Latest / The Retail Podcast / Scaling Global Businesses with Language AI | Bryan Murphy, CEO of Smartling

Transcript

Scaling Global Businesses with Language AI | Bryan Murphy, CEO of Smartling · The Retail Podcast

• 0:00
  So how do we, how do our customers keep up with that
• 0:04
  amount of content cost effectively, right.
• 0:07
  And so the keys to that are automation and also making the,
• 0:11
  the cost of translation more effective.
• 0:13
  And this is where we're using AI agents in effect to help
• 0:17
  automate the the the translation itself and improve the quality
• 0:21
  and measure that quality. Hello and welcome to the retail
• 0:35
  podcast. Now today we're going to go not
• 0:38
  technical, but we're going to look at a technical aspect of
• 0:41
  something that's a big pain for the people that realise that
• 0:44
  it's a pain. And as a, as a, as a, as a
• 0:46
  business owner who's tried to do multiple language across the
• 0:50
  world from a for, for the retail news website.
• 0:53
  It's a problem that I totally understand.
• 0:55
  And I've heard a lot of people talk about translation.
• 0:57
  It shows, but it feels like Brian and we're going to come to
• 1:01
  you. It feels like such a complex
• 1:02
  area or some of the experiences that people have had with
• 1:06
  different tools. Do I just use an A, a Google
• 1:09
  plugin? Do I just Google track?
• 1:10
  You know, it just seems like, and I think there's a lot of
• 1:14
  money left on the table when it comes to conversion.
• 1:17
  And more importantly, the reason I was looking at it was for SEO
• 1:20
  optimization to have my local language website, you know,
• 1:24
  optimised to do this. So I'm joined luckily by Brian
• 1:28
  Murphy, CEO of Smartlink. Brian, why don't you tell us
• 1:31
  what Smartlink does? And you're actually have I got
• 1:35
  Smartlink right? Am I saying it right?
• 1:37
  Yes, Yeah, yeah. No, you're right.
• 1:40
  What what do you take us through, what Smartling is and
• 1:42
  then your journey to to CEO? Sure, of course.
• 1:46
  So Smartling is a a software company that helps companies
• 1:51
  like IBM, Disney, Tesla, Spotify create multilingual experiences
• 1:59
  for their customers around the world is the easiest way to
• 2:02
  describe what we do. And we do that with a
• 2:04
  combination of software and AI and we sort of solve the problem
• 2:09
  of, of, of doing exactly what you just described, which is
• 2:12
  creating these, you know, translated, localised
• 2:15
  multilingual experiences, which traditionally has been slow and
• 2:20
  expensive to do. And that's, that's really the,
• 2:23
  the job that we're here to do is help our, our, our customers
• 2:27
  automate that process, make it seamless and dramatically reduce
• 2:31
  the the cost and turn around time of producing very high
• 2:35
  quality on brand translation. Yeah, I think being a small
• 2:40
  business, I obviously went for the cheap option, which ended up
• 2:43
  getting me in far worse state because I ended up with a whole
• 2:47
  bunch of dead links that were generated for translations that
• 2:51
  were in, in, you know, multiple languages.
• 2:53
  That anyway, it just became so problematic for me that I just
• 2:57
  thought, well, I, I had to retreat.
• 2:59
  I, I gave up trying to solve the problem because the, again, I'm
• 3:04
  a small business. I am not Disney or, or Tesla.
• 3:07
  They're they're dev teams are totally different.
• 3:10
  But I just, one of the things I heard at NRF, which was really
• 3:13
  interesting when one of the CE OS was talking about how he gets
• 3:18
  on brand in different languages. It's actually a lot more complex
• 3:23
  than people take give it credit for to, to translate your
• 3:26
  meaning, your vision in a way that's natural and not computer
• 3:31
  sounding. I'm just curious how, because I
• 3:34
  was looking at your, your, your propensity to translate well is
• 3:38
  really high. So you're actually translating a
• 3:40
  lot of the words accurately. But how do you then move into
• 3:44
  sort of keeping that brand feel? That that's it in a nutshell.
• 3:49
  I mean, I love, I love my translation apps, my plugins,
• 3:54
  Google Translate, for example, or any of these in there on my
• 3:57
  phone. And they, you know, when you're
• 3:59
  in situations when you need to be able to communicate with
• 4:02
  someone who speaks a different language or you just can't
• 4:04
  remember a word, they're super helpful, right?
• 4:06
  They they kind of get the gist of of what is being set.
