Learn about Language AI
Website: Language AI | Helpshift
About Language AI:
Enable Global Support with Automatic Machine Translation: Helpshift’s Language AI enables customer support agents to provide multilingual support, letting consumers engage in their native language. Here are the key features:
Accurate & Contextual Translations: Advanced machine learning continuously improves translation accuracy and understands industry-specific terms.
Integrated Workflow: Seamlessly integrates into existing Helpshift platform, no extra tools needed.
Data Security: Strong commitment to data privacy and security compliance.
In this episode, we dive deep into the world of Language AI, a cutting-edge technology that is revolutionizing player experiences.
Our guest for this episode is Dominic Kelly, the director of technology for Language AI at Keywords and Helpshift. With over 20 years of experience in language technology, Dominic brings a wealth of knowledge and expertise to the table. He shares insights on how Language AI can enhance customer support, real-time translation, and the overall gaming experience.
Here are three key takeaways from this episode:
1) Real-time Translation for Seamless Support: Dominic highlights the significant benefits of integrating Language AI into support tools like Helpshift. By enabling agents to provide real-time translation, players from different language backgrounds can receive support in their native language. This not only enhances the customer experience but also improves player engagement and satisfaction.
2) Customer-Centric Game Development: Dominic emphasizes the importance of game developers prioritizing their customers and considering their language preferences. He urges game developers to constantly ask themselves, “What would you want?” and “What would you expect?” By delivering games in the languages players speak, developers can create a more inclusive and immersive gaming experience.
3) Globalized Customer Support: Dominic discusses the power of scaling customer support globally with the help of Language AI. By leveraging this technology, game developers can provide efficient and cost-effective support to players worldwide. The ability to localize, translate, and test games in different languages opens up new opportunities for game developers to expand their player base and reach a global audience.
Overall, this episode sheds light on the incredible potential of Language AI in revolutionizing player experiences. It’s a must-listen for anyone in the gaming industry who wants to stay ahead of the curve and create unforgettable player experiences.
Swell AI Transcript: LanguageAI.mp3
Intro Welcome to the Player Engaged Podcast, where we dive into the biggest challenges, technologies, trends, and best practices for creating unforgettable player experiences. Player Engaged is brought to you as a collaboration between Keyword Studios and Helpshift. Here is your host, Greg Posner.
Greg Posner Hey, everybody. Welcome to the Player Engaged Podcast, and we have another special episode today. Today, we’re going to be focusing on another technology that we have here at Keywords and Helpshift, and I’m joined by My colleague and my friend, Dominic Kelly, he’s the director of technology for this technology called Language AI. And Dom, you want to do a quick introduction of yourself and who you are?
Dominick Kelly Yeah, yeah, sure. Thanks, Greg. Thanks for having me on the podcast. So I’ve been in language technology now for the guts of about 20 years. And I’ve been in the Gibbous family for about three. I focus on everything AI, a lot of it. So machine learning, machine translation, content management systems, also working with other technologies and can integrate them as part of the customer engagement team. So any kind of technologies that touch the customer, I have to get a little bit involved within the globalized service line. So when it comes to localization, translation, testing.
Greg Posner So for people that are listening and may not be familiar with keywords, let’s give a high level of understanding that keywords is broken into three parts. Create does what you think. They create games from scratch, from nothing to creating a full working game, whatever you need help with. We then have the globalized division, which will help translate as well to provide QA services for companies. And then we have the engaged service line, which is actually the customer support side of things. And we can help with community management. So. When it comes to machine translation, that’s falling under our globalized because it’s helping translate your game as well as engaged because it’s helping agents translate real-time text into their native language. So you can help support multiple languages, even if you only know one language. So my question to Dom is an interesting one is, why should companies, when you’re building a game, be interested in localizing a game? Why spend money there? Why focusing on providing different languages if you’re building your game?
Dominick Kelly I want to flip the question for you, right? Because these, these are always very simple. Podcasts always give you the questions. It’s sometimes good to flip them. So the real question shouldn’t be, why are you doing it? Is it what your customers do or your players do when they can’t play the game in the language that they use every day? That’s a question to ask. So what do they do? Are they going to games generally that are in their languages first? more than likely. If they have any issues or challenges or problems with it, what are they going to do? They’re going to dump it for something that’s better or something that’s local for them. So when you want to compete, you have to compete with the understanding that players have so much choice today. There’s more games produced today than anywhere, right? There’s games on your mobile phone, there’s games everywhere. So if they’re not easy to use, easy to access, and they make the player feel that they’re part of the community, then they’re not going to play them. And you’re trying to retain those players. You’re trying to retain those users. And that’s where localization is the most important part. because you want to make people feel like they’re part of that community and translation of languages is the fundamental thing. Because if you don’t speak the language, they ain’t going to stay. And that’s generally what I see. And now, of course, it’s when you’re looking at localization, you should also consider things like, you know, okay, well, how much money you’re going to make out of that market, not to say that money should be the only reason why, but that is a reason. and you might also want to say well maybe you know that English proficiency in that particular region is quite good so therefore maybe we don’t translate straight away until we see how good the game is because often you know translation localization isn’t cheap it’s a service you have to work on so yeah when you’re thinking about it I would always say think about the player first think about the experience you want to give them and then work backwards so that means then looking at the cost for it so you know that for example that your main focus for a particular title might be France and for that reason you should definitely translate in French And then you might want to add more languages in the future, but that’s how I would take it. Everywhere your player is playing the game, you should be able to find the language. And if you don’t, you should have a really good reason for why, because they’ll ask you.
