Latest / Inside Commerce: Ecommerce Strategy, CX and Technology Podcast / EP275: The Current State of AI in Ecommerce And Where We’re Headed With Visual Content Generation for Brands
Transcript
- 0:00Firstly, you're the boss, not the AI.
- 0:02You should know exactly how it's going to respond to everything,
- 0:04whether it's a chat bot or whether it's Gen.
- 0:06AI or whatever we come on to. I think you should know what the
- 0:10output of the AI is going to be. And then you have a second part,
- 0:13as you just said, it's like the data that you're essentially
- 0:15putting into the AI is everything like that is the
- 0:18value, so. You listen to Inside Commerce, a
- 0:21weekly ecom strategy, technology and CX focus podcast hosted by
- 0:25experienced digital consultants and re platform specialist James
- 0:27Good and Paul Rogers. Get insights into ecom strategy,
- 0:30leading brands, hear from industry thought leaders and
- 0:32leading tech vendors with practical advice to help you
- 0:34improve your online CX. This episode is sponsored by
- 0:38Shopify. A very topical discussion today.
- 0:41We're going to be talking about the current state of AAI in
- 0:44e-commerce. And you know, nobody's talking
- 0:46about AI, obviously, and where we're headed with visual content
- 0:49generation. So kind of a split in the
- 0:50episode, but the second bit's really important because this is
- 0:52one of the big growth areas. Lots of practical insights and
- 0:56we are welcome back today Liam Quinn, who's a technical
- 0:59consultant. So hey, Liam, could have you
- 1:01back on the podcast. Hello, thanks for having me
- 1:03back. Paul's already slagged off the
- 1:05first one, so. I didn't remember.
- 1:07It wasn't very memorable. Yes, yes, for, for, for everyone
- 1:11listening. Paul's comment when I reminded
- 1:12him we'd already had Liam on was it was, yeah, it wasn't very
- 1:15memorable. There we go.
- 1:17This one's going to be because not everyone's as cynical as you
- 1:21right in Liam. Let's get an intro because some
- 1:23people know you, you were previously like run technical
- 1:27side of Ivonne. You now work with Ivonne, still
- 1:29a consultant capacity, but you're on your own thing.
- 1:31And I think you're doing something quite interesting
- 1:33because you're, I know you're doing lots of stuff, but one of
- 1:35the areas is you're building out some Shopify apps to help
- 1:37retailers leverage what we're talking about today, which is
- 1:39the potential of AI. So do you want to give people a
- 1:41bit more of a, a flavour of what you're doing at the moment?
- 1:44Yeah, perfect. That was a good intro.
- 1:45But yeah, I was technical director.
- 1:48So it's just about 5 year anniversary, Paul, actually of
- 1:51us working together. But for the last sort of four
- 1:54months, I'm splitting my time now between still doing
- 1:59consulting above on and building out products.
- 2:03So sort of digital products with the core of AI the.
- 2:08Yeah, currently the start. There's a couple of those that
- 2:12hopefully be imminently released, but Shopify apps were
- 2:18essentially trying to bring the trying to bring AI to to
- 2:25merchants that either haven't got the resource or time or the
- 2:31understanding of like how to get the best out of it.
- 2:33So just simplifying and bringing the value to brands and
- 2:37businesses. Great.
- 2:39And so I'll ask the first question.
- 2:42So I guess like obviously lots of brands been talking about AI
- 2:46and you know, we work with lots of different teams have been
- 2:50lots of different businesses that have different levels of
- 2:52understanding. Some have played around with it,
- 2:55some have built their own kind of products on top of things
- 2:57like open AI. Do you speak to a lot more
- 3:00people than I do around this topic?
- 3:02What do you think the current state is for brand and how have
- 3:06you seen like adoption and usage so far?
- 3:09And also like, have you seen much or like how quickly you
- 3:12seen that growing? Yeah, I think, I think it.
- 3:16Feels, still feels super. Early for like especially in
- 3:20ECOM for some reason, but I think still feels feels really
- 3:24early in terms of the adoption against like the rate of change
- 3:29and improvement in AI. So it feels like it's AI is
- 3:33moving at a, you know, deeper trajectory than I've seen a lot
- 3:38of brands like really, really utilising aside from everybody
- 3:42using ChatGPT and things like that.
- 3:46But it's definitely, I think over the last six months we've
- 3:48definitely started to see adoption like we we've had
- 3:51clients, Paul, I think over six months ago there was people that
- 3:55were dabbling with it and playing with a few different
- 3:57things and like experimenting. In the last six months we've
- 4:00probably seen production use on more and more brands.
- 4:06So I think it is increasing just.
- 4:09Yeah, still early. Yeah, I think so.
- 4:14We've had it like we've seen a couple of clients doing a few
- 4:18different things, but there's been so examples like Mangoes,
- 4:21like the high profile brand where they're experimented 6
- 4:25months or so ago with AI Gen. AI imagery and then turned out
- 4:31to be really successful. And I think they're continuing
- 4:34use and factoring in that into their processes.
