Latest / Inside Commerce: Ecommerce Strategy, CX and Technology Podcast / EP280: The AI Workspace For Product Catalogues - How Emfas Automates Quality Improvements & Solves Data Challenges
Transcript
- 0:00So trying to bridge the gap between the the product team and
- 0:03the e-commerce team. So for some of our customers,
- 0:06we're actually even integrating with their PLM systems like for
- 0:10the product life, life cycle management.
- 0:12And in case that you have such a system that is so you don't have
- 0:15to have that, but but if you do, then you can get a really good
- 0:18sort of automatic pipeline of designing products and then
- 0:22going into Mfas and then we automatically enrich them and
- 0:25they are ready for your e-commerce.
- 0:28You listen to Inside Commerce, a weekly ecom strategy, technology
- 0:31and CX focus podcast hosted by experienced digital consultants
- 0:34and re platform specialist James Go to Paul Rogers, get insights
- 0:37into ecom strategy, leading brands, hear from industry
- 0:39thought leaders and leading tech vendors with practical advice to
- 0:42help you improve your online CX. This episode is sponsored by
- 0:47Shopify, another really interesting and highly relevant
- 0:51topic, which is about AI driven product data fixes, how you can
- 0:54automate quality improvements and solve data challenges.
- 0:57And this comes up in every business main pool ever worked
- 1:01with any whether that's re platforming or just people
- 1:03trying to improve how they drive product discovery in the online
- 1:06store. There are always product data
- 1:09quality issues in the business and the sources and and causes
- 1:12of them are myriad and some are technology based, some are
- 1:15process based, some are skill and knowledge base.
- 1:18So we've got a really interesting technology vendor Co
- 1:22founder on today who background is in head of dating technology
- 1:26for a lead in ecom retailer ask it, who we had on the podcast
- 1:29last year, but has now taken his brand and retail knowledge and
- 1:33turned that into a really interesting product.
- 1:36So a warm welcome to the Co founder of Enfast, Veda
- 1:39Troyenburg. How are you?
- 1:41Thank you. Really nice to have you guys.
- 1:43Yeah, appreciate you joining us again, second time in the
- 1:46podcast. How does it?
- 1:47Feel. Yeah, it's like a recurring
- 1:49thing. Yeah, well, you've gone from
- 1:51from the retailer e-commerce leader to the technology
- 1:55supplier now. So this is going to be
- 1:56interesting. So before we start getting into
- 1:58the the questions around like the the that this this concept
- 2:01that AI and certain product fixes, let's talk about you
- 2:06MFAST. So how have you got to being a
- 2:08Co founder of MFAST from where you were?
- 2:10Asked. Sure.
- 2:11Yeah. So I started out working at
- 2:14Ascot and leading the sort of tech development over there for
- 2:18the past five years. And during my period there like
- 2:21I've seen like a bunch of different technological
- 2:24projects. So we did a big RE platform to
- 2:27Centra back in the day. I've also done some work within
- 2:30like CMS integrations and the CRM even.
- 2:34And then more recently I also did a PIM project and I think
- 2:39during that time that was like similarly when LLMS and like
- 2:45large language model technology became widely available.
- 2:49And at the same time I had my two Co founders now Rasmus and
- 2:54Carl, they were building this large language model
- 2:57applications in the education tech space.
- 3:01So around 2023, we sort of like put our our heads together and
- 3:08we took a look at like e-commerce and how you can
- 3:12actually leverage AI and large language models to improve the
- 3:16sort of product enrichment problems that all brands are
- 3:20struggling with. So yeah, we sort of put the
- 3:22one-on-one together and and now we we found that in fuss.
- 3:28Great. And then your, so your strap
- 3:30line is the AI workspace for product catalogues and obviously
- 3:33you've just given us a bit of a high level intro.
- 3:36Can you talk us through what that means in practical terms
- 3:39for brands and end users? Yeah.
- 3:43So one thing that we have found now that we've spoken to like
- 3:48over 100 brands is that like every single brand and and like
- 3:52e-commerce person is some way or another using ChatGPT or or
- 3:58similar technology in their day-to-day.