Greg Posner I love that point. And you know, it’s, it’s interesting, right? Because there is a ton of competition out there these days, right? You’re playing a match three game. And just because you’re the biggest match three game in town, if I’m not immersed in the game, if I’m not feeling it, if you’re not providing the support or tools in my local language, I’m just going to download another copy of the game. And it’s easy for me to say, because I’m an English speaking player, and most games are available in English, but you bring up the point of monetary value. And you don’t necessarily, you do want to think about it. We can say we don’t want to, but you do, right? And the English market may not be the most profitable market for a game. There’s other markets out there that are going to bring your bring you more money than the English speaking crowd. So is that something you talk about with customers when you’re there, about what markets might be the best markets for you to go into? Or is that something that they’re primarily making a decision on their own?
Dominick Kelly Well, we do talk to them about that. It depends on what kind of level of relationship you have with the customer, right? Because some of them, they just fire and forget. They give it to us and they ask us to do a service and we provide the service. But that’s fine. And in other cases that we provide technology as well. So the AI stuff in other cases, we provide services and technology. So we do. Where we’re at, we’re at the partnership level, where we’re doing all that kind of stuff. The conversations are more at that level where we’re saying to them, okay, so you’re launching your game, you know, across all these different languages, and these different markets. And then we talked about culturalization. So is your game applicable for that market? Have you considered that you can’t use blood here or the cultural aspects? If you’re using Nordic gods in your game, for example, is that okay to do that? things out in the UAE, maybe not, because there’s some gods that conflict with their religious beliefs. So you have to have these kind of conversations about, you know, ethically and culturally, is that game suitable for launching it and why you should translate it and maybe you should change things for it. So in the source content before you actually And then when you localize, understanding the impact and making sure you always have a story for anything that might be controversial or funny or whatever it is, because every time you try to be funny, there’s always someone who stands around and says, that’s not funny. I’m offended. So you have, because of that, you have to make sure you have a story, right? That this is why we did it. And these are the reasons why, if you don’t have a story for why you’ve got a certain cultural, um, let’s say adaptation or whatever it is, and you don’t have a story, then you’re generally going to offend a hell of a lot more people. So definitely have your stories and have your logic in place. So we do speak about that. We also speak about then, for example, when a game won’t be running for five or six years, and they’ve localized it into four or five languages, it’s doing quite well, but they’ve noticed that, I don’t know, let’s say the Vietnamese market, there’s a small amount of players playing it, they might say well do you know what maybe to generate revenue or maybe to you know increase our user base out there we haven’t done anything yet maybe they’ve got some dlcs they’re working on that could be quite interesting for that region so therefore localize at that point and that gives you access to a new market a new set of players you can expand your your player base and that region too so there’s things like that we talk about as well where a game has got to its peak and now it’s slowly tapering off and now they’re trying to figure out other ways to get some of those players back into the game or attract new players from different markets. We see a lot of those conversations, but that’s always at the partnership level, not on the fire and forget guys, because they just, they give it to us and say, off you go.
Greg Posner Those are your FIFAs, right? Your FIFAs, your Maddens, your games that are just, hey, we just needed another language. We know what we’re doing here. But you bring up a good point a little earlier is, well, not a good point, but one I want to build upon is you said a little later in the game life cycle when, hey, this game is starting to tamper down. Maybe we want to try and hit up a new market. from the perspective of building a game as an engineer, is it easier for me to do it before the game launches because it’s lightweight, it might be a little smaller, or does it not make a difference to the stage my game’s in? It’s the same process to implement something like this.
Dominick Kelly Well, when you’re localizing, you do have different factors, right? So like, for example, font sizes and, you know, cultural things and stuff like that, that you might have to change in terms of your source content, whether it’s acceptable. So yes, you do need to consider localization and translation in the beginning. For example, the worst thing you don’t want to find out is that someone’s hard-coded a load of strings in your game, and when you go to localize it, then you translate everything into Arabic, and then there’s 20% of it that’s left in English, and you’re like, oh, what happened here? And that only happens at the very end, and because it’s happened at the very end, then your delay to launching that game in that market becomes quite big. So the idea is to shift things left in the process where you can, so where you can detect things early, where you can detect things sooner, where you can start to gather the information about what you’re doing, the quicker you can start to deploy fixes so that at the end or the right of the process as well as we like to say, where time is tighter and things are tougher to get done and budgets often tighter too, you don’t end up putting the pressure there. So try to shift things left. So when you’re looking at localization, when you’re looking at testing, when you’re looking at QA, when you’re looking at all this kind of stuff for globalization of a game, you should always try to get it involved as early as possible, embed the people in your teams if you can do that, and get that stuff sooner rather than later. So yeah, if you can do it. One particular thing is like Japanese, for example, is a double white language. So you know if you wonder why sometimes a game might run a little slower because well it takes twice the amount of memory for Japanese and silly things like that that you know like you might not even consider. Or for example, that your game maybe all the text that it doesn’t display right to left correctly. So therefore now you really shouldn’t be doing Arabic because you need to re-engineer maybe part of your game engine. So there might be things in terms of localization that will highlight limitations or problems you might have in the actual game itself. So you might have what you call like lock engineering issues that you need to fix that are for the developer to fix but for the lock engineers to help them fix them if that makes sense.