- 4:37So yeah, I think it's, it's definitely on the increase.
- 4:42But yeah, there's still the like, a lot of friends together.
- 4:45I think the of our client current bodies probably maybe
- 4:49done the coolest, well, not necessarily the coolest thing,
- 4:51but they use it the most in terms of day in, day out.
- 4:55So they work with Nolo, who we've had on before, Holly
- 4:58Rhodes and they're doing quite a lot around translation, QA of
- 5:03content, that highlighting of content wire, certain US PS
- 5:07aren't mentioned and various other kind of bits within their
- 5:11kind of PLM PIM workflows. And and then World of Books was
- 5:16another one we had on recently that we're using it more for
- 5:19content generation. Yeah.
- 5:21I. Was going to go on.
- 5:22To like the. Current body example, but yeah,
- 5:27already, no, we'll come back on to that.
- 5:28But I think I Andrew, current body actually, it's like I was
- 5:33thinking about earlier, it's a good yardstick of the direction
- 5:36that the industry moves. I think like he's got together,
- 5:39he's quite a bit on like, yeah, where things are going tech
- 5:42wise. And I think, yeah, the facts
- 5:44that they're adopting it more and more in the processes, it's
- 5:48like an indication for some other Brexit.
- 5:50And we'll come on to like third parties and how because I think
- 5:53some of the stuff that people are doing early right now, a lot
- 5:55of third parties are like starting to enter the market to
- 5:58cover some to make some of those areas a bit easier to kind of
- 6:01break into. But I guess second question, so
- 6:06kind of following on from that, where do you see the biggest
- 6:09opportunities short term for kind of e-commerce brands and
- 6:13businesses in terms of like investing into different areas
- 6:17of AI? Yeah, I think I don't really
- 6:19like the question. Like a few people asked this and
- 6:22it's like what's the obvious quick wins?
- 6:26But I think more because I think eventually we get to a point
- 6:31where AI is like laid through all aspects and processes
- 6:35through a brand. And actually the the, the
- 6:39biggest value completely depends on like the make up of the team
- 6:43processes of that brand goes through, you know, all of their
- 6:46operations. So, yeah, I think it can be
- 6:49completely varied in terms of like highest value, highest
- 6:54value return for different brands.
- 6:58But I think some of the obvious ones like customer service is
- 7:01probably the obvious area where we've seen like biggest
- 7:06adoption. And that just makes perfect
- 7:09sense for the majority of brands to be moving to the customer
- 7:12service suite that's more efficient and takes off a lot of
- 7:18the manual effort. I think other areas like so like
- 7:23data entry and validation, we've started to see more like
- 7:28managing translations. You said as part of that.
- 7:30I think all the anything where there's like a lot of manual
- 7:34effort and the requirements like accuracy and efficiency will be
- 7:39picked up by AI. So yeah, I think the data entry
- 7:42validation like reporting we've seen on Census Paul, like that
- 7:47report inside is sort of an easy entry point.
- 7:51So brands where you know you've got the post patcher survey,
- 7:55some of the brands get 10s only 1000 responses a month where
- 8:00it's actually just too much for even if you are the human sat
- 8:02there full time working through that, it's firstly probably not
- 8:08realistic for them to get through.
- 8:09Secondly, you can now do that with AI more efficiently, more
- 8:14accurately and sort of get better results anyway.
- 8:17So, so yeah, that side, I think there are a few of the the areas
- 8:21right now overseeing. One sub point that I wanted to
- 8:24make, just as like a general frustration with AI for customer
- 8:30service, I've had two horrendous experiences recently with
- 8:34companies I think historically would have been way better.
- 8:37And obviously using AI is way more efficient and there's lots
- 8:40of productivity games, etcetera. One was finals.
- 8:44So I bought a hat and it's significantly cheaper on
- 8:47flannels. So I bought it through flannels
- 8:50and then when I went to return it because it's tiny, which is.
- 8:55But when I went to return it, it didn't recognise my number.
- 8:58So I started the only route which was live chat and it was
- 9:02all AI and the whole thing was terrible.
- 9:05And then in the end I found an e-mail.
- 9:07I went through e-mail and like all the cons just didn't make
- 9:10sense. So I was like no contacts.
- 9:13So whilst there's clearly a huge opportunity here and people are
- 9:16doing loads of cool stuff in this space, I think that and the
- 9:20other one was going to be hotels.com where it's just
- 9:22really poorly. It didn't really understand most
- 9:25types of queries that I put in front of us.
- 9:29I feel like there's a real risk to brands creating a really
- 9:33negative experience here as well as alongside the obvious
- 9:37opportunity. I think the issue is like a lot
- 9:40of the early, you know, people getting in early implementing
- 9:44AI, it's like everything was everything chatbot focused.
- 9:49And then it's quite easy to stick a chatbot on site that's
- 9:53like not trained with customer data, not trained with brand
- 9:56data. And it's and it's technically
- 9:58working, but actually there's nothing behind it other than
- 10:01sort of. What do Mattel, Victoria,
- 10:06Beckham, Sennheiser and Glossier all have in common?
