#282 RWS CEO Ben Faes on Why They Partnered with Cohere

0:00Ben Faes:Behind this LLM partnership, there's a fundamental shift in how we think about localization.

0:11Florian Faes:Hey everyone, and welcome to SlatorPod. Today on the podcast, we welcome Ben Faes. No relation. So Ben is the CEO of Language Solutions Integrated and Super Agency, RWS. Hi Ben, and thanks so much for joining.

0:24Ben Faes:Hi Florian, thank you for having me.

0:26Florian Faes:Yeah, it's the first time I have a Faes on the podcast, so that's good. That's a premiere. So there's a lot to talk about today, Ben. I want to talk about status as a public company, the Cahir announcement, language we were pro-launch, repositioning over the past couple of years of RWS, software and AI, and of course, a ton more. Now I want to start with a hard question, Ben. What are the markets and investors missing about RWS?

0:49Ben Faes:I don't want to fall into the trap of, I think, every CEO of a public listed company in the UK, especially at the moment. Everybody thinks their company is undervalued. So, you know, it's not really for me to say if it's wrong or right. What I can tell you is what the market thinks. The market thinks that we are operating in a market that's disappearing. And I think you're very well placed to have an opinion on that element. I think fundamentally the market doesn't understand the role of super agency or LSP, LSIs. The second thing which is more linked to us is I think we have had some disappointing times in the past where we missed on our guidance.

1:47Ben Faes:And the market is quite cruel. If you lose their trust it takes a lot of time to rebuild it. So I've been focusing really on that. The first guidance that I gave, we hit it. And I think fundamentally where I would say the market is wrong is, but that's the only way that they work, is they are pricing the past, they are pricing what they think it used to be. And the fundamental difference is it's hard to know where we're going. So I've given some optimistic, I think, but quite reasonable guidance for the future of RWS in terms of revenue growth and profitability growth. And we are really focused at that.

2:32Ben Faes:It's something that you need to learn. You cannot spend your days watching at the stock price. There's no reason why the stock moves in one day or not. We haven't released any specific news. And so as CEO and as management team, we really need to focus on the long term and making that company great for the future.

2:55Florian Faes:So you're originally from outside the industry. I want to touch on that. Now, what do you tell people that don't know the language industry about what RWS does? What's your extended elevator pitch, I guess, for the large company you're running?

3:10Ben Faes:It's funny you ask that because I think I've been struggling a little bit at the beginning, you know, at dinner tables to say, what is it that we're doing? And so it really helped for me to spend a lot of time internally and to understand what we were doing, which is quite complex and quite diverse. So the way I explain it now is we're providing technology solutions to enterprise to help them operate confidently in any market. And we do that, we help them protect their assets, their IP in any market. We help them engage, communicate with our customers confidently in any market, in any language.

3:52Ben Faes:And we help them also understand the environment better. So I think we are a positive force to help companies expand their operations abroad.

4:06Florian Faes:Maybe for the people who don't know you, maybe tell them a bit more about your experience in the past. I think you were with Google, YouTube, AOL, Concentrix. Tell us a bit more about that. And also, how did that prepare or not for the role you have at RWS today?

4:22Ben Faes:So, not going to do my own CV, but I did start my career at L 'Oreal in marketing, which taught me to look at the detail. You know, when you work on the luxury product, every detail counts. And I must say I'm a bit of a maniac sometimes, a slacker for going the extra mile, making sure that every detail is right. And I moved to the more internet world in early 2000. One of my best friends was managing Netscape in Europe at the time. And he told me, you know, your beauty creams, it's great, but there's a revolution out there. you need to join and so I found this role at AOL and my role was to build the whole advertising infrastructure of AOL in France.

5:16Ben Faes:Fantastic, fantastic time. Six times after I arrived there was the biggest internet bubble and you reconstruct from there and you rebuild and you find new solutions. I joined the last three years I was CEO of AOL in France which also was a story of reinvention. AOL was initially an ISP so an internet service provider in France and we sold the ISP part so we only stayed as a internet portal but I don't know if you remember at the time AOL was operating in a private platform. So it wasn't on the internet, it was a private protocol. So, you know, being an open platform when you don't even have a proper website was quite a challenge.

