Latest / Inside Commerce: Ecommerce Strategy, CX and Technology Podcast / EP265: Delivering A Smarter CX: Integrating AI Chatbots Into Onsite User Journeys To Improve Product Discovery, with Asklo
Transcript
- 0:02You're listening to the Insidecommerce podcast, hosted
- 0:05by experienced consultants James Gerd and Paul Rogers.
- 0:08Insidecommerce is your fast track to digital knowledge
- 0:10feature in interviews with leading e-commerce retailers,
- 0:13industry thought leads and the C-Suite of key tech companies to
- 0:16explore the latest in digital strategy, customer experience
- 0:18and tech. Discover what the latest trends
- 0:21really mean and get practical advice based on experience to
- 0:24help fine tune your ECOM strategy and tech stack.
- 0:27Follow us on LinkedIn and reach out to James and Paul if you'd
- 0:30like to discuss an e-commerce project.
- 0:32We hope you enjoy the latest episode.
- 0:35Thanks for tuning in to the Inside Commerce podcast
- 0:37sponsored by Centre and grabbing more about them later.
- 0:40Got another interesting topic this week.
- 0:42I say it every week, but I generally do love the
- 0:44conversations we have on this podcast.
- 0:47So speaking to a technology company that I've known for
- 0:50years, who I really rate and have worked with on several
- 0:53client projects. So Clayvoo, who a lot of you
- 0:57know from product Discovery Engine, have a new product on
- 1:01the market called Asclow. And I'm going to let the the the
- 1:04founders of the business and CEO and CTO tell you more about it.
- 1:07But the topic is around how can you deliver smarter customer
- 1:10experience online and integration AI chat bots into on
- 1:13site user journeys to improve user outputs, but also to bring
- 1:16efficiency to businesses to automate processes, which
- 1:19historically would have taken a lot of human efforts to do so
- 1:25rather than me try and do the like elevator picture what
- 1:29Arslow is. Neil, let's let's you're giving,
- 1:32please. Can you give a bit of an
- 1:33instruction to our listeners? Some will know you, some won't
- 1:35who you are, what you do. Then do you want to position
- 1:37Arslow? And then we can come and do it.
- 1:39We'll do the intro to you, Niraj, when we come into your
- 1:41your first question. Sure.
- 1:43Thank you so much, James, for having us.
- 1:47Yeah, my name is Neilo, one of the founders at CLOW and Oslo is
- 1:51a is a very interesting product. It's sort of part of our larger
- 1:54mission of you know, democratizing discovery.
- 1:59That's what we started with and our goal has been to improve the
- 2:02site right conversion from, you know, your every stores and when
- 2:07you about improving site right conversion and we focus
- 2:11exclusively discovery as our core focus, every touch point of
- 2:16the discovery, we want to optimize.
- 2:18So we we did both search collections recommendations and
- 2:22then we found that OK, product pages, you know, they lose 30%
- 2:26of the visitors that come from directly from marketing paid ads
- 2:29or social ads, you know, bounced.
- 2:32It is as high as 30%. So you lose that, you lose the
- 2:34insight. It's like a dark art for
- 2:36marketeers, right? The customer was on my product
- 2:39page and he or she left. But why is it just pricing that
- 2:44we assume by default? Why actually customer left?
- 2:49Can I somehow go back to the customer?
- 2:51So we had these conventional chat boards that try to, you
- 2:54know, help the customer. We have clearviews of the world
- 2:57who tries to text the e-mail and then go back to the customer.
- 3:00But why? And can we actually capture the
- 3:03customer at that moment itself? And that's where we came across
- 3:07FAQ section. So that product page is getting
- 3:10messier. They have a lot of information
- 3:13and shoppers really don't kind of pay attention to it.
- 3:16And this is this is where we kind of, you know, thought that
- 3:19OK, how we can build the FAQ kind of to the modern age, which
- 3:23is like a true AI age. And that's where you have the
- 3:26kind of the element of conversation also embedded into
- 3:30it. And that became ask low.
- 3:31So in nutshell, what it does is that it it adds little widget on
- 3:35your product page and then it's a live FAQAI.
- 3:38It takes care of kind of generating the questions.
- 3:41And if the shopper has any particular question, they can
- 3:44also ask that and it takes care of answering it.
- 3:47And then there's a back end panel where you can actually see
- 3:50the whole product and the performance and yada yada.
- 3:54Yeah, I I really like this idea and and for those of you
- 3:56listening, we are going to get into more like more depth about
- 3:59what how it works, the AI, the benefits and also some case
- 4:02studies. So do stay tuned to listening to
- 4:04a couple of really interesting ones.
- 4:06What I love about this is having been in e-commerce for many
- 4:09years now, I'm like one of the old men of E com now.