- 10:11They power their businesses with Shopify.
- 10:14Shopify is the most innovative and scaled commerce platform on
- 10:18the planet. It also happens to have the best
- 10:21converting check out on the planet Shopify Powering the
- 10:27future of enterprise commerce. Find out more at
- 10:30shopify.com/enterprise. Yeah, and The thing is AIAI
- 10:36doesn't cover over bad process and bad data and bad
- 10:39information. And if people put things in
- 10:42without thinking through the user journey and thinking
- 10:44through the insight and knowledge that's needed to
- 10:46educate these models to, to be able to respond effectively.
- 10:50Yeah, this this is where I think the, the, the big maturity has
- 10:54to happen is, is these tools are amazing, but you can't just let
- 10:58them loose and hope they work. You've got to rigorously control
- 11:02and test them. I've had, I've had a couple of
- 11:04brands who turned off their AI chat bots because they were
- 11:07getting loads of complaints from customers about the quality of
- 11:10response and inaccuracy and response as well.
- 11:13And when you've got, especially when you've got brands that have
- 11:16got specific product, product compliance requirements as well,
- 11:21you have to be very careful. Yeah, I think 100% it's maybe
- 11:26what and on this year more people, I think we're getting to
- 11:30the point where it's like not just hand everything over to AI
- 11:34and hope that it does really well.
- 11:36It's like the firstly, you're the boss, not the AI like you
- 11:40need to you should know exactly how it's going to respond to
- 11:43everything, whether it's a chat bot or whatever it Gen.
- 11:46AI or whatever we come on to. I think you should know what the
- 11:49output of the AI is going to be. And then you have a second part.
- 11:54As you just said, it's like the day data that you're essentially
- 11:59putting into the AI is everything like that is the
- 12:01value. So if you have no put no data in
- 12:04or it's got access to nothing, then it is going to be just
- 12:08completely generic, yeah. I remember having that
- 12:10conversation with, with Neil a about their AI like live Q&A
- 12:17thing. I'm on the PDPS Asklo and that
- 12:19was the whole thing is, is the, the, the, the bot is really
- 12:22intelligent and its ability to, to understand and process the
- 12:26data and understand the different types of data that sit
- 12:29within it, attributes, categories, you name it, and
- 12:32then interpret the natural language of the customer.
- 12:34But if the source data is crap or inaccurate, and let's face
- 12:38it, we've worked with a lot of brands who might be very good at
- 12:40what they're doing in terms of growth, but there's always data
- 12:43inaccuracy, there's always product issues where the quality
- 12:45is not good enough. So yeah, 100% it's a often the
- 12:50what underpins this is the bit that that needs the focus
- 12:53initially. But let's let's move on to the,
- 12:55the Cray because we're not going to do the operational side
- 12:57today. And actually, we're going to do
- 12:59a mini series in the not too distant where we're going to
- 13:02drill into some more that operational back end software AI
- 13:04is doing cool things like warehousing wise, etcetera.
- 13:07But let's talk about like content.
- 13:09I think this is an area that a lot of brands, e-commerce teams
- 13:12are interested in getting smarter in terms of their
- 13:16process, being able to like in, you know, rapidly speed up
- 13:19output as well. And I know you've done a lot of
- 13:21work and you've done a lot of research around visual content
- 13:24generation. What what's your where's your
- 13:26head at the moment? Like how?
- 13:28How can or should brands be using AI from a visual content
- 13:33generation point of view at the moment?
- 13:35What's what's the potential? I think so.
- 13:39I think over the last say I've been let's say 12 months ago
- 13:45when more and more bits of Gen. AI came out.
- 13:48I've been sort of dabbling with it since then, where actually at
- 13:51the time it was like, this is like cool or some interesting
- 13:53things, but for a brand there's like no consistency, you know,
- 13:59little accuracy and it weren't really usable.
- 14:02I think it's I think it's massively gone under the radar
- 14:04how far it has come over the last 6 to 12 months in terms of
- 14:09what you could with the right with the right effort and sort
- 14:14of time and yeah, time and effort put in the output that
- 14:19you can get from yeah, Gen. I Gen.
- 14:22AI automated imagery and videos in particular and and written
- 14:27content as well. I think the written contents a
- 14:30bit more yeah. There's a bit more examples out
- 14:34there of people using it and it's gone from just, you know,
- 14:37ChatGPT again, generic ChatGPT descriptions that people put on
- 14:42site and weren't very good. So now you can be trained
- 14:45completely on like tone of voice, what sort of language the
- 14:49brand uses, even like, you know, localised like Fred Perry might
- 14:55use a different tone of voice and language in different areas
- 14:59of the world or areas of England.
- 15:02The written content you could get now is like heavily trained
- 15:06on that can be really accurate. And yeah, imaging video now as
- 15:10well. So I think they're still like me
- 15:12and Paul were talking about before.