Greg Posner I think what’s fascinating to me there is that I never really put thought into, hey, different languages need different types of memory because of how much processing power it’s going to need to compute. That’s fascinating to me. It’s something I would have never really even thought about. And it’s something that I guess people should be aware of because you don’t want your game to be clunky just because it’s running in a different language. You still want that smooth experience. So how do you make sure that happens? And I know it’s just a factor I never really thought about would play a big role in this.
Dominick Kelly Yeah, you’ve got to think about it. All the things you can think about, you know, like if you’re launching a game and you know, for example, you’ve only tested it, for example, on stream and you haven’t tested it then on any of the applications that might be out in Asia, and that’s a market you’re going to go after, then you’re going to lead yourself up to a bit of a failure there, right? You know, that it might not perform as you expect it to, so you might not get the users that you expect it to. So always try to put yourself in the players’ shoes, always try to consider the cultural aspects, the performance aspects, of playing that game in that region. So language is only part of it. Localization is only part of it. Testing is key to all of it as well. So making sure that everything runs correctly. So that’s, I think, a big difference with keywords and the life that we live here is that we are all pretty much working in games 24 hours a day, seven days a week. So we understand what it means to build a game and launch a game. And like you said, with the co-dev and the create guys, they’re creating stuff and they’re co-deving stuff and they’re doing artwork and we hear the stories of stuff before. Well, we don’t hear the stories because they can’t tell us, but after they’ve launched and we start to see the stuff being released, we start to go, okay, well, that was part of us. You know, we did that and that’s really quite cool. So, you know, because we live and breathe it from beginning to end, we see a lot of the good things and we see a lot of the bad things. And, you know, when we own that partnership level with customers, it’s good to be able to say, look, you know, we’ve learned these things, we can share them with you. And it might help your game be more successful if, you know, if you’re running on a tight budget to launch a game, or if you’ve got a certain amount of investment and you need to make a return investment for it quickly, then you need to make sure you spend your money in the right places and, you know, You want to make sure the players get what they need as well.
Greg Posner I have a two-part question here, and you have to excuse my ignorance for understanding how engineering and development work. If I am building a game, right, and I want to translate it and I build it in English and I want to translate it into French, boom, I’ve signed up for Language AI. What is the lift to actually setting it up from my perspective? And then maybe a year down the line, I say, you know what, I want this game to be in Finnish too, and Dutch as well. Is it because I’ve already done it from English to French? Is it a lot easier to then just say, Hey, add this language, add this language, or are we talking about a whole new engagement?
Dominick Kelly you’re going to get a load of things you’ve learned from one language to another language, so like we have languages that are in the same family, like romance family and so on and so forth, where a lot of the things are going to be similar, you’re going to run into similar problems. Then you’re going to have problems which are more for, you know, the right to left languages and so on. So there’s quite a few families of languages that once you’ve done one, it does help you understand what to do for the other ones. That’s true. In terms of language AI and the technology and the solution, so fundamentally, you need data, right? So if you want to use machine translation, artificial intelligence, you need data. So that means people will always be needed from the get go in the beginning. If you don’t have any data, then you need to build the data. If you don’t have any data already built, then you need the humans to create that data. So they might be, for example, translating the first maybe versions of the game until we have enough data to build and customize an engine that is good enough for you to be able to use. And what we always do is we use a thing called TER, which is called Translation Error Rate. So this is basically where we look at human translation, we look at machine translation, and we compare the two and say, What editing would you have to do on this machine translation to get it to this human translation? That’s your productivity. When that productivity tipping point from the engine goes, whoop, OK, so now it’s really, really good. The engine makes sense. That means you start using it, and that means you get productivity gains. You can do things faster. You can produce more translations quickly, and so on and so forth. So you have to do this on a language-by-language level. So every language is a learning curve. So you should be expecting to spend more money in the beginning to collect the data. And then as the game evolves, and you may be doing, you know, you’re halfway through and you have enough data to build an engine, you might see some of those costs kind of come down and also some of the speeds increase, because you’ve got the data. If you’re part of a franchise, And you’re doing one game and another game and another game. Oh, Language Eye loves that because you have huge amounts of data to use. It will gobble it up, and you can do some really great stuff. And often, you can create engines for particular content types, like one just specialized for French in-game content, one just specialized for your marketing content, one just specialized for dialogue, because all that content is very different. So you have differences on a language level to build an AI. And then in a language, you’ll have content types within that language. So you have mainly marketing, in-game, and dialogue are the three ones that you tend to see. So yeah, a lot of customization, a lot of data manipulation, and trying to get the best out of them. But we tend to use the terse core to manage the productivity, because it’s the easiest one to be able to say, is what we’re doing making a difference? And is it actually having a positive impact on it? Because you shouldn’t use AI for the sake of using AI. You should always make sure that it is productive.