- 4:13I remember when, when like Q&A stuff on PDPS became really
- 4:18popular. But actually the biggest
- 4:20challenge was scaling that when you've got 10s of thousands of
- 4:23skews, you know, hundreds of thousands, if not millions of,
- 4:26of visitors and it becomes unmanageable.
- 4:28And a lot of people just switched it off.
- 4:30It was unmanageable, but the content was really useful.
- 4:34So I the ability to generate this at scale.
- 4:38And I'm really, This is why I wanted to have this conversation
- 4:40because I think a lot of people listening will find this really
- 4:42interesting. And, and when we get into the
- 4:44case studies, go and try it out because I had that classic.
- 4:47How good is the response? Is it really accurate?
- 4:49And there's a couple of sites where where the quality of the
- 4:51response was, it wasn't just a felt like a butt, in fact, like
- 4:55someone actually properly answering questions.
- 4:58So, So what I guess a question, one more question for you,
- 5:04Nilay, before we get to to Naraj on this is, is business cases
- 5:09because there are we talked about something.
- 5:11There are so many AII tools out there and it isn't a hype cycle.
- 5:14We all know it. There are people pushing AI
- 5:16tools that might not resever any specific underlying need, so how
- 5:21can people find a genuine use case for this?
- 5:24Greben create digital flagship stores for iconic brands of
- 5:28today and tomorrow. We are the number one Sentra
- 5:31agency in the world with more than 50 Sentra implementations
- 5:34to date. We are also e-commerce design
- 5:37specialists, platform and tech partner agnostic.
- 5:40Try our new e-commerce design subscription.
- 5:42Visit greben.com. Yeah, excellent point actually.
- 5:46And it's, it's true. And, and, and I think, you know,
- 5:50like in clear, we tried to always develop anything that has
- 5:53a clear kind of in a use case and conversion value.
- 5:56But I, I think in general for retailers, it's sort of, there
- 5:59are so many applications everybody claims about AI.
- 6:03So how to kind of, you know, cut through that noise.
- 6:06And as you rightfully said, there are so many tools also
- 6:09available. I think one of the things or,
- 6:11or, or some of the things actually, I would kind of really
- 6:14focus on as merchant, if I ever to choose is to focus on
- 6:18outcomes and what actual outcome this AI tool is going to produce
- 6:23for my shoppers or for me, how they perform.
- 6:27Don't just take it like what says on the tin.
- 6:30You actually test it on your own and if possible, test on the
- 6:35real user data and not just, you know, what you think.
- 6:40Because if you think from the conventional applications point
- 6:43of view, you already had a lot of user test data available for
- 6:47some of these new tools, you don't have actually right type
- 6:51of user data available. So it's very important to
- 6:54actually test it in the live user environment and understand
- 6:57that is it really working or not.
- 7:01One of the best ways that is I learned from these actually to
- 7:04test any application is how it performs.
- 7:07When the answer is no, when and there is more information, how
- 7:14does your system perform, how graciously it fails?
- 7:20Does it admit that it's failing? I think so you when you try to
- 7:24say, in software engineering we used to say exception handling.
- 7:27So I think in the same for AI tool, I would say that you try
- 7:31to find these exceptional cases, use cases or edge cases and
- 7:35understand how it performs. The idea is not that whether
- 7:38it's getting right or wrong, it's about how it actually
- 7:41performs in those settings. Will those responses, will those
- 7:45outcomes be acceptable to your shoppers or not?
- 7:50That's the key thing in my view ready to AI applications.
- 7:55I really like that failing gracefully.
- 7:57And I think, I think that's couched in, in the whole
- 7:59customer service, customer care language as well because it's
- 8:03not about technology then it's about people and you've got to
- 8:06cope for those, those education. I like that.
- 8:08And we're going to get into a bit more detail.
- 8:10So Niraj, before I start asking you large questions, please
- 8:14introduce yourself to our guests.
- 8:15Let them know what your role is as well.
- 8:19Yeah, thank you. Thank you very much James for
- 8:20having me here. I'm one of the Co founders and
- 8:24CTO at Gabu. My background is actually in the
- 8:28field of AII. Started my career about 20 years
- 8:32back when role based systems were very popular.
- 8:37But we have seen now things changing significantly.
- 8:41We see deep learning, we see introduction of large language
- 8:44models and more importantly it becoming available to everybody
- 8:50whether you are an LP scientist or not, but you are able to make
- 8:54use of that. Yeah.
- 8:56So that's my background. I've seen that journey and
- 8:59that's what we have been trying to use every latest technology
- 9:07that's available at Cleo. Yeah, I, I love it when you
- 9:11speak to people who've who've been working with things like
- 9:14machine learning AI for many, many years.
- 9:16It's quite amusing health. There are some who are
- 9:19pretending it's like a new thing as well, which is because it's
- 9:22not. But you're right.