• 4:10
  But if you're a brand, right, if you're global brand like an IBM
• 4:14
  and you're producing 20 million, thirty, 40 million words of
• 4:17
  content every year around the world, you're not it.
• 4:20
  That's not good enough. Just like anything like it's
• 4:23
  there's professional grade, right?
• 4:24
  Yeah. And that's really what we do.
• 4:27
  And so a couple of things need to happen.
• 4:29
  Number one is number #1 you need automation, right?
• 4:33
  So you have to have this seamless plug in to your what we
• 4:36
  call the brownfield tech stack, your content management
• 4:38
  solutions, your websites, all these things.
• 4:41
  We, we have that integration that we do so allows the
• 4:45
  automation this translation then once, once it's now once it's
• 4:48
  automated, then we have a variety of solutions that
• 4:50
  actually do that translation, ranging from the most expensive
• 4:54
  and the most traditional, which is to use a straight up human
• 4:57
  translator that's specialised in that content type and in that
• 5:01
  language pair, right all the way to neural machine translation.
• 5:06
  And the what we do that makes it this professional, that makes it
• 5:10
  different, that makes it on brand is that we use a
• 5:13
  combination of neural machine translation, AI and when we need
• 5:19
  to human expert. And that's what allows us to
• 5:22
  produce highest quality translation that sounds like,
• 5:27
  you know, IBM or Disney in any language that you speak,
• 5:31
  incorporates the correct gramme, glossary, style, guide, tone of
• 5:36
  voice, etcetera. So in whatever language you are
• 5:40
  translating into, it sounds like IBM or it sounds like Disney or
• 5:44
  it sounds like Tesla and they have very different brand
• 5:46
  voices, right. So it's super important for our
• 5:48
  brands is that it sounds like they're brands.
• 5:50
  And and so do you often integrate into the CMS then
• 5:54
  they're all OK. Yeah, because we interviewed,
• 5:58
  gosh, I, I should remember who we interviewed last week, CMS
• 6:02
  provider of NRF. I'm just not.
• 6:06
  We integrate to just about all of the major.
• 6:08
  All the major CMS providers. Yeah, all the major, so
• 6:11
  contentful, Adobe, you, you name it.
• 6:14
  We we integrate to the to the CMS provider whether content
• 6:19
  stacks sanity, a core story block.
• 6:22
  Coral Yeah. Storyblock.
• 6:24
  Thank you. There we go, I knew what to get.
• 6:26
  As we had, we had the CIO of NRF on, on the on the podcast and he
• 6:32
  was telling us that, you know, obviously when the big show is
• 6:34
  happening, that website has to be translated into 5 or 6 core
• 6:39
  languages. And that and that content lives
• 6:42
  in so many different places, trying to create that user
• 6:45
  experience in one language. But story block were like, well
• 6:49
  that's why you would use story block because we can help you do
• 6:51
  that. That's, that's exactly right.
• 6:52
  So they're they're a good example of a good partner of
• 6:55
  ours. So we have a, we have a plug in
• 6:58
  that enables their customers to seamlessly create localised
• 7:02
  content. We, another good example of that
• 7:05
  from a ecommerce perspective is Shopify.
• 7:07
  We just recently rolled out an integration to Shopify which
• 7:11
  allows their, I think they've got nearly 6,000,000 stores now
• 7:15
  and 75 different countries. And one of the things that we
• 7:19
  found is, is that number one, and this is always a surprise to
• 7:22
  people who speak English, is that the Internet doesn't speak
• 7:25
  English. OK, 70. 75% of Internet users
• 7:29
  don't speak English and 87% won't buy from an English only
• 7:34
  website. Yeah, it sort of makes sense
• 7:37
  when you think about it. So the point is this is that for
• 7:40
  the Shopify merchants out there, stores that need to have that
• 7:45
  have an international business, right and need to have accurate
• 7:49
  and on brand product descriptions, website copy
• 7:53
  etcetera. Sparkling is a is a super easy
• 7:56
  solution to use that is significantly faster and less
• 8:01
  expensive than the traditional way of translation and
• 8:04
  localization. Yeah.
• 8:06
  I mean, when you really think about, you know, e-commerce
• 8:09
  expansion just across Europe, 7, you know, 8 core languages that
• 8:13
  you would need to translate to. And especially if you've already
• 8:16
  come out into Europe. I don't know if you're an
• 8:19
  English speaking company or even a French speaking or a German
• 8:23
  speaking. It's just how, how are you going
• 8:25
  to grow if you don't do this? And it's, it's interesting you
• 8:28
  mentioned 40 million, is that the amount of words that major
• 8:31
  sites there or thereabouts, Is that what what you're seeing in
• 8:34
  terms of how many words have to be covered?