- 4:01And I think this is a good indication that I mean the
- 4:05technology is here and it's solving a lot of problems, but
- 4:09there is no real efficient way or like systemic way of applying
- 4:14those types of technologies on your tote like whole product
- 4:18catalogue. So we are sort of bridging the
- 4:21gap of making the technology available, but at the scale of
- 4:25your whole product catalogue in a really easy to use UI.
- 4:31And I mean, since every e-commerce brand have their
- 4:34product catalogue, it's sort of a a very common problem that
- 4:38that all brands share and that we're solving across the
- 4:41industry. And in terms of like what are
- 4:47the biggest issues that you see from a catalogue quality issue
- 4:50perspective and like what level of impact do you see?
- 4:54And then, yeah, kind of how does that translate into Empire?
- 4:58Yeah. I mean, from speaking with all
- 5:02of these brands and I don't think I'm alone about this, that
- 5:06I've seen a lot of like nightmarish experiences where
- 5:11where you have this like really complex and big product
- 5:14catalogues that are managed manually and this sort of
- 5:20quality is sort of falling apart or it's like really difficult to
- 5:25stay consistent across your whole product catalogue.
- 5:28And I think for many like brand driven e-commerce like brands,
- 5:37it's really important to be able to provide a consistent sort of
- 5:40story and consistent product like a data sort of experience
- 5:49to the customer. And seems like brands are
- 5:52struggling a lot with this. If some products are sort of
- 5:56subpar to that quality, then that will be be very visible in
- 6:02in an overall like really nice looking brand.
- 6:06So yeah, I think data quality issues issues are are everywhere
- 6:10and for like brand heavy e-commerce stores, this is a a
- 6:15big, big problem. And I've, I've a bit of a sub
- 6:19question, this kind of eats into the final question where we'll
- 6:23ask you a little bit around kind of where the product's going.
- 6:26But I've obviously seen the product and you've kind of shown
- 6:29it to our team. And at the minute, I think
- 6:31largely you're kind of focusing on like existing text based
- 6:35attributes and like translations, that kind of stuff
- 6:38with ASCA. A big part of your PDP
- 6:42proposition was the kind of transparency DPA star.
- 6:47And then I guess like where, where do you see this going?
- 6:49And I don't know if you're hopefully you're happy to answer
- 6:51that, but is it a case of, you know, you plan to move into
- 6:55visual content, you plan to move into some of that, Yeah, more
- 6:59supply chain based stuff like how how deep do you intend to go
- 7:02around kind of catalogued it. Yeah, that's, that's a really
- 7:07interesting question. So I think we've sorted out
- 7:10with, as you say, we're solving all of the immediate problems
- 7:14regarding like common product related data.
- 7:17So, so most of the things that you will see on a regular PDP.
- 7:21And currently we we solve like all of the text based attributes
- 7:27where they can be like categorisation, product
- 7:30descriptions, but also a lot of SEO things like SEO meta
- 7:34descriptions and meta titles. And what's, what's super obvious
- 7:39for us as a company is that visual content is, is definitely
- 7:43next in yeah, on our road map. So working a lot with yeah, like
- 7:48imagery, product imagery and being able to, to translate from
- 7:53image into text attributes and like features.
- 7:58So being able to extract that from images automatically.
- 8:01And this is something that we're actually rolling out in, in beta
- 8:05right now and we have widely available for all of our
- 8:07customers in March. So I think that's going to be
- 8:10super, super exciting. But but All in all, I, I would
- 8:14say that like just solving text based content questions is, is
- 8:21really the majority of the work that's most brands are are
- 8:25pouring into this. So we started out in the in the
- 8:28big sort of problem space, but now we're moving into visual.
- 8:33You've obviously got a lot of experience working in retail
- 8:37side yourself and working across multiple teams in an on channel
- 8:40environment. We've worked with lots of
- 8:42different businesses, whether they're pure on lines, large
- 8:45Omni channels. And you'd see typically see like
- 8:49commonality around what the causes are for the disconnect
- 8:54between data that ends up on a website versus what the ultimate
- 8:58level of data should be in order to sell products effectively.