- 9:23The the point you made, the democratisation of this about
- 9:26the accessibility to even small businesses now is amazing.
- 9:29So with that in mind, let's get into this.
- 9:32Because PDPS FAQs giving people relevant information to help
- 9:37take away barriers to purchase. I think any retailer listener
- 9:40will be like that sounds perfectly good sense.
- 9:42How is the AI actually working to generate these answers?
- 9:46Centra is the only commerce platform built exclusively for
- 9:48fashion brands. Over 300 fashion brands run on
- 9:50Centra, including icons like Paul Smith, Jeff Ave. and Nudie
- 9:53Jeans. Central solves business logic
- 9:55challenges for fashion brands. I know of the platform and they
- 9:57are on a mission to help you scale up rapidly globally across
- 10:00all online and offline channels. Take your fashion brand global.
- 10:03Faster, further, stronger today. Yeah, so seems I'd like to start
- 10:10with seeing how the user lands on the page on the website.
- 10:16There are multiple ways a user can reach to your PDP.
- 10:21It may be through the search box that we have up there.
- 10:23They may be directly landing from a Google or for that matter
- 10:28any search engine website. Our objective as Nilay said is
- 10:33to make sure that you know customer finds why they have
- 10:39landed there. So the Oslo application that we
- 10:43are talking about here today, it's a FAQ application which is
- 10:49allowing you to question the details about the product
- 10:55itself. But in our road map, what we
- 10:58have is that as Neil I said, when you ask for something that
- 11:03the answer is no, how can we suggest something else to the
- 11:08customer which is actually fulfilling the customers
- 11:12requirement there. So the technology that I'm going
- 11:16to talk about now is not about only FAQ sections which are
- 11:21available today in the market, but that is also taking into
- 11:24consideration what's we, what we have in the road map in the
- 11:28coming day. Now when we talk about the AI
- 11:32technology behind this, there are a couple of buzzwords. 1 is
- 11:37RAG and that is retrieval of maintained generation.
- 11:43That's something that we are using behind the scene here.
- 11:47But what it means in simple terms is that earlier we had
- 11:51search boxes, somebody would come there, they would think
- 11:55about the keywords that can guide them, the most relevant
- 11:58products. But with the introduction of
- 12:01large language models, it is allowing somebody to express
- 12:04what they want in the language they want to choose.
- 12:09It can be English, it can be the native language as well, it can
- 12:14be the actual product that they are looking for, or it can be as
- 12:20simple as the requirement that they are after.
- 12:25What large language models have made it possible is that it is
- 12:29able to understand the actual requirement and it is able to
- 12:33translate that into what we are looking for.
- 12:36The way Reg applications work is that they take this query into
- 12:43account. They fired that against our own
- 12:47merchant catalog database. They try to find out which
- 12:51products are more relevant for that query and they try to
- 12:54deliver the answer in the natural language, in the
- 12:57language that customer had used over there, right?
- 13:01So Reg is what we are using behind the scene.
- 13:06But the very important point to to mention here is that what we
- 13:10have developed here is far from the standard Reg applications
- 13:15that we see available in the market.
- 13:19It's very advanced behind the scene.
- 13:22It's not only the vector database that we are using, but
- 13:25behind the scene it is actually using all that we have done in
- 13:29last 10 years with Clebu search. So it is using the Clebu search
- 13:33power there to find out which products are more relevant to
- 13:36the question and it uses that as a context combined with large
- 13:42language model to provide the right answer.
- 13:47Yeah. I thought I find this really
- 13:48interesting. And this leads on to the next
- 13:51question quite nicely because going back to when Q&A became
- 13:56important and that you would have like a customer can submit
- 13:59a question, then a customer service team would type it in,
- 14:02you know, bizarre voice in the early age were rather the first
- 14:04to launch it. the IT relied on so much knowledge in specific
- 14:11teams where you had all this knowledge built up over the
- 14:13years and it wasn't always well documented.
- 14:16That's so I guess a key question.
- 14:18This is and, and retail and ecom leaders out there will want to
- 14:22know this like what depth of product information is required?
- 14:25What other information does Osclo need to make sure that
- 14:29it's able to deliver a quality of response versus how much is
- 14:32it able to to to use AI to to generate and infer information
- 14:39even if it's not explicitly available?
- 14:43Yeah. So I think to understand that we
- 14:47have to look into what kind of questions today we are seeing
- 14:50that customers are asking within Jet, within our chat interface.
- 14:55So of course they are coming on a product page.
- 14:57They want to know about whether the product is meeting their
- 15:02requirement, which means the usability.
- 15:07And also so that is if the product is perfect for their own
- 15:14use case, then they would want to know about the availability
- 15:17of the product. Then they would also want to
- 15:20know about things like the the pricing.