Greg Posner That’s a key thing. And that’s going to lead us into questions and not quite yet about the rise of generative AI and that. But before I go there, So just again, for understanding how the tool works, right? If I’m playing Marvel Strike Force, right? And there’s a saying like, hey, here comes Spider-Man, right? Is it translated Real-time like is it then like if I’m playing in Spanish, is it gonna translate it real-time? Or is it already programmed into the game for all the languages needed?
Dominick Kelly That’s a good question. So generally, that’s already translated. So that’s already translated in the past. So usually, you know, when you build a game, you’ve built it, you’ve created all your assets. It’s in something maybe like Git, Perforce’s version, something where you have like a repo with all your assets in it. That content then will come out. It will have a branch for when you’re building it and launching it, and it’ll have a branch for maybe when you’re localizing it, maybe two branches for when you localize assets. And that stuff will get pushed out to keywords for localization. If we have an engine already there, great. If we have human translation assets in a database, that’s great, because then what we do is we’ll take whatever you send us, leverage what we’ve already translated, apply the machine translation on top of it, and then have our human process step in where we post-edit that using our professional linguists who know the game, who’ve played the game, or are familiar with the game. and they post edit it and then it goes to a reviewer and that person is also like an expert you might say like a language expert for that particular IP and they then review it and then it goes back so what you end up having is still people in the process but the localization part tends to be a lot tighter because you can now provide all the content translated, whether that be partly machine translated with AI and also leveraged from the work you’ve already translated in the past with humans. So we try to leverage the both to maximize the speed. And there’s no point in sending stuff to a machine if the human’s already done it as well.
Greg Posner So here’s where I want to try and trap you and get you in the corner here. Why is this machine learning? I mean, text has been translated for years, I can go to Google, I could type in translate this from Spanish to English or translate, like, what’s the machine learning aspect of all this?
Dominick Kelly Yeah, so Google is what they call like, you know, miles, miles, miles wide, but only inches deep. So it has a lot of data, but not much wisdom. And as the wise man would say, you know, to turning information into wisdom is tricky. And generally people, you know, have that ability. So AI doesn’t really have the ability to do that. You know, at the moment, what AI tends to do is it will go out there, grab whatever it can, and then try to formulate a response it thinks you want. Now, that doesn’t tell you what you need to know all the time, it tells you what it thinks you want. Right. And that’s the hard part, the disinformation, misinformation, half AI at the minute, right? This is the challenge with it. So you have to make sure when you customise and train an engine, you feed it the right data, you feed it the right information, you get it to understand the context of what kind of thing it’s tackling. So when you say machine translation, yes, okay, there’s an AI model inside of it that’s learning the language and the nuances of what you do, you feed it. But it’s not, as I say, artificial intelligence, as in it’s automatically understanding what you’re saying to it and then creating a response based on true understanding of what you’ve asked it and giving it back to you. It’s not doing that in real time. There is a lot of machine translations that are learning. But yeah, you have to give it context. It’s really, really important. So you always still need some kind of human in the loop. Unless you’ve trained the engine very, very well, and it understands the context. Like, for example, in player support, right? So in player support, you have very small, short, fast turnaround times that you need something instant that will work. So you need to make sure you feed it a lot of small chats. You need to feed it a lot of that kind of dialogue for it to get an understanding of how humans interact. And it’s normally quick. It’s not 500, like Elder Scrolls, 9 million words, for example. Massive amounts of content with huge amounts of context. And you can’t really feed an engine into that and expect it to get the context 99% of the time. Or to the point where the player isn’t going to spot all the problems and the context challenges that the machine will not spot. And therefore, that’s why you definitely need humans in the loop. So where it comes learning, it’s the data you feed it and the cleaning process and giving it to the engine. That’s the machine learning part. So yeah, you know, AI is overused a lot. There is an AI model which learns, but It’s machine learning is the process really. There’s very few true artificial intelligence technologies out there that actually are thinking for themselves. And if there are, please send me some because I need a few because instead of hiring people, I can just use AI.
Greg Posner Boom, it’s a lot easier. I just get AI to create. I can’t just type that in somewhere else.