6:12Ben Faes:And so we redesigned, reinvented the whole platform and launched all kinds of new services. and I was poached by Google in 2008, roughly a few months after they bought YouTube and the mission was to launch YouTube in Europe. Launch in terms of, you know, they had spent what seemed an enormous amount of money at the time,$1.6 billion, I remember, and there was barely any advertising on the platform. So what do you do with this? so I joined YouTube which really took some convincing there used to be a video platform at the time in France called Dailymotion who was I think 10 times bigger than YouTube and the first meetings I had with the media agencies I was laughed out of the room you know keep your dogs on skateboard for yourself we are serious people here but then again you know Perseverance, redesigning the product, coming up with new ad formats and famously the skippable ads that we invented at the time.

7:26Ben Faes:Made this platform now I believe the largest media company in the world in terms of revenue on audiovisual content. so I think what I really learned is that you need to keep being aware of what's going on keep listening to what users are doing consumers clients keep reinventing yourself and I was lucky actually at Google to to change job almost every three years so I touch on the the media side, the partnership side. And the last three, last five years almost, I spent at Google Cloud, again at the early days of the cloud explosions, and trying to convince company that actually their content was much safer in the cloud than on-prem.

8:22Ben Faes:And all this now is almost common sense, but you need to start with really strong conviction and start selling and engaging with clients. What my cloud experience did is to understand the complexity also of the infrastructure behind to make things happen smoothly. Arguably, I had worked for almost 15 years on the ad side and it's glitzy and it's fun to go to Cannes every year, but the hard work is also on the infrastructure that happened in the background. And I left Google in 21 after 15 beautiful years. I really wanted to have a different experience. And I joined this amazing company called WebHelp, who was the largest customer service outsourcer in Europe.

9:18Ben Faes:And there you learn something completely different. You learn to manage people at scale. I was given my first job. I was in charge of our UK operations, so CEO of the UK division. And, you know, from day one, I had about 20 ,000 people to manage in the UK, in South Africa, in India. And of course, you know, that it's in call center. But when you go and visit those places, you have to really engage and understand how to motivate everybody, how to give a purpose to all those people that work so hard in your company. And that's where, you know, the culture and the purpose comes as so important. It's great when you work in a company and you pay the fortune, people stick to their job.

10:17Ben Faes:But, you know, when it's more linked to what it is that you're doing. And, you know, I've met some fabulous people in all those places that really took pride in helping customers around the world, helping solve a subscription problem, helping solve a technical problem, helping solve their financial issues with their bank. so I guess that's the two sides that I'm using the most at the moment is the technology side and how to think product, how to build solutions and also the human side which is important and probably the most difficult one.

11:00Florian Faes:For those of us that are like maybe 45 plus, how would you compare internet 96 98 to like AI 2026? Just that that'll be weird for anybody below 30 or under 30, but for the older ones, it feels a bit similar-ish sometimes. I have friends that are like calling me like, oh my God, I just, you know, connect the claw and I have these like Mac minis here and I'm so excited. I'm like, the last time I saw this was like 97 with Dreamweaver and like spinning up a website.

11:31Ben Faes:There's one saying that I kept using is innovation is never going to be as slow as it is today. It's kind of a spin, when you think about it, you say, what's that mean? It means that it keeps accelerating. And so today it feels that we're innovating at an incredible pace, but tomorrow is going to be even faster. And I feel that, we all live through the launch of the first iPhone, spinning your first website, as you say. Now, it's been, I've learned to use, I've learned first to use AI deep research, I think back in September. Now, every weekend almost, I'm coding a website, sometimes professionally, sometimes inventing a game with my son.

12:26Ben Faes:It's something that was unable to do before. So the democratization of tasks, the ability for everybody to personally innovate is fantastic. And so I think we're wasting a little bit of the gift that we're being given. And we're spending a lot of it on mundane things. But we are an incredibly lucky generation to have access to so much intelligence and capability at scale. It started for me, it started really also, I mean, you could argue that there's a democratization of media. Everybody was its own media. That's what you're doing here. And so you didn't need to have a broadcaster to have access to millions of people.