- 15:26Often they may want to negotiate the price as well.
- 15:33Yeah. And how soon it it can be
- 15:35delivered, for example. So these are the different kinds
- 15:39of queries that we are seeing. What that means that product
- 15:45data is required of course. So whatever product catalog that
- 15:49we receive from the customers is something that we are indexing.
- 15:52But together with that, there are certain elements which are
- 15:55dynamic. For example, if I'm logging on
- 15:58to a website from from particular location, that
- 16:01product may not be available to be delivered to me within a
- 16:04certain time frame that I'm expecting.
- 16:07So this information becomes very dynamic.
- 16:11So we are able to collect this information from the from the
- 16:14page directly for that individual user and utilize that
- 16:19to actually answer the question as well.
- 16:24Yeah. So in short, it's not only the
- 16:26product data, but whatever data is required to answer this
- 16:30Different types of intents is is what we are also using behind
- 16:34the scenes. Yeah, makes sense.
- 16:37Neil, I, we, when you were initially showing me this and I
- 16:41got really excited by the whole prospect of it, one of the
- 16:44questions I had was because you, you have different products in
- 16:48the Clayvoo family. You've got your product
- 16:51discovery engine, which a lot of people know about.
- 16:54You've got Moi, which is I guess an AI assistant chat bar and
- 16:58you've got Arsenal, which is focused on one specific area at
- 17:00the moment they're coming back to that failing gracefully.
- 17:03There will be times when it can't find an answer.
- 17:04There isn't either it can't create one, it doesn't have the
- 17:07knowledge or there isn't the content there to display it.
- 17:11And that's when a, a brand will say, right, how do I automate
- 17:14pushing them into my customer service flight?
- 17:16And maybe it's pushing them into a request form to submit an
- 17:19enquiry to follow up or So what? Where are you going with this?
- 17:24How can brands hand over inquiries when the customer
- 17:26actually needs to speak to some of these more expertise?
- 17:29Yeah. So we experimented two different
- 17:32kind of, you know, methods if you like.
- 17:34So 1 is like this embedded widget method, which is a live
- 17:37FAQ, what we call it. And the other method we had was
- 17:40ask product AI. So you have a button coming on
- 17:43the product over the product title and then you click on a
- 17:46button and you start, you know, typing your questions.
- 17:49Now in that method, we found that people were more actually
- 17:54even though whether the answer is there or not, but there were
- 17:57more inquiries going towards that.
- 17:59OK, I want to talk to a human agent because they were seeing
- 18:02that more as a chat bot. But we made this as an embedded
- 18:05widget as part of the product is itself that became like a live
- 18:09FAQ. We are seeing less of that, but
- 18:12we, we are actually and many chat bots now have become modern
- 18:18like Zendesk and, and core GS and all these platforms have
- 18:21become more modern who are supporting the handover
- 18:24mechanisms as well. So so you could actually
- 18:27transfer the handover to them. I'm still struggling to
- 18:31understand that because the experience would still be
- 18:34different. And for the from the shopper
- 18:36point of views, I feel that seamlessness in the experience
- 18:39is very important. And hence, I would say that
- 18:43these chat companies could use these kind of API within their
- 18:47experience itself so that as a shopper you are not confused,
- 18:51you know that, OK, you understand what I mean.
- 18:53Like if you're completely different experience on the
- 18:55chat, then then as a shopper, you may just leave the store as
- 18:58well. When we also are getting
- 19:00customers who are saying that, look, I want AI to answer
- 19:03everything so that expectations are rising up.
- 19:08And they said, OK, if there is a real edge case where AEI cannot
- 19:12answer you create a ticket you send to us and we will answer
- 19:16the question to the customer. These are of course more
- 19:20relevant to those categories where response is not very time
- 19:23sensitive, but it's it's also emerging now because early
- 19:27again, I think if you think about like two years ago, the
- 19:30level of response human like responses that you can give you
- 19:35know is is just changing. I think it's rapidly evolving to
- 19:39a point where you can almost actually provide human like
- 19:45response. So as long as the contextual
- 19:49knowledge is available to the system, you can actually, you
- 19:53know, go pretty far now. Where I I the future of this is
- 19:59fascinating for me because the the user journey, but you're
- 20:02100% right is you can't just have push from one bot to
- 20:05another bot. I don't, but actually using the
- 20:08APIs and you're you know whether it's gorgeous or Zendesk have
- 20:12the API ability to to trigger tickets.
- 20:14Correct. The ability to do that.
- 20:16But I guess I have the logic of what you can trigger a ticket
- 20:19for and be able to fire that. So you're in, you're in the
- 20:22Ascalo journey and you have the ability.
- 20:24Do you want, would you like to hear back from customer service?
- 20:27Yes, OK, give us your name, your e-mail address, etcetera using
- 20:31the the the standard form. I think that that will be a
- 20:34really nice and smooth journey. Yes, absolutely.