Dominick Kelly They’re not there yet. If they were, I’d be using them. They’re not there yet. You still need people. But people in combination with machines that are intelligent or give perceived intelligence That’s where the future is right now. Many people now are using technologies to augment or to aid what they’re doing, to help what they’re doing, to remove some of the menial tasks that they’re doing. That’s really kind of where AI and machine learning is kind of stepping in now. You know, where, for example, before people were right, you know, there was always someone designated to take notes. Now, most things are dictated, they’re recorded, and you have your notes created for you automatically. Might not be perfect, but at least you have 65% of the donkey work done. So now someone only spends 15 minutes collating an hour and a half meeting notes. This kind of thing is useful, right? But for an AI to truly understand a game, no, that’s not possible yet. There’s AI that we’re working with Mighty, which is an acquisition that we got back in the beginning, end of last year, sorry. Well, basically what we’re doing is the AI really, or the bots should we say, they play the game. So they have to understand all the events that go on in the game, all the player actions that go on in the game, and they play the game. And what that means is that not having to sit through every single screen in order to work out where text is and stuff like that, this bot now goes through and does this for you. And it starts to triage information. It won’t tell you what the bug is or what the problem is and stuff like that, but it will help identify the things that the humans now need to take intelligent action on. So that’s the key part, that the AI and the machines and the automation is really augmenting the workload for the humans and helping us basically focus on the parts where we add value, all that will pose the biggest risk to what we’re trying to achieve. So in other words, this case is making sure the game plays well. And then, of course, when we have a bot doing that, that’s collecting a huge amount of data. So now we have all the screenshots. Now we have exactly where the events happened. Now we have that. We can feed that into localization. Now what happens is when our translators work and they have the machine translation, they have the image, they have the context of where it belongs, now they know how to translate it correctly. Whereas before, if you were to ask a publisher, where does this string belong in this game? They were like, we think it’s here. They’re not sure themselves, because like I said, it could be a massive amount of data. So having this information that we can provide in terms of the automation feeds it through. So this whole machine learning process, I think it’s more about machines are there to design, that machines should be designed to give context to humans to help us make smart decisions. I don’t think AI should be making decisions for us. That’s not where I want to be, because I prefer having my own two hands and my own understanding of
Greg Posner And I don’t need robot AI surgeons. I mean, I trust that maybe more than the human hand, who knows. But there’s a few things you mentioned that I kind of want to build upon, which is interesting to me. The first one being that the machine learning then is really just taking in the context of the game, understanding what’s actually happening in the game so it could provide the right type of translation. Not just one-to-one, as you mentioned. Google is great, but it’s very shallow. It’s no S and Vs. It’s not personal like that, where the model is different. You also interestingly brought up Mighty. And Mighty, for again, listeners may not know, they’re a QA platform that’s part of Keywords. It does this really cool automated game testing, no more grandma’s boy kind of game testing, if you’re familiar with that. But you mentioned earlier that sometimes you’ll translate Arabic and there’ll still be some English characters that exist throughout the game. Is that something then you can automate with your QA testing tools to, hey, look for any errors that might exist within the- Exactly.
Dominick Kelly That’s exactly it. So if you imagine, let’s say you have a medium-sized game, 50,000 streams. That’s a medium-sized game. So if you have 50,000 streams, that could be thousands of different UIs. Maybe it’s only one UI, which is great. It doesn’t have to be a big UI. But generally, it’s going to be spread across a lot of different parts of the game. And you then have to have someone sit there and go through and test it all. Now, that’s absolutely fine. If it’s in English, you probably won’t notice they’re hard-coded. Do you know why? Because you haven’t had to do anything to the text. The second you have to localize it and you have to edit that text, you start to go, oh, that’s been hard-coded. Oh, that’s here, that’s there. So this is where localization sometimes, as I mentioned, it can kind of highlight the things that you may have missed when you’ve been putting the game together in terms of how the text is structured. So that’s why, you know, it’s great that Mighty does this because you can go through and it can indicate a lot of stuff. It can say, you know, this has actually been translated, you know, correctly. Also, the Proofbot technology, which is the part of technology I’m really interested in, is that what it does is it actually tells you this has been translated and it’s actually French. So it can tell you that it’s French and it can say the text in the image is the same as the text in the game. So in other words, the text we’ve localized and the text that’s in the game is the same. So it can give confidence score, which is really useful. It’s not going to tell you it’s right. It’s going to tell you it’s in context, but it’s going to give you some kind of confidence score. So once again, giving the context and triaging the data for a human to make a smart decision on the actual stuff. That’s the beauty of it. And that proof bot bit is a part of the test bot. So they have build bot, which builds the game, makes sure the game is not broken. They have test bot, which plays the game, so takes the role of the player, goes through and plays the whole game. And then they have proof bot, which is kind of like a weird guy that sits in the background and just takes pictures of everything. and he’s collecting all that data and that’s the bit that I’m like for an AI person The more data you collect earlier on in the stage, understanding what all those artificial players are doing, that should be great for understanding what real players will do. That would be great for understanding how we can pass things through for localization. That’s great for understanding generally how the game will play. And if you ever wanted to realize as well, at that point, you can say, what would happen if we had 10,000 players in the game right now playing this part? well, chuck a load of pops in it, and it’ll give you the idea. So, asking those kind of questions and trying to figure out what performance might be, you know, when you’re doing your English or French or German versions, this kind of technology is really useful. So, yeah, it’s relatively new to the group, but we’re all quite happy about it. And, you know, I think also working at how we can feed this into other parts and other technologies in the group, you know, is part of my job as well. So, that’s going to be my life, I think, for the next year or two.
Greg Posner So let’s go to that because this is where it gets exciting for me personally, right? And I, as being a sales engineer for the Helpshift side of things, we, for listeners who aren’t familiar with Helpshift, we provide a platform that allows to provide customer service so agents can communicate on our platform directly with your players or your users. And where it used to be, you’d manually have to copy and paste text if you want to translate it to a different language. We’ve now taken this language AI model and built it into our dashboard, allowing you as an agent to translate any inbound message to whatever language you speak. So if you have a bunch of English speaking agents, you can have a global game that players can speak whatever native language they want. And keeping with that player experience, making sure the player feels immersed, we can translate the text for the agent to be Spanish to English, and then the agent can reply in English and translate it to Spanish. All in real time. It’s super cool. It’s a great way. And English is one thing, but if you want to, say, put your support in Manila or somewhere else that maybe is a little cheaper, they can all speak English. What was the decision to integrate that into Helpshift? I mean, it seems like common sense. So you might just say, yeah, Greg, it’s common sense. But like, it makes a lot of sense. And it’s something at Helpshift we wanted for years. And it’s not until we became part of the Keywords family that this happened. But how did this get called out, I guess, for lack of better words?
