to have in place to be successful there. Also, Paul from Wheelock Global added his take in last week's podcast about what's needed. So it's not a straight line from an LSI to a data provider, but their LSIs have, I think Paul said about 50 % of what they need to be successful in that.

16:46It's not a bad starting point, I mean, frankly.

16:48Florian:Exactly. Compared to another outsider. And, you know, you've seen a number of LSIs kind of pivot completely and others haven't. But like, it's definitely something that you need to look at. Yeah. I mean, I've had that as a, let's say, talking to people in the market where I'm saying, okay, what are you open for? What are you looking for in terms of maybe acquisitions? And it's come back like data for AI, data support services, that really this is an area that people are kind of, some people are prioritizing from a strategic perspective. So when you talk to them on the advisory side, they're saying they're not looking at acquisitions in that space?

17:28Yeah, exactly. I mean, on the one hand, looking potentially for acquisitions on that space, but also themselves sort of trying to double down within that space. So, you know, double-sided.

17:38Florian:And it really shifted in the report, we're laying out three phases, kind of pre-2020 and 2020 to 2022, and then from 2022, you know, with the chat GPT moment. So I think really the central bottleneck in AI has shifted from building the base capabilities, kind of those foundation models, training the base capability of the model to now ensuring deployment readiness. That is not jargon. That just means these models, if you want to deploy them at your company, they need very specific data to really work, right? They work up to a point, the base model is, but then you need to add specific additional data and that's where the data for AI industry is coming in.

18:29Florian:So the industry is increasingly focused on data that makes these AI systems reliable and safe and useful in real world operational environments. So capability data was the past or this still is needed of course, but deployment data is now where demand is really accelerated. And there's different sets there, adaptation, alignment, adversarial, and evaluation datasets. This is all essential for tailoring general foundation models for specific tasks, behaviors, and industries. And this also meant that the kind of workforce shifted a lot, right? In the 2018, 19 kind of Appen glory days, like it was low cost general annotators.

19:11Florian:and now you need much more specialized expert knowledge labor. That's very different, right? Where's that coming from? I mean, is it people that have come from other professions that are sort of training up somehow within the data-free market or is it sort of graduate of programs that might be somewhat suited to this? Great question. I think it's, yeah, let's dig into the report even more. So we're going to schedule a podcast with our head of research. who wrote that report to answer some of these questions. You should join and ask her questions. I will. Yeah. So we also found that Frontier AI Labs are shifting their sourcing from kind of the saturated public web data towards licensed and proprietary data.

19:56Florian:You know, in the past, you just scoured the internet and I guess that one's done. Every Reddit thread has been ingested into some model already. Yeah, look, we could go on here, but I just do recommend that people go and get that report. We're scheduling another podcast specifically on that report. We're trying to get people on the podcast from the space. We're hosting the panel at SlaterCon remote with people who are running big companies that are active in this industry. So yeah, we're doing our part to make sure that people learn about this and the people that are already in this space get additional market intel on it.

20:36Florian:it. Why don't we head over to M &A? I think there's been a couple of fundings and like one somewhat complex buyout there in Italy. Yeah, there've been a few things happening since we spoke and I think this is not all of it, but some of the sort of transactions that we've chosen to highlight. So start with the complicated or most complicated one first, I suppose it would be the buyout of Star 7 by a private equity fund, Argos, with the idea being that the fund is going to acquire sort of 75 % of the company, which is currently public, somewhat public. Some of its shares are public. So there's a whole sort of level of complexity there.

21:22And I think specific to sort of the Italian market.

21:26Florian:So the 75 % are owned by private investors, like by Dante and then the Star, the Swiss based company, the LSI. Yeah. And then they have this kind of tiny free float or like the quarter of the shares, 25 % is a free float on the Milan stock exchange. Okay. Huh. Yeah. I mean, it's not easy. It's kind of not easy to parse in terms of, you know, what exactly is happening from a transaction perspective for somebody coming in from the outside like myself. But yeah, I mean, I think just generally we could say we can look at some of the reports that come out. So in terms of the offer price, I think we had that at 11.9, 11.90 euros per share, which seems to be a bit of a premium over the six month average price.

22:19But then again, only of those 25 % of shares. So, you know, not quite sure what to make of that in particular. Overall, so I think the 2025 revenue was down and they have faced some performance pressure. So revenue down about 3%, nothing crazy, but, you know, still some pressure, while net profit also down 14%. So, yeah, I mean, it'd be interesting to see what the fund does with it once this all, you know, kind of goes through the necessary clearances. And, yeah, I think initially there'll be the 75 % owner, but with the objective of at some point probably taking over the 25 % shares also.

23:08Florian:But it's been a while since we've seen a fund, kind of a PE fund or another buyout fund, by something related to the language industry. So yeah, maybe they're coming back now. It's interesting. I mean, they are firmly within localization, but I mean, they say they've seen a decline within the localization, which was primarily its core business and somewhat of a shift towards what they call experience and product knowledge services. And I think they are emphasizing, you know, somewhat naturally the strategy around the integration of AI and developing this knowledge as a service approach. So let's see what the fund decides to execute on.

23:54Florian:That's what we're doing. We're doing a podcast, which is knowledge as a service. It's free. You can listen to it on. Knowledge as a free service. Knowledge as a free service. All right. There is another interesting one that is also from a world I'm not familiar with manga localization distribution. Manga. Manga, yeah. You said you like this one. Yeah. I like this one. Why do you like this one so much? Yeah. It brings back memories like Akira days back in like the 90s and now my kids are reading all these manga comics. I have absolutely, I mean on a personal level just absolutely no knowledge of manga or comics or anything.

24:38Florian:They're so popular. I mean they're incredibly popular. Yeah, I know. I think I'm, well, that's what I'm saying. Not the right generation, maybe. Yeah, so there was a New York-based company called Global Comics that raised$13 million in funding, first of all, and then went on to acquire a company called INKR, based out of Singapore, which is the manga localization and distribution platform. So therein lies the manga component. I think the aim here is for Global Comics to go deeper into the content localization side also with the aim of helping publishers to release the manga or the translated manga editions kind of closer to the original publication date and I mean something I think is pretty interesting within this space is what they were talking about as sort of the piracy element so the fact that a lot of manga is not actually sort of officially translated but you have these homegrown fans that are probably trying to localize and translate as much as possible.