- 15:13I think for some pre, you know, like a Frankie shop or a
- 15:18self-portrait, I think you still get the materials not 100%
- 15:22accurate or the, you know, certain bits not 100% accurate
- 15:25to be used overly on site. But I think, I think right now
- 15:31it's probably for 80% of brands, you could get an end result
- 15:35that's like super usable, a lot more cost effective, a lot
- 15:39quicker than trying to produce it yourself.
- 15:44Yeah, so a couple of things I've found really interesting is, is
- 15:47you're saying like the the quality of the tech has
- 15:50increased rapid. It's that classic curve of of
- 15:52tech improvements, isn't it start, don't it gets rapidly and
- 15:56actually human adoption lags. But Adobe, you've got big, big
- 16:00businesses like Adobe launching Firefly to do a a, a Gen.
- 16:05AI videos and images. And some of the quality I've
- 16:07seen, I know, I think it's only on beta.
- 16:09I think you have to be an A customer and apply for but some
- 16:12of the quality I've seen in the outputs for some of the animated
- 16:15videos, it's really crystal clear and short.
- 16:18I don't know about the brand lens control over at the top of
- 16:21it, but the quality of the actual the videos is good.
- 16:24And there's, there's some other nice use cases that come out in
- 16:27the last like 6-12 months for brands or businesses who are
- 16:30using this as part of their creative process.
- 16:33I know next one or the 1st the next one, a really good YouTube
- 16:35channel. They they talk about, you know,
- 16:37what they're up to within the business.
- 16:38You're really like. And one of them was they've used
- 16:41Gen. I I to create product concepts
- 16:44for new lines. And then they test those out on
- 16:48their customers to see whether the the designs are popular, not
- 16:51popular, what they like, dislike and they fine tune and that then
- 16:54informs their products and planning and buy in process,
- 16:58which I think is really, really smart.
- 17:01Yeah, I think, I think you remember when we touched on, you
- 17:04know, like these SL Portrait or Frankie Shop, maybe the premium
- 17:08brands were it's not 100% perfect for your main products
- 17:13gallery. I think there's still I think a
- 17:16lot of people would be surprised the quality you could get and
- 17:19for things like, you know, blocks on site where it's more
- 17:22how you could style that product rather than it being like the,
- 17:28you know, the main product image.
- 17:31And the other one that I think is probably the use case for
- 17:34right now is the basically like meta ads or social ads.
- 17:39So being able to quickly output a lot of high quality testable
- 17:47assets that, you know, people might only see for seconds or
- 17:50minutes. And then, yeah, doing that at
- 17:54much higher skill than you'd be able to manually do.
- 17:57I think that's probably a starting point.
- 17:59But yeah, I think there's a lot of, I think that in the next six
- 18:04months will be we'll see a big change in that because you can
- 18:07get like super high quality video and image now.
- 18:13I think the, the difficult or not the difficulty we have, but
- 18:16we have obviously Vivon is very fashion focused.
- 18:20So we immediately go towards clothing and like model shots.
- 18:25But I think the opportunity for like consumer products, like
- 18:29even FMCG products is much bigger.
- 18:32And like if you've got massive catalogues as well, you know,
- 18:35some in dynamic product cards and everything 'cause I think
- 18:38there's going to be loads of opportunity there and current
- 18:42limitation. So I just mentioned fashion, but
- 18:45like how do you like, how far along do you think we are And
- 18:50how kind of viable do you think the kind of outputs are for kind
- 18:56of like higher level high fidelity, like yeah, model
- 18:59charts, branded content, etcetera.
- 19:02And then what human elements do you or like?
- 19:04How could AI fit into a human workflow to best to achieve the
- 19:10optimal assets? Yeah, I think, yeah, without
- 19:14going to plug, I think what I've been working on is trying to
- 19:16simplify the process from a merchant or you know, like econ
- 19:21director, whoever's, whoever's managing it to get that quality
- 19:26output. So I think you can right now
- 19:30with a reasonable amount of manual effort involved like
- 19:35train on, train on the brand and what style of photography you
- 19:38want to get output and get something like super usable.
- 19:44But I think the limitations are that bit of effort upfront.
- 19:49So it does take effort. It does take, yeah, a bit of
- 19:53manual input and training and a few different steps to get
- 19:56there. So I think trying to simplify
- 19:58that, I think the other limitations are have been like
- 20:02consistency. And then probably brand, so you
- 20:07know, accuracy on really complex designs or like you know, brand
- 20:15logos and things on on products. But I think there is ways around
- 20:20those that it's, it's trying to, Yeah, it's trying to solve those
- 20:26problems in a way that doesn't take huge amounts of efforts for
- 20:30the team. But and I think right now you
- 20:33can get around those. I think in six months time
- 20:35though that won't be an issue. That's sort of even been talked
- 20:39about now. So I've got a general question
- 20:42on this. And Paul, I know you have lots
- 20:44to to say because you work with a lot of, of creative teams and
- 20:47I know you're also very hot on creative execution when people
- 20:51are like launching websites as well, down to the control of
- 20:55the, the design and the visual sharpness of that.