- 20:39Yeah, you're looking forward to seeing the the the next
- 20:41iterations of this. So let's let's come on to the
- 20:44user intent signals this scenario.
- 20:46I think it's really interesting. We did a podcast a few months
- 20:48back with Made With Intent, David Mannheim, who's the
- 20:52founder there, and they're all about just doing all the data
- 20:56modelling around user intent signals to pass on to other
- 20:58systems, for example, yours. I'd love to understand how you
- 21:01do So. What signals are you using to
- 21:03understand user intent and what what are they used for?
- 21:11Yeah. So I think the most important
- 21:15bit is what they have asked so far, understanding that past
- 21:22history of all the questions and it's not about the questions
- 21:26they have asked on this particular product itself.
- 21:31We have to understand what they have been asking, the kind of
- 21:34the questions they have been asking on the other products as
- 21:37well because that gives us some information like where this
- 21:41customer is coming from. The second thing is what type of
- 21:46products they have been visiting on the website itself.
- 21:50For example, price is something that you can easily identify the
- 21:55price range for this particular customer, what price range they
- 21:59are looking for, the kinds of the product they are looking
- 22:02for. Yeah, so that would be the
- 22:07second bit. Also, the advantage that we can
- 22:12bring to the customers who are using Claboo Search already is
- 22:17that we allow merchants to actually merchandise their own
- 22:24store whereby they want to promote certain types of
- 22:28products as well. So when it comes to recommending
- 22:31products in our session, we are able to not only bring the
- 22:37personalized products to the customers to to their
- 22:40preferences, but also we are able to look into what merchant
- 22:46wants to promote. So that's also something that we
- 22:50track. Finally, there's something we
- 22:54call self learning as part of clay whereby we are identifying
- 22:57the trend on the website. So whatever is actually in
- 23:02trend, whatever the other customers and that's where the
- 23:05collaborative filtering actually comes into playing its role.
- 23:09It is telling us what is entrained on the website.
- 23:12So that is also something we are utilising here when it comes to
- 23:15recommending the the suggestions to to to the shoppers.
- 23:23Yeah, I like, I like this because if I understand
- 23:26correctly, because at the moment you're focused on on answering
- 23:28questions, but you're launching product reqs.
- 23:31Soon. Can you talk us through a bit of
- 23:34that actually? Because I think again, this is
- 23:35another thing that will be really interesting to ECOM
- 23:37merchandisers, is that is it going to be, you know, will
- 23:42there be rules where people can control what products are
- 23:44recommended? Is it all going to be automated
- 23:47using the AI? And what?
- 23:49Yeah. Is it going to be personalized
- 23:51based on just in the way that Claview has personalisation?
- 23:54Will personalisation be part of those wrecks in Osco?
- 24:00And since you are absolutely correct, so today when we talk
- 24:04about the search experience, we are actually monitoring all the
- 24:07activity on the website by that customer.
- 24:10And accordingly we are bringing products not only in the search
- 24:14but also on the category pages and also in the recommendations.
- 24:18The similar or the same technology will be used here as
- 24:22well where we will identify what kind of products the customers
- 24:27are interested in. Earlier personalization was
- 24:31based on the kind of the the search terms that customers have
- 24:36used, the kind of the products that they have clicked on.
- 24:39But now you can imagine they are actually interacting with the
- 24:42application. So they are actually telling us
- 24:45a lot more about their own requirements.
- 24:49So definitely personalization will be the key here.
- 24:53We want to show very relevant results to the customer.
- 24:58But when we talk about the rule based kind of thing, there are
- 25:02two things I want to highlight here #1 whatever merchants are
- 25:05merchandising behind the scene, they may want to promote certain
- 25:09brands or they may want to promote certain types of
- 25:12products on the website. So yeah, that will be also taken
- 25:16into consideration. But there is something called
- 25:20custom intense that we are going to today.
- 25:24It's possible Merchant is telling us that, hey, if the
- 25:28question is about providing certain types of promotions,
- 25:35please answer it this way. But this is something we will be
- 25:39bringing into our something we call Asklo Hub, where merchants
- 25:45will be able to tell us how to answer certain questions as
- 25:49well. So that's also something that we
- 25:52will be considering when we provide answers to the
- 25:55customers. Excellent.
- 25:58Yeah, that's really interesting. And so Nilay, rather than me
- 26:03asking this question about whether it can be used with
- 26:07other product discovery engines, I know that you said it is.
- 26:10So you don't have to be using Clayber to use Arsenal, right.
- 26:13You could use it with if you have a different product
- 26:15discovery engine. The intention is Arsenal
- 26:17different it will take. Nicely.