Dominick Kelly Well, we were, you know, we were working before I helped you became part of the Gibbons family. Myself and a few of the group in AI guys were like, you know, we need to do this kind of thing. We need to integrate into support because where are our customers interacting with their players? Where are our clients interacting with their players the most? And customer engagement was one of them. And one of the biggest real, like, things that upset me was there were so many players playing the game that had problems and they don’t want, they don’t even want to respond when they have a problem, because they know that the response they’re going to get is going to be in English, and they can’t direct their answers effectively, or they can’t have effective communication. And what that basically means is that they play the game for a little while, if it’s anyway buggy at all, And that’s the challenge where you don’t have that kind of player support in multiple languages. And it kind of frustrated me, because I have a lot of Polish friends and Russian friends, and I know all around the world, right? And every language that you’re speaking, sometimes you get good support and bad support. And the majority of the thing, being Irish and being able to speak English is great, because generally we get supported all the time for everything, everywhere. which is fine, right? But that was the main reason why I got called out, you know, that we need to be doing it. And in companies that deploy a solution like this, if you’re doing a follow-the-sun model that players have been playing anywhere all the time, that means you need to have a facility geographically placed to have people, and then you have to make sure that those geographic places have the right resources because it’s very hard to find I don’t know, French and Korean linguists or, you know, people that can speak their language combinations in France. It’s very hard to find German and Ukrainian speakers in the Philippines and so on and so forth. So it’s like, where do you put these people? That means you might have to have several people, you have to have people working remotely and on site and so on and so forth. So it becomes all complicated, right? And the huge amount of money investment basically means that people didn’t bother and that’s where you leave that situation in the beginning we end up with that don’t provide the support. So the idea is simply use the machine to augment the human again, right? So you have the agent, they can sit there, they can be anywhere in the world. And rather than also have like agents that are sitting there, for example, they handle French and they only handle French. So that means you have agents sitting there doing nothing until the French issues start coming in. and then they’ve got working, and then all of a sudden you’ve got loads of Mexican guys who are working over there doing Latin American, Spanish, and they don’t get any tickets right now, so they’re doing nothing. So all of a sudden you have like these agents that they go to really high peak times and times where there’s no tickets, right? So then you have to, you know, kind of watch and look about staffing. So when you take language out of that mix, what does that do? That immediately goes, right, how many tickets, regardless of language, do I need to process? now and at peak times. How many agents do I need to cover that amount of issues? There you go. Now you just basically streamline your thinking of exactly how you need to support your customers, because language doesn’t become a barrier. If you can put the AI there, it removes language as the barrier completely. And then you start really thinking about how you staff. And then also those people, those are the people who are doing the tickets all the time, and their knowledge retention, and the ability for them to understand the game. They stick around longer. They will be doing more hours, so they’re probably going to get paid more as well. So this is all good things. And then when you plug that into something like Helpshift, which is such a fantastic tool for working inside applications and providing support for customers that they don’t have to go outside to a different application to start doing requests, they start to get that feeling that the player actually gets the feeling that they are actually being supported. Um, and that means they’re engaged and that means they play more and they don’t have to wait, you know, like five hours for a response. They can get a response in five seconds because you’ve got the agents there. You can do the language anytime, anywhere. And, and that’s really the fantastic bit about it. So it shifted the thinking, you know, and then like you said, uh, Greg as well, you know, then you start thinking about, well, where do we want to have our agents? You know, is it Manila? Is it. wherever you know you can pick wherever you want to put them and doesn’t mean you don’t have native speaking linguists in certain regions, of course you can, you can still do that. You still have agents in certain markets where you see that the revenue is so big that you want to provide, you know, local agents in those, of course you can, but now you can think about, you know, a dual approach where you can do both. You can send it all to Manila if you want to, or whatever it might be, but you can think about where you want to do it. You don’t, you’re not kind of siloed into, into doing something a certain way simply because you don’t have the resources at hand. So yeah, it was a real good, it was a real good idea. It made common sense for everyone. And, uh, you know, the amount of savings that we made in massive for some customers, but I think really the shift. In actually being able to scale customer support globally was the big thing. You know, it’s really, really massive thing when you think about it.
Greg Posner There’s a few notes I took while you were talking through there, and I’d love to kind of just rerun it back here. And first of all, beautiful sales pitch for Helpshift. It was one of my finer ones there. So thank you for that. Hopefully now I’m going to wave my hands like this, but hopefully we could show you an overlay of what what language AI looks like in the Helpshift dashboard. You can see, and hopefully my video editors find a way to edit this so it’s not just me doing this, but we can see it. So if you’re following our YouTube, you’ll see this. But a few things you said there that were very interesting. One is you basically talked about staffing. You have the ability to staff for global support without needing someone all over the globe. You can provide 24-hour customer service at a lower cost because of the ability to have everyone speak whatever language the tool is set to. I think that’s fascinating. You talked about retention, and retention is an interesting one. We often hear quite a bit that one of the highest countries for revenue for players is Japan. If you’re not speaking Japanese, the Japanese market’s not going to give you time of day. So being able to translate your text as well as your agents. I mean, people don’t think about support as being a part of the game, but we recently heard a story, which was interesting, of a German game where all the users were coming to support and talking in English, even though they were German users. And what they heard from their customers after a few interviews were that The whole experience of the game was actually an English game. It was kind of like a mythological thing and players wanted to continue that lore and customer support. So they were just continuing speaking in English and it confused them on why. But it’s interesting that support is part of the experience and you want to make sure that they feel like it’s part of that experience. And from a monetary value, I mean, again, from a savings, from a cost perspective of being able to choose where your agents are located, how often are they working? What hours are they working? Again, I said monetary value and I already talked about my Japan example. There seems to be a lot of benefit for having these stuff in help shift or in the support tool so that the agent can do this real time translation. And that was a lot better, but I think a lot of this made sense that I seem to sum that up properly.