• 8:38
  Oh, can I? I'm just curious.
• 8:40
  Yeah, yeah. I was at 200,000.
• 8:42
  Then I was like, oh, that's really big.
• 8:44
  And then I saw a few million as the next one, but 40 million.
• 8:47
  I was like, wow, that's, that's incredible.
• 8:50
  We have to give you a perspective.
• 8:51
  We have 16 billion words in our database.
• 8:56
  OK, so all. Right.
• 8:57
  It's, it's a good as we, we, you know, we, we translate across,
• 9:00
  we help our customers translate across 450 plus language pairs
• 9:06
  And yeah, we get about and we probably, we probably translate
• 9:09
  about 7 billion words a year for our customers.
• 9:13
  Yeah. So just a few words here and
• 9:15
  there. Yeah.
• 9:15
  So one of the as a, as a tech person, I've, I've had this
• 9:20
  conversation time and time again on premise or in the cloud.
• 9:24
  Where'd you sit with that? Oh, we're, we're, we're in the
• 9:27
  cloud. We, we're cloud native platform
• 9:30
  and that's by far and away the most efficient solution.
• 9:34
  We, we have a continuous development cycle.
• 9:36
  So from a technical perspective, we do 3 / 3300 production
• 9:40
  releases per year. So what is that 11 per day?
• 9:45
  So our customers all are always benefiting from very, very rapid
• 9:50
  innovation. And how do you see on the topic
• 9:53
  of innovation, obviously there's so much, there's so many
• 9:56
  commercial grade AI translation tools or agents that are sort of
• 10:01
  the next thing that everyone's going to agents.
• 10:04
  Where'd you sit in that world of agents and and AI?
• 10:09
  Yeah. So we, that, that's a great
• 10:11
  question. We talk about that a lot.
• 10:12
  We we're, we're constantly if if you look at a road map right
• 10:17
  now, it's, it's loaded up for the next 24 months in doing
• 10:21
  exactly what you're describing there, which is helping automate
• 10:24
  the very, very arduous task of, of translation and and
• 10:27
  localization. And you know, once again, when
• 10:29
  you're doing 7 billion words per year and we think that that
• 10:33
  number is, you know, we see that content on the Internet is going
• 10:37
  to double in the next 5 years, right?
• 10:38
  So everyone knows that there's more and more content.
• 10:40
  So how do we, how do our customers keep up with that
• 10:44
  amount of content cost effectively, right?
• 10:48
  And so the keys to that are automation, which you know, AAI
• 10:52
  agents are, are super helpful there and also making the, the
• 10:57
  cost of translation more effective.
• 10:59
  And this is where we're using AI agents in effect, to help
• 11:03
  automate the, the, the translation itself and improve
• 11:06
  the quality and measure that quality.
• 11:09
  I'll give you a great example. So if you're once again, take
• 11:12
  Spotify, if you take Shopify, right, 175 different countries.
• 11:15
  So you're a brand manager or an e-commerce manager and you're
• 11:19
  producing all of this content right in product descriptions,
• 11:22
  etcetera. And you needed to be in 1520
• 11:24
  different languages. Well, most people don't speak 15
• 11:27
  or 20 languages. So how do you know that what you
• 11:31
  just wrote is on brand and accurate, right?
• 11:35
  In the past, you would have relied on a human translator to
• 11:38
  review all that and give you the thumbs up.
• 11:40
  Problem with that is it's super slow and it's very expensive.
• 11:43
  So obviously the right answer is like, well, let's get AI to do
• 11:47
  it. OK, Well that's really what
• 11:49
  we're in the business of doing is combining the, the SAS
• 11:53
  platform that we have machine learning AI and that human to
• 11:58
  create this really optimised process for, for translating
• 12:01
  content that has also a quality assurance aspect of it to it.
• 12:06
  So right, so now you can feel confident I can produce all of
• 12:09
  this content and I know that smart link's quality check that.
• 12:12
  In fact, we actually guarantee quality as a result of that.
• 12:16
  So you can produce, you can take all the content, produce it in
• 12:19
  20 different languages, we'll run it through the system super
• 12:22
  fast, super cost effectively, and then guarantee that content.