- 26:19Yeah, I really like that actually because really smart
- 26:21way to do that product firmly because it can enhance what
- 26:25you've already got to play with, but it doesn't limit people if
- 26:27they are using something else. Got it.
- 26:30Yeah. So let's skip to the the really
- 26:33interested bit. We, I, I teased it earlier to
- 26:35people about case studies you showed me, but obviously people,
- 26:39most people are listening on a player rather than watching a
- 26:42video. We're not doing screen shares.
- 26:45So can you talk us through what view is the best current
- 26:47implementation like how it's working and what what impact
- 26:51it's had for the retailer? Yeah, sure.
- 26:56So I'm going to talk about, you know, 11 recent case study that
- 27:00is travel pro. It's a Canadian kind of you know
- 27:04high end luggage store and you know they they have currently
- 27:11implemented this on European storefront and they saw in about
- 27:15one month usage time they saw 30% reduction in bound straight
- 27:20from PDP itself. They also saw that about 10% of
- 27:28their entire revenue was actually driven by ASCO, meaning
- 27:33how we measure is very conservatively.
- 27:37So those who actually asked question, not just AI generated
- 27:41question, those who actually asked some question and then
- 27:44they bought the product as well, the same product that they asked
- 27:47question for. So we use that information and
- 27:52about 43,000 users actually interacted with Asklo widget in
- 27:59this period itself. And these are unique users
- 28:03interactions that we had on a larger scale in.
- 28:07We have now so far answered about 2250 thousand questions
- 28:13from different users and about maybe 1.5 to 2 million different
- 28:19sessions that we have already captured so far overall.
- 28:23And we are seeing about 5000 questions per day on average
- 28:28being asked by users on on this system today.
- 28:33Yeah, I see. That's impressing.
- 28:35And I actually, I Rd. tested it because I'm a, I'm a typical
- 28:38consultant. I listen and then I go, does it
- 28:41really work that well? And then I have to go look at it
- 28:43myself. My wife laughs at how deeply
- 28:45cynical I am in life. But yeah, when, when I looked at
- 28:49it, I asked some, I asked some basic questions and I'd expect
- 28:52it to obviously give it give some, some standard answers.
- 28:54But what I really liked is when I asked some of the somebody
- 28:57like, you know, my children, my children are clumsy.
- 29:00If if they drop it, is this going to dent badly?
- 29:02So, you know, pretty, pretty natural language, not a really
- 29:06like T word focus thing. And the, the consideration in
- 29:11the response was really smart. It articulated why it was going
- 29:15to be protected from dentin, the specific facets and attributes
- 29:19of the product that would enable it to be protected.
- 29:22A really, really like nuanced answer rather than just yes, it
- 29:26will be protected and that I really liked.
- 29:29So I would I would suggest anyone go and have a look on
- 29:32travel prior and ask ask it some tough questions.
- 29:35Actually you, you are right, one of the in one of the electronic
- 29:39consumer appliance stores, somebody asked that about, it
- 29:42was about the fridge and somebody else said can my kids
- 29:46hang on the door because they usually do, you know, and, and I
- 29:50think this is what I meant earlier that, you know, test the
- 29:53edge cases. Yeah, it doesn't mean that it
- 29:56needs to answer perfectly, right.
- 29:59But test the edge cases to understand how actually it
- 30:03performs. So I think it's it's very
- 30:07important to to do that. Yeah, 100%.
- 30:11I, I, oh, it's always good. I mean, going through a re
- 30:15platform at the moment and you'll know because you've been
- 30:17involved in so many projects where you are part of the text
- 30:19that's being built and you do all the core user journeys and
- 30:23it's great. We're all happy.
- 30:24And then you, you mess around in the way that some crazy users do
- 30:28and you go, oh, oh, we might have broken that.
- 30:30So yeah, 100% edge case testing is great.
- 30:34And I also see at the moment from what I've understood is
- 30:39it's, so it's you've got Shopify integration, what what's the, is
- 30:45it able to be integrated with other platforms?
- 30:47Is it Shopify centric? Great question.
- 30:51So it is not so as an app, it's Shopify app.
- 30:57It works seamlessly. You know, within 5 minutes to 30
- 31:00minutes Max, you are up and live and running if you're on
- 31:04Shopify. If you're not on Shopify, you
- 31:08can also sign up and try and integrate.
- 31:13It requires some technical knowledge for you to integrate
- 31:16it. Shopify is literally click and
- 31:19go. You know, it's like literally
- 31:20click and play. Or if you're a Cleo customer,
- 31:23then you already have your catalog.
- 31:25So it's very easy to integrate by just creating the widget and
- 31:29adding the Cleo API key and it's just up and running there.
- 31:33And of course then then depending on the platform with
- 31:35the remittant or Shopifybee covers, there's a script that
- 31:38you are adding. But it's pretty, pretty
- 31:40lightweight to be honest, like extremely lightweight.