Dominick Kelly Yeah, no, that, that makes sense to me too. And I hope the listeners, um, you know, uh, if they want to see or hear it, you know, reach out to you about it, but I would, every time they build a game, I would just say, please just think about your customer. Just think about the player. Think about the experience to have and ask yourself every single time, what would you want? And what would you expect? And whatever you expect in the language that you speak, that’s what you should deliver. That’s the fundamental thing you should always do when you build a game.
Greg Posner Have you heard feedback from the users on Help Shift who are using the tool and their experience with it so far?
Dominick Kelly Yeah, yeah, you know, security has been quite, you know, from the actual client point of view, because when you’re going in and out different technology, you’ve got to come pay stuff and you’ve got to get agents to do it and you’re putting it in maybe your whatever kind of tool you might be using, and therefore that might not be secure, and you won’t be taking data out of places. And that’s not really what’s happening now. You know, you need to make sure that you have your data and your client data and your client communication in lockdown. And so they were really happy that the process was secure, that there was no potential leaks for data out of the player engagement or the customer support functions now in businesses. And also, you know, like that that whole like relief of okay well we don’t need to worry about the game being localized now into russian or into spanish or into whatever language right that you might want to be able to do because you can provide support no matter what so it means that the player engagement part of the process now can react and it’s not the thing that reacts at the end. It’s the thing that just says, well, okay, we’re launching in the new market. Great. What’s the expected uptake? to be and we can make sure we have the agents in place to do it and that’s as simple as that and you can imagine that the stress is removed because player engagement itself is stressful you know like you have a lot of people invest a lot of money in their games as well you know they have to they have to feel they have to feel like they’re being looked after and have to feel like they have that care and you know the agents have to deal with a lot of stress too so they have to be professional They can’t just be someone you get in and pull out and leave. They have to be able to deal with sometimes young children who’ve invested money or maybe even high swing players, the VIP players who’ve got like 50, 60 grand invested in the game. Those kind of players expect a certain level of experience. The feedback we’ve been getting is that AI is working really, really well. Don’t have any challenges with it. Love the fact that it’s opened up this way of thinking. But it hasn’t changed the fact that they can still go back to doing human customer support where they need to as well. We’re hearing really good feedback. They’re also asking us, where else can we use AI now? So because it works here, they say, OK, well, can we now use this in our game? We say, yeah, OK, you can. And we’re kind of doing it anyway. So here is where we can use machine translation or AI here. Because you can leverage an engine, for example, you might have started Create for Player Support. And then you can maybe use that baseline engine to build something on top of it to build online for a game. So you can do that as well. So there is ways to take what you’ve learned from the Engage experience and use it in other areas too. So it’s kind of what’s operationalizing your technology that we have at play.
Greg Posner I think my one main takeaway I took from that, the most priority one to me was that it’s all two things, right? It’s all secure built into the platform, right? You’re no longer copying and pasting and putting it into a chat GPT or a Google, right? This is customer data. It’s sensitive data. Why are you going to throw it onto third-party tools? It may not seem like it’s risky, but there is an inherent risk when you’re copying data out of a platform into another tool that’s public-based. The second one is that it’s still human-based, right? Yeah, we call it AI and it’s machine learning, and there might be me translating English text into Russian because I don’t know it, but it’s still me as an agent. You’re not working with a machine. You’re not working with a robot on the other end. So you’re still getting that human support despite maybe me not knowing that language. And I think at the end of the day, there’s a comfort knowing I’m still working with someone, even though they may not know my language. And I think that’s an important perspective for players that need to know that. I think those are most of my questions. I have some couple of normal podcast questions, which for people that don’t listen to the podcast, I normally send a list of questions that we may talk about. Dominic, I’m just putting on the spot here because he’s a friend. First one’s easy one. What did you have for breakfast?
Dominick Kelly Oh, what did I have for breakfast? That was just a cup of coffee and a banana, to be honest with you. There you go. Yeah, very sensible, very sensible. Not the full Irish that I would like to have, but the wife has me on a bit of a diet. So, you know, I have to look after myself. I turned 41 last week. Congratulations. Now I need to… Now it’s bananas and coffee, make sure everything runs properly. Keep an eye on the cholesterol and stuff like that. So, yeah.
Greg Posner Cool. What did you want to be when you were growing up? I’m sure you didn’t dream of being a machine language, machine learning translation tool creator.