- 9:02Instead of me rambling about what I see, I'd love to hear
- 9:04from you. What do you think of the the
- 9:06main causes? What?
- 9:08What have you seen and where does this inaccuracy creep in?
- 9:12Yeah. So I think this is one of the
- 9:15big problems with an e-commerce that that all brands are
- 9:19struggling with. And and I would sort of
- 9:22summarise that as being the it's a natural disconnect between the
- 9:28product team or like you know, production team and the
- 9:32e-commerce team. So you have, you know, the the
- 9:35product team who are designing products or or purchasing
- 9:38products. And then you sort of have to
- 9:41have this kind of hand over into the merchandising and e-commerce
- 9:45team. And this sort of bridge between
- 9:49is often very long and there are different teams, different
- 9:52stakeholders. So it's difficult for them to to
- 9:55work efficiently. And along this sort of Rd. you
- 9:59have all of this scattered and very manual processes.
- 10:04So like Excel spreadsheets that need to be filled out and like
- 10:09files that need to be gathered and uploaded into Shopify or as
- 10:13your e-commerce platform, for example.
- 10:15And I think that providing the the correct tools to, to bridge
- 10:21that gap is exactly what we're trying to solve here.
- 10:24So trying to bridge the gap between the, the product team
- 10:27and the e-commerce team. So for some of our customers,
- 10:31we're actually even integrating with their PLM systems like for
- 10:35the product life, life cycle management.
- 10:38And in case that you have such a system that is so you don't have
- 10:43to have that. But but if you do, then you can
- 10:45get a really good sort of automatic pipeline of designing
- 10:49products and then going into Mfas and then we automatically
- 10:53enrich them and they are ready for your e-commerce.
- 10:58What do Mattel, Victoria Beckham, Sennheiser and Glossier
- 11:03all have in common? They power their businesses with
- 11:06Shopify. Shopify is the most innovative
- 11:10and scaled commerce platform on the planet.
- 11:14It also happens to have the best converting check out on the
- 11:17planet Shopify Powering the future of enterprise commerce.
- 11:23Find out more at shopify.com/enterprise.
- 11:27Yeah, I always find it interesting.
- 11:29We're working with businesses to see where product data ownership
- 11:33sits as well. And, and often I find the
- 11:37biggest problem is that everywhere there might be 3 or 4
- 11:39people who own data, different parts of data.
- 11:43So you, you might have like they, they owning the SKU and
- 11:47the core product naming and categorisation sits in one team.
- 11:53Then owning like the all the cost and pricing stuff sits
- 11:56somewhere else. And then you've got all the rich
- 11:58from bet. And actually it's it's amazing
- 12:01how how not having a a centralised view of it or
- 12:05centralised owner means that that inefficiency creeps him
- 12:08from the start and then it's pain in the arse to fix
- 12:10retrospect across a large product.
- 12:12Set, one thing that I can mention just like anecdotally
- 12:16from speaking with with a lot of customers that we have now is
- 12:19that most brands don't actually know the whole.
- 12:24There's no one person who knows the whole scope of like where
- 12:29the data comes from. Even the brands themselves don't
- 12:32know that. So what we often do when, when
- 12:34we come in to, to help a customer is we, we help them
- 12:38just mapping out all of the different data points, where
- 12:42they come from, who owns it and where it ends up.
- 12:46So, so that's a big part of what we do as well.
- 12:49Yeah. And that is hugely important
- 12:51because it's amazing. And this isn't a criticism of
- 12:53any business because we all know when you were, when you're busy
- 12:56and you have so much to ask you on, inefficiency is inevitable.
- 12:59It just happens. But over time you get that, that
- 13:02product data debt where you you aren't even aware of some of the
- 13:05gaps because it's not, you can't see the impact.
- 13:08So that's, yeah, I think that's really nice.
- 13:11But I'm going to, before I, I'm going to, I'm going to end up
- 13:13sidetracking down a little rabbit hole here because I find
- 13:16this subject fascinating. But let's get back to the, the
- 13:19premise of your product is about AI automation.