- 20:58And AI generation. AI can generate images.
- 21:01We all know it. And it can generate images from
- 21:02prompt. It can take image inputs from
- 21:05you and it can embellish them. There are different ways that
- 21:07you can use this tech, but fundamentally each business has
- 21:10a different set of brand controls, brand guidelines, and
- 21:12also how stringent they are within the what you can do
- 21:17within those creative roles. And some are completely
- 21:19obsessive to the last pixel and others are a bit more loose and
- 21:23free. So my general questions about
- 21:26what is the role of creative direction alongside AI?
- 21:30How, what, what type of expertise, what type of people
- 21:34are needed to make sure that this is executed well, not just
- 21:37technically because there's a technology expertise, but
- 21:41there's the visual oversight. What, what, what do people need
- 21:44to think about? And, and from what you're
- 21:46seeing, how are people handling this?
- 21:47They don't just let a load of wild stuff get out there are.
- 21:54You going first, Paul? Well, I was going to let.
- 21:57I was going to let you go first. You go first then.
- 21:59Yeah, I can go. I think, yeah, I think there's
- 22:03obviously a lot of conversation around AI replacing jobs and
- 22:08like who's actually at risk and and whatever, whatever.
- 22:11I think the like art director, creative director sort of roles
- 22:16are even more important for all of these.
- 22:19But I think while AI and I think what we've just been talking
- 22:23about, you can get like at scale a lot of outputs to high
- 22:28quality. I think the bit that separates
- 22:33separates that going forward is the creative director, art
- 22:36director, that sort of role overseeing it and actually
- 22:40achieving again, going back to you're the boss, not the AI is
- 22:44not the boss, you're the boss. It's like actually achieving
- 22:46what the vision is and AI is just action in action in the
- 22:52task. I think one thing is an example,
- 22:55not really econ related, but that the Coca-Cola advert that
- 22:58everybody obviously like LinkedIn went wild for around
- 23:03Christmas time. Should the, the AI can
- 23:07technically now deliver Like it was pretty good quality in terms
- 23:12of literally the, you know, visual aspect of, of what it's
- 23:15outputting, But without that creative director role of what
- 23:18the point of it is and storytelling, the emotion of it
- 23:21all, that sort of thing. It's pointless and it's not a
- 23:24good end result. A good example of of an
- 23:28equivalent that a company put out there without the kind of
- 23:32consideration around, like the brand elements with Toys-R-Us.
- 23:37So they created what was on paper a pretty good asset or
- 23:41like piece of content, but like they were missing so much of
- 23:45like the nostalgic elements of Toys-R-Us.
- 23:48I think it had the giraffe, but it didn't up with music and like
- 23:51they were like, it was all just a bit rubbish.
- 23:52And I think that's a good example of where like I
- 23:56personally see the trend at the moment is creative direction,
- 23:59art direction, photography, like all of the technology around
- 24:04these areas people are investing more in.
- 24:07And then it's more of the development and like execution
- 24:10where people are investing less in.
- 24:12So I think for like the medium term, the, the role of like the
- 24:16people that are directing, their value will go up and then
- 24:21execution will gradually disappear over time.
- 24:25But I do, I would imagine over time that these like models will
- 24:29be trained on how to develop like, you know, unique artistic
- 24:36direction. And I think over time, it'll
- 24:39require less time to deliver the projects that, yeah, currently
- 24:44are completely bespoke and unique.
- 24:47Yeah, I think there's, I do think there'll be always that
- 24:51element of the human creative direction being being needed
- 24:56because some of it is like what's happening right now in
- 24:59the real world and and things like that factor into what that
- 25:03content should be. And I think even obviously
- 25:06content writers were probably the first one where it's like,
- 25:09you know, our content right is dead because AI can now do that.
- 25:12But actually good content writing will be is like nuanced
- 25:16to the brand can reflect what's going on right now in the world
- 25:19might feel like, you know, like subtle gig at a competitor of
- 25:26things like that where actually, and then it's woven through not
- 25:29just the name of a product or a poster, but it's woven through
- 25:32everything. And I think that actually a good
- 25:34content right here is even more valuable.
- 25:39But using AI to sort of, yeah, run the task and and get it done
- 25:44at scale. It it's also the governance
- 25:46piece, isn't it? There was a high profile case.
- 25:48I remember Paul Connor, it works for Fiat Vivonne and flagged
- 25:53this of a one of the the couriers live chat thing that a
- 26:01a user managed to get it to abuse itself and say that it's
- 26:04crap because whoever had set it up hadn't obviously put the
- 26:07right guard rails in place. And when using any Gen. 2 tool
- 26:13and you've got to 1st validate how that model is working, what
- 26:16AI it's using what models it's using to drive that and what
- 26:20safeguards there are in to stop them because they they all have
- 26:24to only be able to take from legitimate content.
