With Andy Andersen, Growth Product Manager at Brave

Below is an automated transcript of this episode

Antoine Rey (Host): 0:18

Hi, everyone. My name is Antoine Rey, and I’ll be your host uh today for this uh episode of Global Ambitions. And today my guest is Andy Andersen, who’s a gross product manager at Brave. And Andy, welcome to the program.

Andy Andersen (Guest): 0:35

Hey, thanks for having me. It’s great to great to chat.

Antoine Rey: 0:38

Where does uh this podcast find you, Andy?

Andy Andersen: 0:41

Currently in New York City, where we’re welcoming the winter that seems to be coming upon us now.

Antoine Rey: 0:48

Very good already. Okay, so today, yeah, guess what, everyone? We’re gonna talk a little bit about AI, but we have a little bit of a different take on it. We’re gonna take some specific aspects of it on when you’re dealing with execs, when you’re looking at SEO, which is very different again, and in relation to change management. So diving straight in, some of the comments I heard over the months or so at a different events, Andy, was that you know, everybody seemed to have some stage selling promises of AI and almost have figured it out, you know.

And then this year, somehow in the last few weeks, I’m hearing a more cautionary approach, you know, saying that some things are working very well, others not so well. The main question remains though, how do you address this with your execs and your stakeholders? Because the expectations, I guess, from your point of view are still there, from those execs are still there, and they want to see some form of a time cost reductions and some benefits out of this. Can you maybe expand a little bit on this, what you’re seeing at Brave?

Andy Andersen: 1:59

Yeah, well, it’s it’s a really interesting question because I think you know, when it comes to new technologies and new advancements in the industry, we have to take into consideration, everything from past history to now in terms of where things have come and where the industry’s grown. And you know, when we think about how buzzwords come into play, like for example, right now we’re talking a lot about AI. There’s machine translation at some point. And the reality is right now we’re dealing with a situation where they have this new technology, it’s not really a new technology, let me correct that. But it’s it’s it’s something that everyone wants to talk about, everyone wants to implement into their workflows.

But the reality is there was a lot of hype for situations that it was a magic solution to every situation that people in the product and and space really work on. For the example, of this is going to translate your entire product, it’s gonna do everything you need, all of that. But what we’re actually seeing is that, first of all, it’s not available to do something for every single situation, in the sense that it’s not the AI capabilities are not perfect in all scenarios. And I think the the use cases there when it comes to like when you break it down into all the different areas of that that someone might need a product to be you have to consider that there are situations in which you need curation, you need a human in the loop, and it’s it’s just not perfect. And then the second thing is language-wise, we’re seeing that LLMs are not equal for all languages.

So when you think about French or Spanish or German, they might have a decent amount of online content available for them to train models for, but something like Hindi or Thai or you know, I’m not sure exactly what what the breakdown would be, but the amount of content that they would have, the the bodies of information from which they can train LLMs for is much uh different. And the quality of content is uh is also not consistent. And so you end up with poor quality content that’s not good enough for what you need. And so you end up with this situation where basically there’s a disconnect between what was promised and I think where we’re at. And I think navigating that in this day and time with with execs is interesting because there’s this promise of of all these things, and now there’s a there’s a disconnect that we have to fill the gap for. And I think the first thing that I would say to people that I try to work on is like, what are we actually solving for? When you ask that, what I mean is specifically, is there a budgetary concern when it comes to localization, or is it a time technical concern, or is it merely just the idea of staying on top of new technology?

So I would say if it’s budgetary concern, you need to find ways to implement AI into the workflow in the way that you can do large volumes of content through AI without having to worry about doing every single string from a human. So you can kind of reduce costs that way. If it’s an efficiency thing, then figure out the workflows there, which you know we could get into the nuances of, but figuring out the workflows to save time, go to market time. And then when it when it comes to anything else, if it’s new technology or whatever, then making sure that we are trying to implement things, but also showing that where it works and where it doesn’t, and just keeping up with that. So I guess it really depends on who you’re talking with and knowing your audience.

Antoine Rey: 5:18

And do you feel like in general execs are are like receptive to those explanations? Like I get a feeling that they were expecting the magic button once again this time.

Andy Andersen: 5:29

Yeah, I think you know, execs in general at all companies are are always focused on different areas in the sense that they want to see high-level information, not so much in the details of does it work, does it, does it not? I think we as an industry need to do a better job of doing case studies or showing for and after or giving high-level details in the sense of like where there’s efficiencies, where there’s where there’s opportunities to improve, and not just kind of blanket statement that this is going to fix everything. I’ve spoken with a few people from around the industry, and I think it’s not necessarily consistent. Some really are pushing AI as the solution to everything, and then other companies want it implemented into their workflows. I’ve heard a few cases, not many, but a few cases where executives aren’t pushing at all for AI, which is which is also interesting.