- 13:23So it's about stripping out manual process, it's about
- 13:26trying to provide efficiency. But we are now, I think I
- 13:29believe we're reaching a tipping point in AI where there's so
- 13:32much AI and now there is quite rightly increased scrutiny
- 13:38around like governance and how you use AI intelligently,
- 13:42effectively and how you control it so it doesn't go off and do
- 13:45rogue things. So one of the key questions I've
- 13:47got is about brand controls and, and some businesses are going to
- 13:51be more concerned and this and others, but you know what people
- 13:56are worried about. How do I make sure it doesn't go
- 13:58off and, and write things that are completely at odds with how
- 14:01I want to present my products or safeguards are built into the
- 14:04products? And how can people like control
- 14:05and overlay these these controls?
- 14:09Yeah. So there are two main components
- 14:11within the the Enfast platform that ensure the the sort of
- 14:16integrity of your content quality.
- 14:19So the first one is basically an automatic sort of mechanism
- 14:24where we automatically look at your current content or previous
- 14:30content. So we looked at your, your
- 14:33previous products and all of the content therein.
- 14:36And then we sort of leverage and extrapolate based on that
- 14:40through a framework, a framework called RAG.
- 14:44It's like an AI framework where it's retrieval augmented
- 14:48generation. So, so by extrapolating on that
- 14:51previous data, we maintain the the authenticity and the tone of
- 14:56voice. However, like let's say that you
- 15:00have a big poor quality as your input data, then then of course
- 15:05we shouldn't sort of extrapolate on on that poor data.
- 15:09So the second component is basically the ability for brands
- 15:13to provide explicit guidelines. So as a brand, you can provide
- 15:18this, yeah, it's guidelines basically where you can set up
- 15:23rules for how you want your different content thesis to be
- 15:27generated. And this is something that that
- 15:31all of our customers use a lot. And it's really, really
- 15:34efficient. So you simply provide rules for
- 15:39like the the attributes level, like for the product
- 15:42description, for example, let's say that you always want to
- 15:45mention the the product material within your product description.
- 15:50And then by providing that guideline, the AI will
- 15:52automatically enforce that across your product catalogue.
- 15:57And what's really cool is that we do it both sort of
- 16:00proactively for all of the new generated content, but we also
- 16:05do it retroactively. So we look through your previous
- 16:09content from the lens of the guidelines, meaning that you
- 16:14will find any inconsistencies that you currently are on your
- 16:16site with this in place. Right.
- 16:22And so one of the things that I know that you've kind of built
- 16:27in is maybe a differentiated some of the other things on the
- 16:29market are like, I guess like some level of auditing
- 16:33capability and then off the back of that you kind of prioritise
- 16:37some of the actions and fixes. Can you just talk us through how
- 16:41that works and why you've kind of built that early on and kind
- 16:44of how that helps to add more value to the Ng?
- 16:49Yeah. So, so I think when we started
- 16:53out with M fast, a big part of what we did is is to do this
- 16:57kinds of audits or or alert based system where we look
- 17:02through your whole content with the AI.
- 17:05So our AI will just go through all of your products, all of
- 17:08your content within and sort of surface any of these
- 17:13inconsistencies with the help of the guidelines that I just spoke
- 17:16about. And as the AI brings these
- 17:20inconsistencies forward for you, you can then manually just go
- 17:24through them one by one if you want and just adjust them
- 17:27accordingly. So the AI sort of gives you
- 17:30proactive recommendations for for the changes that that needs
- 17:33to be done and then you can just act on them in a matter of a few
- 17:37clips. So instead of having to like
- 17:40manually go through hundreds or even thousands of products just
- 17:45to make sure that they are like qualitative, the AI will sort of
- 17:51do the The Dirty work for you. And then you just sort of fill
- 17:55in the blanks and do the final manual work on top.
