Conversational Commerce Reinvented with AI - with Lili Kostadinova from CloudOffice

In this talk, Lily Costustadinova, a corporate sales manager at Cloud Office, discusses the significant impact of artificial intelligence on conversational commerce. She shares insights gathered from her interactions with decision-makers and partners across various industries, emphasizing the growing preference for chatbots among customers. The session explores the evolution of customer interactions from physical stores to digital platforms, highlighting the importance of seamless and empathetic customer experiences. Costustadinova also presents examples of how AI is enhancing chatbot capabilities, particularly in understanding nuanced customer requests in multiple languages. She concludes by stressing the necessity of AI-driven assistants for businesses to remain competitive in the evolving digital landscape.

Uh my name is Lily Costustadinova and I'm a corporate sales manager at the cloud office and as a part of my daily job I do work and I do uh talk to a lot of um decision makers, business owners um and different partners u about their business goals and how they can actually achieve those goals with technology. And believe it or not, in the past um two

years, uh the topic has always like or almost always been artificial intelligence. Um Cloud Office is a cloud solution company. Uh we do have three offices uh three locations, but we do work with more than 700 customers all around the globe. We're a Google Premier partner and we actually won partner of the year for 2024 in in IMIA. Uh so what I'm going to share with you

uh in the next uh 20 minutes is actually collective knowledge and a lot of observations gathered from all of these conversations with um as you can see uh some global brands and also some brands um that are like large local brands. And let's dive into the topic. When we talk about conversational commerce, a lot of people have already something something in mind uh some interpretation, some association.

Mainly to be honest, mainly for a lot of people that's chat bots. Uh and now I want to have a quick quiz uh with you. Just don't be shy, just shout out. I'm just going to ask you uh some things so that we can check if we know the landscape good enough. So the first question is um just a number. What is the percentage of the customers

who actually prefer chatbot than waiting in a queue for a human agent? >> 82 and 20. It's actually 62. So it's quite a lot. So the chatbots actually do have their fans. Then the next question is um what what is the percentage of the customers? Think globally because this is from Statista like it's a worldwide right. Um, what is the percentage of customers who actually prefer messaging

um over calling? Think about Jenzi as well. Yeah, it's it's 75. Um, yeah. All right. Uh, and then the the last one again think globally. Uh, what is the average boost of the revenue of the companies who have implemented AI chat bots? Be honest. 2% 30 yeah it's actually um 20% and I believe this is the reason we we actually have this topic today on the on

the stage um and now talking about the conversational commerce um this here is um like let's start from the very beginning actually the conversation and commerce are have always been bound together for like thousands of years because like uh before the digital world the the commercial the commercial things and basically everything in the commercial world has happened uh by conversation actually and think about the physical shop

experience. When you actually walk into the into the shop, what do you expect? You expect the cashier, you expect somebody to instantly acknowledge you that you're there. If you have a question, you do expect uh that there is going to somebody to answer. And of course you expect um empathy and human behavior and um in the and yeah basically you also expect a seamless shopping experience like

basically no queue would be optimal and those are non-negotiables because if there is no cashier if if you have to wait like for 10 people on the cash desk you're probably not going to make the the purchase actually and this is what the customer expects nowadays and when you have when We have the physical shopping experience. Um the the sales the salesperson or or the cashier and

especially in the luxury shopping experience. Um as the seller uh what do what do they gain? What what do we gain uh from this physical experience we gain apart from obviously the the sales opportunity uh we gain the upselling opportunities. Uh also what we what we gain is the immediate feedback. you can know if the person actually wants to be approached or doesn't want to be approached.

Uh right right from the beginning you can just tell it by just physically looking at the person. And of course we have the opportunities to run the to have the customer loyalty and to build the brand trust uh uh based on the human trust and for all of these things are something that that the people are used to when it comes to the shopping experience generally. And

then the uh in the digital era the digital or the e-commerce uh it's here and actually uh because of that thanks to these opportunities um a lot of things are being even solved. We do have the 247 access uh to some uh to some shopping and I mean to all of them that are online also we do have the global reach uh even when you're somewhere not

in uh the not like even if the shop is not open or you don't have that shop in your own city or town you can also purchase from it and you have the seamless transa transactions there are no uh you don't wait on a queue etc but then a lot things are actually lost in the digital commerce. Uh and it's and it's for me I can name

it. This is the conversational part of the commercial experience. Uh so what what's been lost is the human nuance and what also has been lost is the personalization and honestly after all the all the shops and all the physical shops uh are digitalized what is the industry about in the past 10 years uh let's say and I believe also Karina for from Mastercard also mentioned that it's

about personalization it's about bringing back the personal of experience and the conversation into the digital experience and um there has been a lot of um trials and a lot of companies are working in that direction. Um you have all seen the rule based chat bots uh that actually emerged in 2014 with monkeybot. We have all tried them or some other recommendation engines also Google has their recommendation

engine recommendation AI probably a lot of you use or have you tried uh Reva AI the Bulgarian company um so a lot of um a lot of um services and vendors trying to fill that gap but what what happens actually there's always the human fallback there's always the customer support somebody like the human agent that has to pick up the phone. And what is the drawback there?