So I think it’s situational in that regard. But I think if you’re an executive and you think you can save costs somewhere, I think it’s always appealing because it’s like we can improve our margins or something like that, especially in the startup space or in the publicly traded space where margins are are everything. But all in all, I think not to go too tangentially here, but I think one last point that I want to make is that I think it’s also important to keep in mind the actual cost of localization versus you know this idea that it’s such a huge cost compared to other things. When companies spend millions of dollars on marketing budget or something, it’s localization into 40 languages or something really that expensive compared to the value that you get from that. So I think that’s a that’s a conversation that we don’t have enough as well, is that people focus a lot on can we save a few dollars doing translation work, localization work? But in the end, is that really that big of a cost compared to the brand campaign or the ad spend or something like that that companies tend to have?

Antoine Rey: 7:15

And that remains for us to convince them the holy grail a little bit. And like, are you being encouraged internally? Because I’ve seen that at a few clients, are you being encouraged to build your own model rather than using LSP provided or TMS provided models?

Andy Andersen: 7:29

Not right now, I would say. I think there’s always that idea that you could train your own models and have company-specific content. But I think in general, it’s it’s not something that’s pushed, mainly because if you really get into the technical side of that, if you’re trying to do that yourself, it can become very costly and very labor intensive. If you’re going through third parties, there are a few third party companies out there that will help you train LLMs and and you know work something if you can have enough content. But I think the reality is that most companies don’t actually have enough content to build their own models. I think they would have to rely on bigger corpuses of information.

And so I think the data that is available to most companies, I mean, if you just have a few a few hundred strings or something for your product, it’s unlikely to be a great enough amount of data to basically train a model. So I think you can definitely encourage models to lean towards your content, you know, with glossaries with contextual information. And so you can give LLMs, you can feed them, you know, more information. But I think in terms of fully training LLMs from the ground up, it’s not something that you know we’re focused on at the moment.

Antoine Rey: 8:34

And if we focus maybe on some specific aspects for using AI services, we started to see some interesting success with international content creation or multimedia, for instance, with text-to-speech. You and I discussed that, you know, around the SEO side of things. Have you seen some major shift in how we approach SEO in general using AI?

Andy Andersen: 8:59

Yeah, so SEO is a funny space in the sense that I would say, although it’s changed a lot over the years, the fundamentals, in my opinion, haven’t really changed. Fundamentally, if you’re a good website, you have good content, you’re still going to be if your domain authority is good online, you’re going to be discovered online in the sense that you’ll rank for type type different types of content, et cetera. And so I don’t think fundamentally SEO has changed all that much, but I do think the nuances of SEO have continued to change and evolve over time. And I think what we’re seeing now is that as search behavior is evolving, if you visit Brave, you can see Ask Brave as like an evolution or Brave Leo have this, you know, these models you can you can chat with. But across the industry, if you’re just doing generic searches, someone might be using an answer engine like Chat GPT or anything along those lines. It might be using Google’s Gemini. You know, they’re different areas where people are using.

And so what’s happening is that we’re seeing a mix of old behaviors where people just go to traditional search engines like Brave Search or something, and they’ll search what to do in New York this week or what to do in Dublin this week, or something like that. And they might, you know, they’ll get some results and whatever. But then there’s also the people who are going to LLMs and they’re saying, build me an itinerary for this, or even just asking other generic questions specifically, like what’s the best private browser, what’s the best search engine, anything related to us, for example. And when they’re doing that, then it becomes more about are you optimized for answer engines or for LLMs or anything like that? And so that’s when it becomes very interesting too, because now you have to worry about not only your content, but also how your brand is portrayed to the internet. So, for example, your brand reputation is now becoming really important in the sense that if you do an answer engine and you go and you ask a question about what are the top five browsers that I can download or the top browsers for privacy or something like that, and you say, give me a pros and cons list, and you’re looking at the answers are there, it’s pulling from Reddit, it’s pulling from Wikipedia, it’s pulling from many different sources. And so the the reality is that you’re not in control of all of that content anymore.

And so you have to actually think, how are we managing our brand reputation online? Because even in a situation where a customer complains in Reddit or something like that, you still need to be there responding, providing up-to-date information. You need to provide context, you need to say, you know, something because an answer engine will take that into account. If someone’s complaining on Reddit or something like that, and they say your product did something before, like it was broken or something like that, there was a bug, or whatever. If you come back and you answer that and you provide additional context and links or or whatever, then in theory, the answer engine should say, oh, a bug was reported, however, it looks like the company addressed it, or something like that. And so these are these are things that we all have to keep in mind. I think it’s really across the industry that this is gonna take place. And so answer engine optimization, some people would call it generative engine optimization or LLM optimization, whatever, they’re all basically the same to me. But I think you have to think about this in the context that this isn’t only an English thing. This is now getting multilingual.