- 18:01Great. And then I guess the next
- 18:03question. So when you're working with
- 18:05larger catalogues and I think James will have the same, but
- 18:08there's certainly businesses I've worked with that have got
- 18:11massive catalogues and always use product data issues as an
- 18:15excuse for like, you know, not doing certain things or kind of
- 18:18pushing things back. But one of the things you've
- 18:21talked about like a low level so far or like you start talking
- 18:25about a little bit is kind of how you can use rules as part of
- 18:29the automation. Can you talk us through like how
- 18:33you would approach maybe, you know, big business, hundreds of
- 18:36thousands of skews and use the rules to kind of, yeah, support
- 18:42some of this stuff and how you might then have like layers of
- 18:44governance within that? Yeah.
- 18:48So this all comes down to to our sort of guideline engine within
- 18:53the platform where these rules can be applied on different
- 18:58levels of granularity, which means that you can apply rules
- 19:03on the sort of on a on a brand level if you're a multi brand
- 19:07store or you can go down into category level to do it only on
- 19:11on like you know, North Face jackets.
- 19:15But then you can even go further into product level or like
- 19:19specific markets. Even so, some of our customers,
- 19:23they've been able to do market specific content on, on
- 19:27different kinds of categories just to make sure that they,
- 19:30they capture local intent, which is super, super cool.
- 19:35So I mean, one of our customers, they launch about like 100
- 19:39products per day, which I think is a, it's a really high volume
- 19:44where I mean, setting these guidelines up once will give you
- 19:48a really good sort of control centre of ensuring that all of
- 19:52your content will, will maintain the same quality, but also
- 19:57maintain the, the tonality and, and follow all of the different
- 20:02like requirements that you might have on your product catalogue.
- 20:05So I mean our platform works really well on on both fashion,
- 20:10but but I mean even better on on use cases like like what's the
- 20:17word like? Hardware or like maintenance
- 20:23parts, spare parts, you know, technical products that require
- 20:28a lot of different information on them.
- 20:32Yeah, I did. It's funny.
- 20:34Now I look back to when, when I was early days of e-commerce and
- 20:38I was running e-commerce stores for a becoming consultant and
- 20:42one of the biggest pains was there was no technology that
- 20:45would automate quality controls around product data.
- 20:48To have a definition of right, you must have XY and Z for this
- 20:52particular product type. Otherwise that product cannot go
- 20:54live because that's our quality control jet.
- 20:56You'd have to custom build and code logic in order to do that.
- 21:01Otherwise you'd just anything could be put through and you
- 21:04could get, you know, the classic junk in and crap quality.
- 21:07And it's really nice now to see that you could that, that we've
- 21:09got greater democratisation of technology and we have tools now
- 21:12that allow people to be more granular in the controls
- 21:15category specific levels. So that because we all know that
- 21:18some categories, like you said, technical products require far
- 21:21deeper technical specification work and they're a fundamental
- 21:24attribution that you have to have in there.
- 21:25Otherwise you can't sell those products accurately and they
- 21:28won't drive accurate product discovery.
- 21:30So that's yeah, that very interesting to hear.
- 21:32I'm sure there's probably some people listening who will want
- 21:35to tap you up after to find out how they can adapt that to their
- 21:37catalogue. So those listening who are
- 21:38interested, we will obviously have contact details on the
- 21:41landing page. One of the borrow that you talk
- 21:45about on the website with Mfas and we talked about when we were
- 21:48planning this podcast. So there's context aware AI and
- 21:51we love terminology and e-commerce, but that you're able
- 21:55to detect inconsistencies in product data.
- 21:57And you gave me an example. But for those listening who
- 22:00obviously we can't see that on the screen, what the what is
- 22:03this doing? How's it doing?
- 22:05And what is the benefit to merchants?
- 22:07Yeah. So a good example here might be
- 22:12just, you know, a product consists of many different
- 22:16attributes. So you let's say that you have A
- 22:19at shirt where you have one attribute describing the
- 22:22material of the product. And then you have another
- 22:25attribute just which is the the description of the T-shirt.