Dominick Kelly I want to be a games programmer. I’m a qualified software engineer. I did computer gaming, commercial programming, you know, in university. I created my own little physics engine, which works 70-85% of the time. And I thought I was great at it, and I tried to get a job in Cork and different places with different gaming companies, and did a bit of work there. But I ended up working for a translation management technology company, writing parsers for things like FrameMaker and file formats, basically, so I could get content out of it and put it back in after it’s translated. And that was a long time ago now.
Greg Posner About 40 years now. So you wanted to create games. What game was the first game you remember playing or kind of turned you on to gaming?
Dominick Kelly Commodore 64 Kung Fu. Yeah, you don’t even know what that is. You’re too young.
Greg Posner I know the Commodore 64. The game itself, I am not familiar with the game.
Dominick Kelly Go on to the Commodore 64. There’s an emulator, right? Kung Fu. It is the best and worst game ever. Because don’t forget, this is back in the day where literally the game graphics weren’t great and you totally relied on just smashing buttons and doing stuff like that really. And that was like one of my favorite, favorite ever games that I ever played. And then, you know, of course, then, you know, everyone knows that when you start playing games, probably you end up, you know, with your Super Nintendo and your Mega Drive and what have you. So, yeah, but that was my first one. Commodore 64 Kung Fu was the best.
Greg Posner Sweet. That’s a great answer. I love it. I often hear that a lot of people in the industry don’t really enjoy gaming much anymore. Are you still a gamer? And if so, what game are you playing now?
Dominick Kelly I’ve got some games that aren’t trusted favorites that I go back to, which I just shout out to my 2K people. Civilization. I still play Civ IV. I go back to it. I don’t know why. The clicking, the clicking, the clicking. It’s addictive, but yeah, I’ve always liked that so I play Civ 4, Civ 5, Civ 6 and stuff like that, but I play those games. Always go back to Civ 4, don’t know why, just do. Starfield, I’ve been playing that as well, that’s been quite cool, you know, love that game. Fallout 76 I was playing before, but now Starfield has kind of taken over my life a little bit with it. Always go back to, you know, other kind of standard games as well, you know, like Football Manager and things like that, you know, with my son that we play all the time because we always like to think we’re smarter than each other when it comes to picking players. And FIFA and stuff like that. But yeah, the main one probably is Starfield at the minute and that’s taken up my life. But there are a few others on my hit list that people are telling me I should take a look at. But yeah, the Starfield one is really the one at the minute.
Greg Posner Here’s a fun story about me. When I was in college, I went to college in Baltimore, Maryland, and there was a project where I had to go meet a creator of some sorts that inspired me. And I realized that Firaxis was also located near Baltimore, Maryland, and Firaxis makes the Civilization games. So I reached out to them. and I’m going to get his name wrong and I apologize, but I went there and I got to meet Sid Meier and he walked me around the studio and he opened up this closet full of games and said, hey, you can take a game or two if you want. I was just like, oh my God, that was like my dream moment. I got to meet a video game creator that is the prime of his game, making some of the best games of civilization that were ever invented. And it was just this starstruck moment for me that really solidified that this is something I want to do for a living. And it was just such a great experience.
Dominick Kelly Well, I am very jealous. I’m very jealous. I would have loved to have done that myself. It’s great when you get to see the people that are actually involved with the game and the creators of the game and stuff like that. Very jealous, very jealous.
Greg Posner What game did you pick, by the way? It was called Pirates. You got to sail around a ship. It was similar to Civilization. I was never really great at turn-based strategy. I was more of a command-and-conquer type of player in real time. But I respect the game. To this day, Gandhi still crushes me whenever I decide to try and play the game, and it’s just… it is what it is.
Dominick Kelly Yeah, Yuri’s Revenge is always the command-and-conquer one. That’s a great game. Great game.
Greg Posner One of my first online games at an Unreal Tournament. I just remember playing that game online. Just amazing. Dominic, I’m so glad we got to do this. I’m glad we got to talk about the technology. I really think it’s cool stuff. I’m glad we got to hear your story about it. I’m glad to hear that you’re fixing your diet with coffee and bananas. But I think it’s great stuff. I think we’re going to have a lot of information about language AI on our Player Engaged website, on our Keywords website, on our HelpShift website. We’ll be able to share this and I appreciate your time. Is there anything else you’d like to just share or plug while you got the time here?
Dominick Kelly Well, I suppose everyone’s going to, if you are going to the Japanese game show or you’ve been at the Canadian game shows recently, you know, always reach out and say hello, ask if one of us are there, it’s always good to say hello. I love to meet new people and discover new technologies, stuff like that. So if I’m ever in an event, feel free, do not feel afraid to come up and say, hi, have you seen this? Because that’s the stuff I love. So sharing and getting to know new technologies is always a great thing for me. So that’s one thing I would say. Don’t be ever afraid to come up and say hello and share something with me, because that’s basically why I live my life.
Greg Posner Great. Appreciate that, Dominick. And unfortunately we had another potential guest joining us that knew how to sing Frank Sinatra. Fortunately, he’s not here and Dominic cannot sing Frank Sinatra. So we’re going to, we’re going to cut that from the clip here, but, uh, I appreciate everything. I hope we can do this again and talk more about it. I hope everyone that’s listening has a great day and I look forward to talking to you guys again.