13:23Ben Faes:Then you have democratization of the compute. And you can rent a compute unit. So you don't need to have a big IBM supercomputer to start making amazing things. Now we have democratization of intelligence and coding. And so, you know, I think we ought to think a little bit more how to best use it and how to use it in a positive way.

13:47Florian Faes:Let's talk about an announcement you guys made last week. That's why I'm very excited to have you on the podcast today with Cohere, a company that was on this podcast about three years ago and has come a pretty long way since. We had Nick on the podcast like three years ago. So you announced that you're partnering with Cohere for Language Weaver Pro. And I'll give you the floor. Give us the highlights, background, what's Language Weaver Pro, etc.

14:13Ben Faes:Let me touch first on the on Korean partnership. You know, it seems very mundane these days to announce a tech partnership. I think partnership is a word that understates a little bit the relation that we've built with Korea. We started working with the team back in July last year when they came to us because they were about to launch a new model, command A and they asked us basically to test it and to help them make sure that it was a very good model. And from there started really a relation of collaboration with our research team. We have the Language Weaver team and all of their researchers are based in LA.

15:04Ben Faes:and so they worked very closely with the Cohia team to understand and improve their model and we quickly realized that they were on to something very unique very special and so we decided back in September to start working in a more close relationship and to help them train further the models. And we started, I think, drafting the basic of a partnership back in September. I think it took some time for us also to understand and realize, because behind this LLM partnership, there's a fundamental shift in how we think about localization in many instances. You know, I'm still, I'm yet, I don't think it's going to ever happen, but to digest the fact that this industry is based on the price per word, it's done on segment.

16:13Ben Faes:And I think to get the best of this, of those new capabilities and machine, you need to think much broader. The big plus of LLM is their ability to manage much larger context window. And so to understand how they operate, to understand the message, to understand why some of those transitions must happen. So we worked over the last six months plus, and we are now operating a model that we've fine-tuned together, that we've integrated into LanguageWeaver. And so we are using that model as the brain of LanguageWeaver, the translation product. And the tests that we've done are absolutely phenomenal.

17:09Ben Faes:I'm amazed that we are getting to that level of quality in such a little time. And so it's not the end of the road. It's really the beginning for us. And we keep working and fine-tuning models with them. That's why I really call it a partnership. We are selling together. We're fine-tuning together. And it's a fascinating energy that is brought also to our product team to work so closely with a frontier model like Korea.

17:46Florian Faes:Yeah, and it's a very interesting partnership. I mean, we described it in the newsletter last week as like, RWS gets to integrate a very powerful model that must have cost a fortune to build and train, right? That's kind of beyond the reach of any company that's not funded to like, 6 billion valuation or got like a billion dollar funding route. And Cohere in turn gets deployed across hundreds of enterprise clients that took RWS decades to secure and build and have these relationships. So that's very hard to replicate for a tech company, even if you have billions in funding. What are your thoughts on that?

18:20Ben Faes:So absolutely right. you know, I remember back in 21 when I was at WebEup, we were thinking, should we build our own LLM? And there was an opportunity, there was a company that came and say, if you invest 50 million dollars, we can build the bespoke LLM, blah, blah. You know, thank God we didn't do it. And with Cohere, it's a bit the same, you know, that building a foundational model is not within our reach, and there are people who can do this a lot better. I'm going to admit that I was very frightened at the beginning of those discussions with Cohere, because you wonder, am I helping them kill my business in the long term?

19:09Ben Faes:What's the reason why RWS and Cohere our partners in five years? It's basically the exam question. And so I had multiple questions, multiple conversations, especially with the CEO, Aiden, Aiden Gomez. And he summarized it in a very, very simple way. Cohere is not the product. The model is really the brain that powers the model. And so that really ticked it for me. I want to have the most intelligent brain and the product team really focusing on making this the best product possible for customers to manage their translation. And the reality is when the Cohere salespeople meet some customers that want to have translation product, They can use their Cohere subscription, but they need all products to actually manage the actual translation.