- 26:27They can't use copyrighted content without explicit
- 26:31permission. So you have to make sure it's a,
- 26:33it's a legitimate on the one area that I want to flag, as
- 26:35well as this whole thing about copyright awareness, because I
- 26:38think it will probably surprise a few people, the fact that
- 26:41certain assets that you create using AI will not be
- 26:43copyrighted. They'll be public, which means
- 26:45anyone can then use them and they don't have to credit you.
- 26:48And obviously your big thing around brand is creating unique
- 26:50visual identity. It's creating assets that are
- 26:52yours that can't be used by competitors or other companies.
- 26:56And I know that I'm not going to profess to be legal expert on
- 26:59this, so tell people to read up properly.
- 27:01But what I do know from what I've been researching is that
- 27:04the law around AI generated content is developing.
- 27:06It's not rigid and there are different rules depending on how
- 27:11much human input you've put in to create something that is
- 27:15unique, original, personalized and can be demonstrated to have
- 27:19human creativity versus a model has created off of its base
- 27:23level of data and knowledge. So I think this, I, I don't hear
- 27:26it. I don't hear conversations
- 27:27around the copyright impact. I don't know whether you guys
- 27:30here or whether your brands are, but when I speak to people, I've
- 27:33never heard someone say, oh, when you think about the
- 27:35copyright potential of this, do you, do you hear that?
- 27:37Are people talking about it? I've only heard it at the
- 27:40highest level within businesses around the other way, so more
- 27:45around people not wanting to hand over certain information to
- 27:50open AI due to the lack of copyright protection.
- 27:54But I, I think that that will come more and more as it
- 28:00progresses, matures and I think as some of these third parties
- 28:02drive more volume. Yeah, I was going to say the
- 28:05same. I think that's the only sort of,
- 28:08that's the only at times I've really heard it being brought
- 28:11up. I think, yeah, it's a bit of a
- 28:12minefield. I think the issue is that that
- 28:15legal side moves far, far slower than the sort of technology
- 28:19moves. So.
- 28:20There'll be one. It's always 1 high profile issue
- 28:24that starts the issue. And so I'm sure that'll come at
- 28:27some point. So I guess next question, Liam,
- 28:31in your view, because this AI and all of the surrounding
- 28:34technologies with all the money involved feels like things are
- 28:37moving so quickly and to the point where you could build a
- 28:41technology and that it could be obsolete by the time it's out
- 28:44due to developments from another technology.
- 28:47Do you think brands should be investing now for the long term?
- 28:50Like, do you think they should be training models?
- 28:52Do you think they should be, you know, structuring data in a way
- 28:55that allows you to lay on more AI?
- 28:58Yeah. What's your view?
- 28:59Yeah. I mean, I definitely think so.
- 29:01So I think it's at the point where if a brand's not.
- 29:05At least. You know, planning or starting
- 29:07to think about how to implement AI and having a strategy around
- 29:10it. I think it it'll be a
- 29:11competitive disadvantage by not doing it.
- 29:15All the competitors will be, you know, a greater advantage having
- 29:20AI, you know, within their operations.
- 29:23So yeah, I think it definitely is a point where it should be
- 29:26started thinking about investing for the long term.
- 29:29I think you've just flagged like the main caveat, which is, you
- 29:33know, the AI models themselves, even Open AI Llama, all those
- 29:36models are, you know, every two months sort of look completely
- 29:41different. And I think then the tech that's
- 29:42built on top of those is changing constantly, so.
- 29:46Like. I think brands definitely can't
- 29:49pick a technology and expect in three years time it's like this
- 29:52is our three to five year plan with Pixie's piece piece of
- 29:55technology, let's build for it. I think that's not going to
- 29:57happen. I think yeah, the tech even more
- 30:01so than ever. I think that tech landscapes
- 30:03changing more quickly and there's like a need to be more
- 30:08agile and experimental in in being involved in which I know
- 30:12James is typically the opposite view of of the experimental
- 30:19nature. But I think yeah, I think that
- 30:21is a big caveat. And it sort of has to be when we
- 30:24come back to like current body or whatever, there is that
- 30:26element of being sort of experimental.
- 30:30But I think going back to a point that you made earlier,
- 30:34James, the the key part is actually the strategy around
- 30:37what our data is, where our data is held, what the format of all
- 30:41the data is, getting all that into a place where we know that
- 30:45AI, you know, having a strategy for AI and it is going to be
- 30:49based around all of the brands data.
- 30:51So getting that into a place where it can be used.
- 30:54I think actually different to how tech has worked in, you
- 30:59know, the last 10 years, I think that layer of what the
- 31:02technology is will change. But to get the the data into a
- 31:07structure which if it's one piece of tech or another doesn't
- 31:10matter because you've got all that set up in the most optimal
- 31:14way is the key. So yeah, I think yes.
- 31:20But if yeah, but bigger brands I think starting with starting
- 31:23with the data and working outwards rather than the tech
- 31:26working inwards. And something I'd, I'd, I, you
- 31:31know, we, our industry loves buzzwords.