• 12:26
  In terms of the Internet not speaking English from an SEO
• 12:29
  perspective, how, how do you support your customers in
• 12:34
  optimising across multiple countries and optimising their
• 12:39
  SEO so they're ranking high specifically when when they are
• 12:43
  in, you know, retail, for example?
• 12:46
  Sure. One of the things that I learned
• 12:48
  it eBay is that digital footprint matters, right?
• 12:51
  So we would look around the, you know, and it kind of makes sense
• 12:55
  for, you know, the way Google thinks, right?
• 12:56
  So Google looks at like how big is this website?
• 12:59
  How much relevant content do they have and how good is the
• 13:02
  traffic are people, do people go to that site and find what they
• 13:05
  need, right? And all that sort of like builds
• 13:08
  this concept of digital footprint.
• 13:09
  The bigger it is, the more Google thinks you're relevant,
• 13:13
  which you want for, you know, authenticity and for rankings
• 13:18
  and all these kind of things. The mistake a lot of companies
• 13:20
  make is that they're like, OK, great, I'm big in the US, but
• 13:23
  what does your site look like in Germany or France, Japan and
• 13:26
  these other sites? It's probably teeny tiny.
• 13:28
  So Google thinks it's teeny tiny, right?
• 13:30
  And there's always a local competitor there.
• 13:32
  Like in Germany, that's like huge, right?
• 13:34
  That's their backyard. So Google prioritises that.
• 13:37
  So digital footprint matters. And this is where we really to
• 13:41
  help our customers because it's expensive to take all that
• 13:45
  content and localise it. But we're able to do that in a
• 13:47
  super cost effective way that generates very high ROI for them
• 13:53
  because now we're able to take all of that content that they
• 13:54
  have right in the US, let's say, take it over and localise it for
• 14:00
  any of the markets. And Google loves that.
• 14:02
  Like they're like, oh, this is great.
• 14:03
  This is a huge website, people come to it, they get what they
• 14:07
  need, it's localised. So it's a very, very cost
• 14:10
  effective way to expand SEO globally.
• 14:14
  And that's 100% top of mind because the last thing you want
• 14:17
  is to to not rank in, you know, whatever the top three of of
• 14:21
  your industry. If you're a substantial company,
• 14:24
  having a translation element doesn't slow the site down,
• 14:27
  right? Because that's the other thing
• 14:28
  that people are worried about. Is this going to slow my page
• 14:32
  refresh right? Yes, that, that's right.
• 14:35
  Well, speed is we thought about two things.
• 14:37
  We thought about a lot of things, but speed was the other
• 14:39
  one. And so we think about that a lot
• 14:41
  too. So it's really critical that we
• 14:42
  don't impact or negatively impact performance which.
• 14:46
  We don't. Brian, thank you so much for
• 14:47
  giving up your time. I'm conscious of time.
• 14:50
  I want to be respectful to it given the the problems that we
• 14:52
  had earlier on as well. When you look to the future, how
• 14:56
  do you see the future of your industry?
• 14:59
  Sure. So I'm a big believer that easy
• 15:04
  integrated, cost effective is the is the future, right.
• 15:08
  And so we think a lot about that as we're building our road map.
• 15:10
  So I see that as we're able to automate more translation,
• 15:16
  leverage AI to make it more cost effective, we're going to see
• 15:21
  significant uptake in the amount of content that gets translated.
• 15:25
  We see this all the time A companies like I would love to
• 15:27
  translate that, but I have to ration it because of dollars and
• 15:30
  resources, right? So that's really our business is
• 15:33
  I, I see over the next two or three years getting to the point
• 15:36
  where translation localization is completely automated and
• 15:40
  seamless to the to to retailers and and B2B sellers out there as
• 15:45
  part of the AI powered platform like smart link.
• 15:48
  So you don't have to think. It just makes so much sense and
• 15:51
  it's one of those areas that I guess the technical folks in the
• 15:55
  room, the people running your e-commerce engine may have a way
• 15:59
  of thinking, right? That's how they've traditionally
• 16:01
  done it. But having gone through this in,
• 16:03
  in such a sort of layman's way, I can see that the, you know,
• 16:07
  the negative impact that it has. And so doing it right, you know,
• 16:11
  getting those translations right, getting your SEO right,
• 16:14
  There's so many little gotchas that I I think people sometimes
• 16:17
  take for granted. But anyway, Brian, thank you so
• 16:20
  much for giving up the time for talking to us and look forward
• 16:23
  to seeing you at an event somewhere near in the future it.
• 16:26
  Was my pleasure. Thanks for taking the time with
• 16:28
  me.