- 31:44We are not seeing more than 30 minutes for almost any
- 31:47integration that we have done so far.
- 31:50Hey, Trusted and Niraj, I know that you've got some some data
- 31:55about the these impact on page like this question will come up.
- 31:58It comes at all time when people are being told just to add
- 32:01scripts as OK, but is this going to really slow down my pages?
- 32:05Can you share a, a metric on that?
- 32:07What? What's the measured impact at
- 32:08the moment? Yeah, we have done a thorough
- 32:11testing of this change. The the short answer is not at
- 32:16all there. There's no major impact on the
- 32:18Bates feed, but the the additional time that it is
- 32:23adding to your website for the first time it is loaded is 14
- 32:28milliseconds and then the JavaScript actually gets cached
- 32:32by the browser, so when the user is actually visiting the second
- 32:36page, they won't see any any latency ready to pass low.
- 32:45Nice, Excellent. It's always nice to have those
- 32:47measured metrics rather than just a generic statement of it
- 32:50doesn't do much. So let's move on to brand
- 32:53controls, because having worked on a few projects with people
- 32:56around, like assistant rap bots, like the customer service ones,
- 33:01brand controls around what what like pre scripted answers and
- 33:05also like tone of voice and language and, you know, stop
- 33:09words that can't be used are really important to customer
- 33:12care teams. So what what can where you're at
- 33:15now? And I appreciate this as a
- 33:16product that's, you know, Rd. in road map development all the
- 33:19time. What can people do now?
- 33:21You know what? What have you got in mind for
- 33:23the future? When it comes to the
- 33:27customisations teams. So today as as Neelay said, it's
- 33:33a one click installation. So out-of-the-box what they get
- 33:38is that they are able to ask any questions about the product
- 33:42itself. The merchants may want to change
- 33:48the CSS to to make it bit more looking like their own website.
- 33:54So that's something they are able to do it as well.
- 33:58In our dashboard, we actually show what questions have been
- 34:02asked for which products, what were the answers given as well.
- 34:08And this is where if merchants wants to come and they want to
- 34:11change the way the answer should be given for that particular
- 34:15product, for that particular question, they are able to do
- 34:19that as well. But as I had mentioned earlier,
- 34:22there may be some custom intents, for example, something
- 34:27to tell customer about the delivery of the product,
- 34:32something to tell them about the promotions.
- 34:35So those custom intents, they are also able to configure that
- 34:39behind the scene. Also earlier we talked about
- 34:42collecting some information from the front end from the page
- 34:46itself. For example, let's say merchants
- 34:49are using some third party tool to calculate what's going to be
- 34:56the shipping charges or they are collecting.
- 34:59This is very common and where the branded things are being
- 35:03sold that specific descriptions of the products are collected in
- 35:09real time from the brand itself. So if they want to submit this
- 35:15additional any this kind of information from the front end
- 35:19in real time, that's also they are able to customize as it
- 35:23stands today, yeah. Excellent.
- 35:27Yeah, that that controller to do preset questions and answers and
- 35:31edit answers is really, really important.
- 35:34Yeah. So it brings me on to my my
- 35:36final key question. I should think As for you Nila
- 35:39is, is is the integration side and what what what you've got
- 35:45planned there? You know what what what can it
- 35:48do integration wise now? Are there any like pre connected
- 35:51integrations or other third parties or do you have any
- 35:53specific ones in minds that you work?
- 35:57Yeah, we are getting a lot of requests from customers who are
- 36:01going live daily right now. But one of the things I'm
- 36:04thinking is like, what's the last kind of in a mile for
- 36:08Marketeer? Now if if I'm marketeer now I
- 36:11know that shopper is on my product base and the shopper
- 36:17actually had this question rather than shopper left or
- 36:20shopper bought. Either could be the scenario,
- 36:24right? I now have that information, but
- 36:26I don't know who is the shopper and in any of these live FAQ or
- 36:32any of the systems, asking user about the e-mail is sort of
- 36:37detrimental to actually even having them using the system.
- 36:41You know it, you always lose out when you try to ask for the
- 36:44emails and so on. One thing we have planned is
- 36:47sort of, you know, seamlessly integrated with Clavio because
- 36:52many of these Shopify customers we have, they use Clavio for
- 36:55example and click if they are using Clavio and the shopper is
- 36:59on that store, they already have the authentication done, meaning
- 37:03you already have the user information.
- 37:05So in Asklo Hub in the back end we provide, you know, all the
- 37:09questions that are asked, the actual sales and so on.
- 37:12But there we have some what column for user ID where we are
- 37:16actually just generating the user ID at the moment.
- 37:18Instead of that, if you can actually show OK,
- 37:20james@something.com actually ask this question, I feel that your
- 37:27marketing follow up could be hyper, hyper personalized.