- 31:33It always happens. Something new comes up and AI on
- 31:35its own is not enough. We've got to have new buzzwords
- 31:37around AI subbuzzwords and lots of people now talking about
- 31:40agentic, agentic, whatever, however you want to pronounce
- 31:44it. What, what is your view?
- 31:47Apart from the fact that there's always the element of industry
- 31:49that likes to create content around new ideas, What is this?
- 31:53Is this anything different? Is this something that people
- 31:56need to be aware of? Does this mean people need to
- 31:58think about the impact and use of AI in a different way?
- 32:03Yeah, I think it's a great buzzword.
- 32:05It is like the perfect word for a buzzword.
- 32:08I try and avoid using a genetic or, but yeah, I do think AI
- 32:14agents as a, as a concept is, is almost like without being
- 32:20extreme, it's almost like when the internet's gone from like
- 32:22read only, you know, here's pages of text to the next
- 32:27evolution, which is like immersive, you know,
- 32:29interactive, going both ways. So logins and being able to
- 32:36send, you know, to the browser and all that sort of thing.
- 32:39So actually interactive. I think this is the similar
- 32:42switch where instead of just being passing an input, get out
- 32:46and output, it's going to sort of open up.
- 32:52Yeah, all the possibilities of basically handing over a problem
- 32:57and AI is thinking about it going away and actually
- 33:01performing actions. So yeah, I do think it's, I
- 33:04think that sort of out of everything we've talked about, I
- 33:07think this is the biggest game changer, not only not just in
- 33:10econ, but in like technology. So I think it completely
- 33:13transforms how a lot of things sort of work under the hood.
- 33:17I think, I think part of it's driven always say through
- 33:20commercial value because I think most or a lot of businesses,
- 33:24maybe outside of some of the SAS businesses, people are the
- 33:27biggest overhead. And I think investors have been
- 33:31particularly interested in that, that I've had a couple of
- 33:34conversations with different people.
- 33:36And that's like the first thing that's been brought up.
- 33:37And I think, yeah, I think there's a lot of people now
- 33:41building products that are just trying to target roles.
- 33:45And then I also think like Marc Benyar from Salesforce have
- 33:48driven quite a lot of this with their like agent force stuff.
- 33:53But yeah, I personally think I think it is a bit of buzzword
- 33:57and I obviously mentioned it in my little predictions.
- 34:01So yeah, I've obviously fallen into her.
- 34:05But I think that it is a good way to start personally, if
- 34:10you're if you do fall into that category where people are your
- 34:14biggest donor. I think we're like, I think
- 34:16we're so alien what the possibilities of the agent side
- 34:20they're going to be. But I think it's interesting,
- 34:23like there's AI put a, I was dabbling with a few things, like
- 34:27a couple of months ago, I put a video on LinkedIn where it's
- 34:30basically using an AI agent to find a, basically give it a
- 34:35budget and the products that I was after.
- 34:36And it's like, go and find a site, check if the price is in
- 34:40my budget. If so, add to cart and check
- 34:42out. And actually that in theory,
- 34:46that's quite cool to see. But the theory, like the next
- 34:49step of that is actually if you get to a point where, for
- 34:52example, people are using that to do any of their online
- 34:57shopping or, you know, if we're using agents instead of doing it
- 35:01ourselves, you get to a point where there's a certain
- 35:03percentage of traffic hitting sites that are AI agents instead
- 35:07of humans. Like, at what point do you, you
- 35:10know, everything to do with UX and Sierra potentially shifts in
- 35:15the future? Like, at what point do you start
- 35:17optimizing a site for AI agents rather than humans?
- 35:22Yeah. I mean, these are long, long
- 35:24term decisions, but that's sort of what this happens to Doctor.
- 35:28Proper dystopia, this Yeah, we always love those because me, me
- 35:32and my, like, Luddite Misery, who outside of my work, doesn't
- 35:36want tech to do anything for me because I just want to stare at
- 35:39fields and mountains. The vision of humans sitting at
- 35:43home asking an AI agent to go and do everything for them
- 35:48terrifies me. It's like this dystopia film.
- 35:49We're all at home and we're just basically not.
- 35:51Actually. It's like The Matrix.
- 35:52We're not awake, we're just, we're just a consciousness
- 35:55that's feeding AI. Yeah, I mean it is, yeah.
- 36:01It's a strange one, but I think yeah, the agent side up and so
- 36:04daughter that I that sort of concept.
- 36:07So at some point down the line that that is likely where it
- 36:12gets to. So yeah.
- 36:15Yeah, I definitely, I find it interesting because I can see
- 36:18productivity wise, I can see why this is really beneficial to
- 36:21businesses. And you're already seeing this
- 36:23in in the way that that some of the customer service chat bots
- 36:28have the capability to execute specific commands then.
- 36:31So at the basic level, I genetic, isn't it like, you
- 36:33know, what's my order status? Well, OK, give me your order
- 36:36number. You're logged in.
- 36:37I validated that. Give me your number.