- 22:29And having a, a context aware AI is, is really important because
- 22:35if we were to generate a description, we need to ensure
- 22:40that the, the material that is mentioned in the, in the
- 22:44description matches the material that comes from the material
- 22:48attribute. So if it's 100% cotton in the
- 22:52material attribute, then the AIAI will make sure that you
- 22:55know, does it, does it contain like 100% cotton in the actual
- 22:59description? And this sort of context aware
- 23:05AI generations is, is really key to, to our platform because if
- 23:10you were to just generate descriptions or, or different
- 23:13things with the use of AI in isolation, just generating a, a
- 23:18random description for a product, then it won't be
- 23:21context aware. So, so you need to always give
- 23:24it the sort of full context in order for the, the content to
- 23:27be, to be of good quality. And I think that the benefit for
- 23:32merchants here is that instead of having like, I've seen a lot
- 23:37of like really bad use cases where you have sometimes a
- 23:42proofreader, like someone who, who goes through all of the
- 23:45production ready like products. Like in some cases, we've even
- 23:49seen the CEO of the company sort of proofreading the, the, the
- 23:54products, which is not really good use of, of the CE OS time.
- 23:59But what this enables you as a brand is this reduces manual
- 24:03labour, it reduces the time to market and it further improves
- 24:08your SEO like overall visibility online and and in the end
- 24:14conversion rates right by having a good information on your
- 24:19products. Yeah, excellent.
- 24:22And for people with multi store, we'll get onto the international
- 24:27bit next, because I know that you support translation, but
- 24:30translation means different things to different businesses.
- 24:33And in some languages, automated translations don't work so well
- 24:37because they're not. They don't have the context of
- 24:39the language and the direct translations don't always
- 24:42resonate. German's a great example.
- 24:45So can you talk us through how do you hand, how does Zenfus
- 24:48handle it? What translation engines are
- 24:49used how how do you handle automatic automated
- 24:52translations? Plus how can businesses like
- 24:56control it directly themselves to bring in like human based
- 24:58translations where they need to? Yeah, precisely.
- 25:01So I think we are firm believers that AI is sort of improving the
- 25:09the productivity here for e-commerce brands.
- 25:11So within our platform you are able to translate into any
- 25:15language with just a matter of a few clicks.
- 25:19And, and, and this works like straight integrated with your
- 25:23e-commerce platform like Shopify or or Sentra or the likes.
- 25:27And the way we do it this we, we enforce both the sort of, as I
- 25:33mentioned previously, the automatic mechanism where we
- 25:37looked at your previous content and if you have written in
- 25:40German previously, then you will maintain your German tone of
- 25:43voice. But if you haven't, we also
- 25:47leverage the the use of your guidelines.
- 25:50And what a lot of our customers do is they provide explicit sort
- 25:56of translation rules. So if you want to translate any
- 26:00specific word or paragraph in, in a certain way, then you can
- 26:06just provide that and the AI will sort of follow that
- 26:08taxonomy, which is really handy if you have a, a certain way of,
- 26:12of writing out in different languages.
- 26:15But then like, also like AI is, is far from from being sort of
- 26:22on par with just human intelligence.
- 26:26Of course, we're not, we're not delusional here.
- 26:28So, so I mean the, the AI will sometimes just do 95% of the
- 26:34work. And as a brand, if you want to
- 26:37have like really, really high quality translations into
- 26:41another language, one option that we provide is you can
- 26:45generate like all of these translations automatically and
- 26:48we do 95% of the work. And then as a human, you can
- 26:53come in and just oversee the translations within the
- 26:56platform. So that way you can sort of have
- 26:59a human in the mix while having the AI do most of the work for
- 27:05you. But then again, if you're a
- 27:08customer that that like you're satisfied with 95% accuracy,
- 27:14then simply working with translations fully driven by AI
- 27:18is is definitely possible. And then my last question, Sir
- 27:25James has got one more in a second.
- 27:28But a lot of people listening to this or a lot of the kind of
- 27:30brands listening to this, they'll be working with
- 27:33spreadsheets or I'm sure you're familiar with my trip to find
- 27:36the Shopify world. And you know, there'll be a lot
- 27:40of kind of manual work around that.
- 27:41And, you know, there's obviously a lot of risk.