So I would kind of not coin a new term, but I would say localization of answer engine optimization, just like we would say SEO localization or something like that. It’s a new thing, it’s a novel idea, but it’s basically something that we have to be considerate of because if you’re worried about your brand in English, you also have to worry about your brand in any other languages that you support as a company. And that’s why it’s quite an interesting evolution in the traditional SEO space.

Antoine Rey: 12:41

Probably has always been the case, but somehow it’s become a lot more apparent in results in search, and when you’re moving away from that QA to a conversational model, I guess in this case, that makes a huge difference.

Andy Andersen: 12:52

Right. And it has always been there, right? I mean, if we go back five years or something and someone types best browsers or something like that, you have to worry about which websites you’re showing up on and how you’re being presented and and you know those things. It’s it’s still been there. It’s just that now the game has changed a little bit. And so you have to be even more concerned with not only where you are, but how people are talking about you, how you are participating in that conversation, and and that you’re putting out the proper information. So any information that companies don’t put out about themselves officially is open to speculation. And so if you want to release a product or feature for your company, it’s good to have blog posts and press releases that talk about your product and give specific details and information about your product. Because if you don’t do that and the internet finds your product, then they get to talk about your product, and that becomes a definition of your product that’s not officially from you. And so it’s it’s something that companies need to keep in mind.

Antoine Rey: 13:50

And if we talk about change management within your own localization team and organization in general, how do you feel AI is changing and affecting the roles that you have to plan within your team, for instance?

Andy Andersen: 14:04

I think when it comes to the needs of the current time, again, I don’t know if I would say that they’ve fundamentally changed in the idea of like what are we doing? I think it’s still the same, right? We need international product, and how do we get that? We need localization of the product, we need uh customizations by country, we need search engine optimization, app store optimization, all of these things. The fundamentals of the what I don’t think have changed drastically, but the how and the approach to that is is definitely evolving. One area that we’ve definitely focused on as a company, and I’ve encouraged others to do similar is to think about a model where you not only have LSP relationships, but you also have like a language growth manager or something along those lines, which it’s kind of a it’s a position that I came up with like a year or two ago with Brave, but it’s it’s it’s an evolution. It’s a it’s a 2.0 version of a traditional language manager because I think that the idea of a language manager is still needed at many companies.

You need someone who can own and curate your content from the language specific perspective. But in addition to that, it’s important to think about the fact that they’re not only going to be doing just traditional localization anymore, they need to be doing a lot more, they need to be versatile, they need to be adapting to new technologies. We also need prompt writers. You know, you need a linguist who can now write prompts in their own language or at least talking about their own language. So, for example, I know that I’m not qualified to talk about Russian grammar or something like that, or or what to do in different situations. And so I would want a Russian linguist to write a prompt for me that that specifically caters to Russian grammar in terms of what is our formality? Are we using T or V or you know these these kind of situations? Like that’s not a decision that the product owners should necessarily be making in this on a language level. We should have a company strategy, we should have a style, what’s our brand voice and and and tone, but that on a language level is much better decisions made by a specialist.

And so I like to focus on this language growth manager as a person who’s basically the CEO of their language. They’re the ones that look at their language as a business and they say, How can I improve my language business for this company? And so they’re basically looking at different opportunities all the time. They’re looking at how can I not only localize the content, but where can we improve? Can we improve in the keywords? Can we improve the conversion rate for our store? Can we improve in the the paywall in our app, in our mobile application? Can the paywall conversion be improved? I want them looking at different angles of the business, not just in the traditional sense of I have English and I need to put it into French, or I have English and I need to put it into Japanese. Should be more involved in that. And so that’s the way that I’ve chosen to pursue it at this point.

Antoine Rey: 16:59

I think that’s very clever and close to my heart. Like I love to see a role name change there using gross because we’ve seen localization teams changing their names to be perceived more as a gross enabler to that very point, as opposed to just a translation services organization, and they use like words like gross or experience, and now that seems to be taken into the the job descriptions or the role descriptions. In this case, I was actually talking to a recruitment agency earlier on because we’re always in search of those linguistic prompt engineers or someone that has a mix of a technical sort of experience or curiosity, at least on the prompting uh side of things, but also the linguistic experience of a given language. Great. Well, listen, that’s all we have time for today. As always, this is a short podcast. Thanks very much, Andy, for coming in today. It was a pleasure to talk to you, and I’m sure people will find it interesting and can connect with you.

Andy Andersen: 17:55

Thank you. Thank you for the time. It was great to chat.

Andy Andersen

Growth Product Manager

Linkedin

Website