- 37:30It's like literally 1 to one personalized.
- 37:33There is nothing more. You can go closer than that,
- 37:36right? You go and ask a question that
- 37:39is this available in red variant or red color instead of, you
- 37:43know, blue color. Now, I exactly know that James
- 37:46was the person who wanted this fridge, you know, which had an
- 37:49ice dispenser and he wanted the red colour.
- 37:52Like you can't go closer than that and you are not guessing
- 37:55anymore. You have the information from
- 37:57James exactly what he wanted. So that's what kind of we are
- 38:02thinking kind of making it personalized e-mail follow-ups,
- 38:06how we can do that. That's one. 2nd is that we
- 38:09already have built actually negotiation system within this
- 38:12as well. So you can actually if, if we
- 38:15can, the trick is to understand when you offer the negotiation
- 38:18strategy, you actually lower the price as well for the customer.
- 38:22That's something we, we were experimenting and we have done
- 38:24some work on the machine learning level, but we are not
- 38:27kind of yet bringing it out. We are getting a lot of requests
- 38:31as I said earlier of you know, turning this into.
- 38:34So right now Oslo is at the product level, right?
- 38:38We are getting questions about can you provide it at the store
- 38:41level? Same aslo, can you product store
- 38:43level and that becomes Moi. So Moi will become then a fully
- 38:47conversation. We already have Moi today, but
- 38:49it's more rule based system. It's more rule based than lesser
- 38:54Gen. AI based.
- 38:56So we want to flip that and make Moi also fully open-ended, kind
- 39:00of, you know, Gen. AI based with some rules still
- 39:03in preservation to ensure lesser hallucination and, and, you
- 39:07know, accurate information. So and then of course, we would
- 39:11love to work with, you know, chat bot systems, like I said,
- 39:14but I want to make sure that the experience is seamless.
- 39:17I am, I don't believe that you should have like different kind
- 39:20of windows opening for shopper that actually, you know, mess up
- 39:23the experience. It'd be.
- 39:26Basically like what you have now when you land on a site with
- 39:28cookie pop up newsletter discount pop.
- 39:33Yeah, exactly. It's, I think that's such an
- 39:35important thing is the handover pace and, and this is where the
- 39:39the the e-commerce customer service teams listening need to
- 39:44really think it through is what is the user journey they want to
- 39:46have and what helps they want to provide.
- 39:48It's such a. Question.
- 39:49And it actually it works because they, one of the Shopify reviews
- 39:52that we have on our marketplace that I saw yesterday was that
- 39:56they they could reduce 40% service time from the humans
- 40:0140%, yeah. So it already is giving the
- 40:04impact, but I think we want to kind of go next step and then I
- 40:07think that's where we would kind of go with the wider ecosystem
- 40:11and start building the use cases there.
- 40:13Yeah, yeah. Agreed, really appreciate you
- 40:17both coming on the podcast. Always enjoy listening to you
- 40:20and hearing where you're going with your your products.
- 40:23If anyone here is actually genuinely interested and wants
- 40:25to like have a have a demo and see it, how do they reach out?
- 40:29Just go ask Lord dot AI and just go from there.
- 40:34Cool, are you for those of you listening on our landing page on
- 40:37the website, there will be a link to that.
- 40:39So go and reach out, ask some more questions, have a look for
- 40:42yourself on travel Pro. The it's the EU website.
- 40:44So I think it's like eu.travelpro.com.
- 40:46I think the debate I just before we came on, I asked it one about
- 40:52the child proofing of the case and the answers.
- 40:55Great. I'll stick that when we when we
- 40:57put our post out and LinkedIn, I'll stick that example on
- 40:59because I think it's quite cool. Yeah.
- 41:01And it and it works on all European or all languages for
- 41:04that matter. So, yeah.
- 41:06All right. Thank you so much.
- 41:08Yeah. Thanks, Nilay.
- 41:09Thanks, Niraj. Always a pleasure and take care.
- 41:12Enjoy the rest of your day. Yeah.
- 41:14Thank you so much, James. Thank you.
- 41:17You're listening to the Insidecommerce podcast, hosted
- 41:19by experienced consultants James Gerd and Paul Rogers.
- 41:23Insidecommerce is your fast tracked additional knowledge
- 41:25feature in interviews with leading e-commerce retailers,
- 41:28industry Thoughtleafs and the C-Suite key tech companies to
- 41:31explore the latest digital strategy, customer experience
- 41:33and tech. Discover what the latest trends
- 41:36really mean and get practical advice based on the experience
- 41:39to help fine tune your E Comm strategy and tech.
- 41:42Follow us on LinkedIn and reach out to James and Paul if you'd
- 41:45like to discuss an e-commerce project.
- 41:47We hope you enjoy the latest episode.