- 27:44You've obviously worked with I think Akinio before Oscar, and
- 27:49you're very familiar with Pyms. Where does Mfas sit in this
- 27:54world? Like I think there's probably
- 27:56recently maybe I think there's potentially less demand for PIM
- 28:00than there was three years ago as a lot of the platforms got
- 28:03stronger in this area. Working with spreadsheets has
- 28:06got easier. And again, the platforms have
- 28:08made that better. But where does Mfas sit?
- 28:10And yeah, how does it kind of drives benefit within the
- 28:15platform you should work. Yeah, I think this is really
- 28:19something that we, we worked a lot with where our platform
- 28:24works really, really well with your existing e-commerce stack,
- 28:29meaning that integrating and fast is simply just, you know,
- 28:33it's like an app in, in the Shopify App Store or we just did
- 28:37an API key to your e-commerce platform.
- 28:40So there is no need for like developing or, or any developers
- 28:45at all. Actually.
- 28:45It's, it's really easy to get started.
- 28:48And I think what we're looking at here is you have this
- 28:51spectrum where on one end you have the traditional Pims, which
- 28:57often times requires like hundreds of, of development
- 29:00hours to, to sort of build out and to integrate with your
- 29:04existing stack. And then on on the other hand of
- 29:07the spectrum, you, you also have this new solutions that could be
- 29:12like air table based beams or really like slim and, and
- 29:18customizable solutions where we are sort of trying to, to be
- 29:24the, the solution in the middle, where we leverage all of the
- 29:29like good features of AI and we integrate directly with your
- 29:34e-commerce platform. So you can leverage those ones
- 29:36instantly without requiring to, to implement like or, or
- 29:42integrate the a lot of consultant hours.
- 29:46So we're sort of bringing value directly to the table for, for
- 29:49e-commerce brands. And then again, as we're not
- 29:54this air table solution where you have to, to build out the
- 29:57product model from the ground up, we give you tools to
- 30:01instantly work on your product catalogue within a very easy to
- 30:07use and good UI, which I think is is really key to be able to
- 30:11as a non-technical person work with the the technology of AI on
- 30:17your catalogue. And lastly, I would say that
- 30:21like in general, if, if you are on, on, if you are working with
- 30:25spreadsheets and you're on Shopify or or Centra, then it's,
- 30:30it's really a no brainer to to just at least try out like using
- 30:34Mfas just because like if, if you're not using AI to enrich
- 30:39your product catalogue today, you're really missing out.
- 30:43So we're trying to like embody, like the whole work smarter, not
- 30:48harder mentality here, which is, yeah, really working well for
- 30:53for a lot of Igmos brands. Yeah, I'll be intrigued to to
- 30:57hear actually what, what is the key feedback you're getting
- 31:00because you're you've progressed rapidly for, for basically what
- 31:03is a start up. The product maturity is quite
- 31:05impressive and you've got a lot of big brands already working
- 31:08with you. So I'd love to hear what are the
- 31:11things that how are they most using it?
- 31:13What are the things that they're getting the most value from?
- 31:15Because when I look to the Branisha gave me a quick demo,
- 31:17the bit that I would I really liked was the the dashboarding.
- 31:22Because a lot of times in in it, like in in early stage
- 31:25start-ups, you get the raw product, but you don't get that
- 31:27more maturity around like the dashboards, the alerts and that
- 31:30ability to see at a glance where you've got errors in data
- 31:35inconsistencies, where you've got those specific parts of the
- 31:37cattle and be able to pinpoint and then prioritise.
- 31:39Because as a yeah, as a merchandiser, I know that
- 31:42category A is my killer category.
- 31:44I can see I've got 30 errors in there.
- 31:46That's where I'm going to focus. So that to me resonate really
- 31:50well. But what feedback are you
- 31:51getting? What are people using it the
- 31:52most for? Yeah.
- 31:55So it's been a really like fast-paced journey so far.
- 32:00And when we started out, we like quite quickly realised that a
- 32:05lot of brands they like, we like in the beginning, we pitched the
- 32:10use case of being able to just, you know, generate a lot of
- 32:13copywriting on product level and translating it.
- 32:17But as time progressed and as we sort of take taking a look at as
- 32:22more and more e-commerce brands and, and their product
- 32:25catalogues within, we have been able to solve a lot of a bigger
- 32:30chunk of the different product enrichment processes.