That that usually happens when the things goes out when something is wrong. Like we don't have the we rarely okay not at all but rarely have actually the upsell opportunity when the human is already in the conversation. Usually this is where something's not wrong, something's not right, the website is not working or we we're not happy with with something. And actually, let's face it, like how many

of you have had really disappointing experience with a chatbot? Yeah, I would assume I would assume a lot of a lot of us like we just want to talk to a person or the chatbot doesn't understand what we're talking about. And like by filling this gap and meaning merging back conversation and commerce um we are all aware that the artificial intelligence has and actually artificial intelligence being

accessible and being more affordable actually helps a lot a lot with that. And here's some like brief examples of what is actually possible. I'm just going to show you some small examples. This is actually our our um one of our largest projects that is customerf facing like not not just not behind the pay wall or something. um the chat the technopolis chatbot and here with some uh

very uh small prompt I'm looking for a Dyson haird dryer it s seamlessly look like um something that you would expect that a chat or a human agent is can understand right in the second but what is actually happened here and why is that possible with artificial intelligence and was not possible three years ago uh is that uh first we do have the Bulgarian language you know

that with the semantically rich languages probably most of the balcon languages are like that um it's not that easy to train your to train your machine learning model uh now it's it's there just out of it comes out of the box then we do have I'm looking for what if it was um my Dyson hairdryer is broken this is another prompt that somebody has to actually manually

type in and distinguish both prompts in order for the chatbot to response to response um in the correct way. But here it's again contextual and out of the box and then uh we also have um this um we don't have to map the keywords because here we see Dyson it's written in Bulgaria it's not the act it's not the actual um brand name but again it's it's

there. the the chatbot understood my prompt and here are the examples and then something that's a bit more complex. The prompt there is I'm looking for the cheapest larger uh washing machine which brings uh a different layer of the context which is something that even with the filters on the website it's not that easy to actually to actually get into that result. Uh so those are just

some of the simple things that are can now be achieved and bring um and basically help the people uh help people generally to uh in their own in our own language to describe what we're looking for and like transitioning from transactional to conversational experience. And of course I'm going to briefly mention the virtual assistants, voice bots, um uh conversational agents, you name it. uh but uh this

is something that again as I said from a lot of discussions comes a lot in place and it's not only in the retail also in the banking industry and also in a lot of customerf facing industries uh getting the getting the chat bots to offload the human agents like the voice bots offloaded human agents um is something that a lot of companies are trying to achieve and

some companies actually has already u u achieved For example, Wendy's uh the fast food chain in McDonald uh in the fast food chain in America uh and also Commerce Bank uh they're obviously a large bank. They both have released with Google technologies uh their voice bots in 2023 and now they do have um their um and the results are public. They do have really good um results

and now they are not they're not even hiring more customer support but they're actually um reducing a lot of for example in Wendy's now 60% of the drive-th through cashiers are voice bots. So it's it's here and it's getting there also with the balcon languages. And now I'm not going to talk about the conversational uh suit for from Google side. Just one thing I also want to

mention because we do work really closely with uh with the Google engineering product teams. Uh one thing is obviously there are all the the technologies there the AI enhancements are there but what's in the core in the core is the conversational core. the text to speech, speech to text, natural language processing, translation. So like even even Google obviously acknowledge that we are aiming to have the seamless

conversation uh and on top of the seamless conversation we are building with enhancements etc. and now yeah you would assume that's the end of my presentation but that's just the beginning not of my presentation but of um of the conversational commerce generally uh because what I have uh showed you for now is the current state uh where we currently are with the chat bots and voice bots

what's actually happening but what's uh what's going to happen well first for everyone in order order to have the chat bots and in order to have the voice bot you have done already one really important step and and it is actually transferring your data to a machine understandable data sets. Um you have vectorized your product catalog you have all of your knowledge bases into embeddings. Uh so

you basically you have done the heavy lifting with your data for some of some of the businesses if not all of them. That also includes actually the duplicating your data, preparing the data generally, but also optimizing it for a large language model models prompting and apart from that you have your platform. So your digital digital space then you have your data prepared and you have your conversational

interface. So uh meaning you have your chatbot or the voice bot but they are only uh but you can like your customers can only reach them if they open your web page and if they pick up and dial your number uh but uh there is one other like we don't exist in a vacuum. So basically we have to be aware with the entire ecosystem and for example

something that is sounded futuristic a couple of years ago but now it's getting there and there uh I believe a lot of you are actually using already JPT for recommending products and this is something the hyperscalers are already working on and it's hey Siri where can I buy a haird dryer because my mine is broken and where is the closest uh store for buying a dress uh

for for tonight. It's open now. And can you propose something? Yeah, just you you get where I'm going. So, this is going to be expected at some point and in the states it's already there as well. Then, um one other thing that we have to be aware of if it shows are the smart devices. Probably not humane and rabbit because they didn't make it. Um but generally

obvious uh but generally the market is working in that direction having the smart devices the smart pins a lot of money are put in that direction and a lot of uh business people are betting on the smart devices. Um and they do not work with applications they work with AI agents. So in order to get there you would have to be your story would have to be

prepared for for that as well. And then uh last and not least uh also the Google shopping that's connected with the merchant center and cla are also uh accepting now that you can just um get uh give them your product catalog but you can also give them your embeddings. So if you have the control for your embeddings probably um you're going to have more more control of

what they're actually showing. And I only have one key takeaway honestly and it's that the AIdriven assistants are not are not a just a wishful thinking and they're just not nice to have anymore but they are a must to stay relevant in the coming years and thank you for your attention. I hope I have I managed to spark the conversation. You can find me around. Thank you.