20:17Ben Faes:And so you have similarity. You take a product like, I don't know, Lovable, amazing product. They didn't build the LLM that powers Lovable. So I think that's really how I've defined also the differentiation for AWS and for the product team. is to focus on making the best product possible that uses the brain. And the reality is for some customers, if you want to bring your own LLM, happy for you to do so. We also think that we're using our expert, our skill set to fine-tune what is the best LLM to use for that particular use case.

20:57Florian Faes:Give us an example of a couple of ideal customers for LanguageWeaver Pro. I think historically, language weaver is very strong and like kind of very, maybe even on prem and things like that and defense sector and areas like this. But what's the ideal customer for that new product, the pro product?

21:15Ben Faes:It's true. And it's still the case, you know, first of all, it's a pure enterprise product. And that's also one where our minds match with Cohere. They are focusing exclusively on enterprise customer. Probably why they are less known because you don't have a customer B2C interface like ChatGPT or Gemini. But it's purely enterprise. And so for enterprise, for our customers, we have a lot of on-prem deployment for LanguageWeaver. I think it's slightly challenging today to deploy LanguageWeaver Pro on-prem. It requires big GPUs to make it run properly, but we're working on it. You know, what we've seen, there's a spectacular result, for example, for all UGC content.

22:15Ben Faes:UGC content, think of comments that you get. think of product reviews or customer service, customer support tickets this content is quite hard to manage on NMT because there's a lot of misspelling, mistakes invented word, abbreviation and there's a lot of it and so do you want to spend a lot of money translating this it's kind of considered low value by most customers. Language River Pro makes an incredible job at localizing those reviews, those content. And so you've probably seen many, I don't know, Amazon reviews that make no sense because they've been translated and the context doesn't work.

23:08Ben Faes:The result is spectacular because you understand really the meaning. You have people getting very excited in comments and putting yeah! That word doesn't exist. He understands the context and he will translate it by oui! in French. And so that's quite spectacular. You probably have seen the white paper we've published. It was really hard to define the parameter of the test and how to test Cohear. And so the choice we've made is to test it on two environments or two universes, a well-known universe in the localization space, so IT and technology, which needs to be very precise. It's high volume.

24:03Ben Faes:You know, it's translating tech manuals, etc.

24:09Ben Faes:And high-end marketing, which is really hard. Because, you know, that's less world, it's more meaning. And Cohear really works very well on both. So I think it means that we can raise the bar in the type of content that we can turn to machine translation only. And we can really use now more and more those models in more complex environments. So I think we've tested that model with a lot of customers in close beta. And the range of customers that are interested in this model is quite amazing. I'm happy to say we've signed also our first customer last week just after the announcement. So it's a good sign always.

25:01Florian Faes:Good sign, good sign. All right. I want to talk a bit about the new segmentation you started. I'm not sure if you're already reporting in those, in like the filings, but like you have generate, transform and protect, you alluded to it before. So kind of what was the thinking behind, you know, differentiating the business into these three distinct categories and yeah, how has reception been like on just the side of the customers?

25:29Ben Faes:So I think, you know, one of the challenges, I guess two challenges. One is, RWS has grown a lot by external acquisition, the most famous of which is RWS combining with SDL. And the reality is a lot of the operations below were a bit still kept as ex-SDL, ex-RWS, even ex-Moravia. And that's not great to really create a united culture within the group and also to leverage the scale that we have. There's no need to be a super agency if you're just a succession of local agency. I'm exaggerating. So that's the first thing is how do you really get the power of the scale that you've built? And the second thing is investors struggle to understand what we were doing and why we were organized the same way.

26:33Ben Faes:You know, life science was doing also localization, but for different types of clients, is that really the reason why we should separate this? So I've separated into generate, which is helping customers generate better content, either through our publishing tools or through our data services, which is understanding better the world and organizing your data so you can use LLM to create content. Transform, and it's transform rather than localize or translate because, as you know, a lot of the work we can do is transforming from English to English. So you're transforming from a specific audience for a specific context.