- 32:34So basically going from just enriching the the descriptions
- 32:38and now into being like a fully fledged like we, we solve all of
- 32:42the attributes on your products. So one thing that you might
- 32:46notice here is that we are actually kind of becoming or or
- 32:50taking over all of the workflows from the traditional PIM, which
- 32:54is definitely I think the direction that we are sort of
- 32:57heading towards where we are basically replacing a lot of the
- 33:04traditional workflows that that were previously just inputted in
- 33:09in this big forms essentially, which are the the traditional
- 33:12teams. And we are now replacing that
- 33:16with the use of AI, sort of making you 10X more efficient as
- 33:21an e-commerce person without having to to pay these super,
- 33:26super hefty fees of of traditional Pims.
- 33:29So I think the future of e-commerce will will look very
- 33:32different in the coming five years with the use of AI and
- 33:37alongside with the the roll out of visual data within the Enfos
- 33:41platform, we will sort of become the we would become the best of
- 33:45breed product enrichment platform for both like text
- 33:49based attributes and visual. So it's a, it's a bold aim and I
- 33:54like it. I mean, you, you need to, you
- 33:55need to be bold. But I, I think the key, the, the
- 33:58key point about this, the reason we're doing this episode is that
- 34:01me and both Paul and myself have seen the fact that the, the, the
- 34:04industry changes. Paul alluded to earlier that
- 34:05fewer and fewer businesses are looking at Pyms if they don't
- 34:09already have a legacy one in place.
- 34:11And I, I go, I go back three years and I had a lot more
- 34:15projects where PIM was a fundamental discussion point in
- 34:18the tech stack as part of a real platform.
- 34:20Him, I've not seen that in the last couple of years come up
- 34:23once. Whereas as you've mentioned VDR
- 34:26Air Table has come up and there are some really smart people
- 34:29like Ollie Rose, who's doing a wonderful job building a
- 34:32consultancy around Air Table and it suits certain businesses and
- 34:35it solves certain problems very, very well.
- 34:37It can, but it doesn't fit in every business model.
- 34:40And it's yeah, it's interesting to see like a modern tool that
- 34:45is tackling the product data quality issues.
- 34:47And I like the fact that that that the the the futures would
- 34:51be around visual as well as like the the, the written product
- 34:54catalogue data as well. Interesting.
- 34:57So if anybody wants to find out more or wants to challenge you
- 35:00on any of this or wants you to put your money where your mouth
- 35:03is and show them the products and can actually see them, how
- 35:05do they reach out V dot? I think best way is simply to go
- 35:10to our website mfas dot AI and sign up for a demo or just
- 35:15contact us directly there. Otherwise feel free to contact
- 35:19via LinkedIn. That's also very much welcome
- 35:22and happy to answer any questions.
- 35:25Fantastic, let Vida always love chatting to you.
- 35:27Looking forward to seeing the story develop and I hope
- 35:30everyone listening has found interesting.
- 35:32We like to bring you new, new tech stories where we think
- 35:35they're really interesting to the E Comm industry.
- 35:37So do reach out to EDAR if you're on a program anymore or
- 35:40feel free to reach out to me and Paul if you want to discuss
- 35:43anything around the product data challenge.
- 35:44Thanks for listening as always. Thanks.
- 35:47Bye. You're listening to the Inside
- 35:50Commerce podcast, hosted by experienced consultants James
- 35:53Gerd and Paul Rogers. Inside Commerce is your fast
- 35:56track to digital knowledge, featuring interviews with
- 35:58leading e-commerce retailers, industry thought leads and the
- 36:01C-Suite, key tech companies to explore the latest digital
- 36:04strategy. Customer experience and tech.
- 36:06Discover what the latest trends really mean and get practical
- 36:09advice based on the experience to help fine tune your
- 36:11e-commerce. Strategy and tech stack.
- 36:13Follow us on LinkedIn and reach out to James and Paul if you'd
- 36:16like to discuss an e-commerce project.
- 36:18We hope you enjoy the latest episode.