- 36:39I've now gone and checked in whichever system I needed to
- 36:42look that up, find the status, return it and tell you on the
- 36:45front end. So we've got it at the basic
- 36:47level. But yeah, I guess I'm intrigued
- 36:49to see where people want to take this and how many people, how
- 36:53many people actually want broader as customers want a
- 36:56broader service versus just want fundamental questions answered.
- 37:00So yeah, treated in free to say and then we always like to talk
- 37:05about tech on this one. You we talked about a few tech
- 37:09companies, journalists, but in the context of like AI Gen.
- 37:13AI and content side. Which?
- 37:17Third parties that you've looked at, have been working with, or
- 37:19you know other people have been working with have most impressed
- 37:21you and why? Yeah, I think I've made a note
- 37:24of a couple of points, but I think to start with.
- 37:27So for a shop, if I had to start with because not actually
- 37:30because what they've rolled out AI wise is like mind blowingly
- 37:36impressive currently, but I think the opposite of all the
- 37:40work that they've been doing on data structure and architecture
- 37:44of the platform and the way that the themes work and all of that
- 37:47side of the whole platform is literally like gearing up to be
- 37:51the perfect architectures for from best use of AI or like
- 37:57optimal use of AI. So I think they will, yeah, I
- 38:00don't know when, but in in the coming period of time do some
- 38:05really interesting things having I did see actually science.
- 38:08But to your part in the spot last week, I saw a screenshot of
- 38:12a shot Shopify in the admin basically prompt to like how do
- 38:16you want the theme? It was a prompt to generate
- 38:19theme or liquid cut essentially. So I think I mean that will be
- 38:22imminent. I've heard little rumours of
- 38:25certain things that people in within Shopify are doing and
- 38:29using. So yeah, I'm expecting like a
- 38:31major update at some point this year that I think will drive a
- 38:36lot of change. I think actually my point around
- 38:39the risk of developing tools for the kind of within around AII
- 38:45actually think in the commerce space, Shopify is a bit of a
- 38:48threat to anyone kind of building.
- 38:51Yeah, tools in this space. Yeah, I completely agree.
- 38:54So I think yeah, not impressive as as you know something right
- 38:59now that's mind blowing. But I think they're they're
- 39:01geared up together. I think 1.
- 39:04So one that I was going to mention is will be very under
- 39:07the radar, but model guide. So there's a couple of guys that
- 39:10were commerce UI or involved in commerce UI that it related to
- 39:16Ascola. So I know you had the guys from
- 39:19Ascala on a few weeks ago or months ago, but similar sort of
- 39:23concept. But the reason it's interesting
- 39:25is it's basically like super vertical specific.
- 39:30So it's a bit like the way so answering questions dynamically
- 39:38similar to Ascala, but is super like trained on just beauty
- 39:42model guide the IO, but it's like trained really heavily on
- 39:45the beauty industry, beauty and cosmetics.
- 39:49So yeah, and basically relevant to any who were I think the
- 39:56originals or something on there, but to yeah, cosmetic sites.
- 40:00So if you're like allergic to certain things or have a certain
- 40:03skin routine or something, you can dynamically answer questions
- 40:06that are like really personalized to what you need.
- 40:09I thought it was really. I thought the fact that it's.
- 40:13An app that's instead of being sort of tech specific is like
- 40:18vertical specific. I think that could be that could
- 40:22start being a trend where actually the tech side becomes
- 40:27easier to replicate and the when we go back to again, data being
- 40:31the key to the success of the AI think products where they're
- 40:35like heavily focused on specific contexts like beauty.
- 40:42It is really interesting. So yeah, I thought I'll mention
- 40:45that one as as something that could be a direction where
- 40:50products start to go. Excellent.
- 40:53Cool. That brings us to the end of our
- 40:55questions. Although I feel like I could
- 40:57just dive into a side topic of agentic AI for days now.
- 41:02For those you listening, I hope it's been interesting.
- 41:04If you've got any questions and you want to probe Liam further
- 41:08on his views or also what he's doing in the apps he's starting
- 41:11to build out, then reach out. We'll have the contact details
- 41:14on the landing page. It's also just look up Liam
- 41:16Quinn on LinkedIn. Liam, thanks for joining us
- 41:19again. Yeah, thanks for having me.
- 41:21I enjoyed. It excellent take everybody
- 41:23until next week. You're listening to the
- 41:26Insidecommerce podcast, hosted by experienced consultants James
- 41:29Gert and Paul Rogers. Insidecommerce is your fast
- 41:32track to digital knowledge, featuring interviews with
- 41:35leading e-commerce retailers, industry thought leads and the
- 41:38C-Suite key tech companies to explore the latest digital
- 41:40strategy. Customer experience and tech.
- 41:43Discover what the latest trends really mean and get practical
- 41:46advice based on the experience to help fine tune your E Comm
- 41:48strategy and tech stack. Follow us on LinkedIn and reach
- 41:52out to James and Paul if you'd like to discuss an e-commerce
- 41:54project. We hope you enjoy the latest
- 41:56episode.