Marketer vs AI: Who Actually Boosts Revenue? - with Alex Danchenko from Yespo

This talk explores the growing impact of AI technologies in the marketing sector, highlighting the speaker's experience as a co-founder of Yespo, an omni-channel customer data platform. The speaker discusses how AI tools, such as predictive segmentation, product recommendations, and continuous AB testing, are reshaping marketing strategies and improving business performance. Emphasizing the significance of customer data, the talk illustrates how advanced algorithms can enhance marketing efforts by analyzing user behavior and preferences. The speaker also addresses the future potential of agentic AI, where marketers may soon be able to leverage AI as an assistant for campaign management. Ultimately, the session underscores the importance of human involvement in harnessing AI to drive better business outcomes.

Uh hello everyone. Very nice to be here. It's actually my uh third time I'm here in uh Bulgaria and third time on the stage. And uh I want to start with this important question that probably all of us thinking about. Okay. The AI is rapidly growing. The implementation of AI is just everywhere. And uh through these three my previous speeches, two previous speeches the first time I

was asking a question. So who of you present here Chad GPT? Thank you. The last year I was asking who of you regularly uses chat GPT. And today I want to ask who of you regularly uses chpt or uh cloud on everyday basis for their job. So we see how this new technology became like very very uh common for us and we even don't know how to

work without it anymore. And uh certainly there are some benefits and drawbacks. But the question here is who is actually responsible for the great results of a company? Is it marketer or is it new technologies and AI that helps us work better? A few words about myself. I'm as was presented a a co co-founder and the director of Yespo the omni channel customer data platform. We help

e-commerce businesses to collect the data about their customers and then make the most efficient communication that will bring uh more revenues and more happy customers to this business. And as a customer data platform, knowing a lot about your customers, we understand that this data is vital for building effective AI based technologies. And let's start with a a little bit of statistics. The AI market is booming and

growing rapidly. And here we see how investment in AI technologies are growing every year. And even most scary thing is actually this graph. It shows how in different tasks AI is getting comparable with the efficiency of people. the the tasks like image classification, visual common sense reasoning, uh the nature language interference are the areas where artificial intelligence is already better than human and okay that's scary you

know the common question are we going to lose our jobs uh let's try to understand how actually AI influences the You know the technologies based on AI are on the market for a while already and uh predictive segmentation, product recommendations, marketing automations, they are on the market for several years. It's not something completely new. Uh and there is a set of tools that we as a marketer

sometimes use. Some use product recommendations, some already implement predictive segmentation for their campaigns. Some generate images or text for the their messages and the biggest shift that we expect will be the next big wave is the agentic approach where all those technology later available separately will be combined in one interface where you where AI as a marketer's assistant will help you to manage all the campaigns all

the marketing in your company and probably this makes everything even worse. Yeah, there's going to be someone who's going to do all my job. What the hell? Let's dive deeper in how exactly those technologies improve business performance. The three main areas, product recommendations, predictive segmentation, AI assisted AB testing and agentic AI. There's four main trends. Product recommendations is basically the technology where AI knows exactly what your

customers want, what they want to buy, what's the next best offer for them. And this technologies also evolve. Before that, we had um so-called collaborative filtering uh modeling that help to predict what people buy based basically on a frequency. The more often people buy some stuff together, the more often the algorithm would recommend this additional product to previous one. The transformers, it's new words. And for you

to remember that I want to tell the chat GPT actually has this T in the end. This is transformers and that's a technology that is already implemented not just for chatting but also for predicting the next uh step of a customer. Uh the model works in the way I'll try to explain it not the way it's described on this slide because I don't like them. So imagine

that customer when visiting your website he's writing his own story. So, I went to the main page, then I went to the category of shoes, then I went to a subcategory of sneakers, then I looked at a product from Nike. Uh, and and that's my story. Imagine that transformer algorithm will see it as a sentence like Chad GPT is something you typed there and then he tries

to predict how to continue this story. And in the same way how he sees the whole sequence of actions of um all sequence of words in the sentence you uh plug into chat GPT. The same way uh transformer algorithm can continue the story for your user and make this whole story complete. And this technology works very well when he understands that person that was looking at bicycle

then he purchased a water uh for sports and then he purchased a gloves. The old uh generation of algorithms would recommend three different products but transformers understands okay bicycle sports water and gloves for uh exercises. This is probably a bicycler. The someone who is who likes bicycling basically. And to complete his story, he will suggest absolutely different products like t-shirt or shorts or uh helmet. And let's

make it even better. Uh the large language models brought us the ability to explain to the machine to the algorithms what actually means each of these particular uh items because in previous generation of recommendations we only basically the output of the recommendation algorithm was if the product ID was 1,67 the next product ID is 1 million68 for instance. Now vectorizing of uh descriptions or images of products

helps us to give the algorithm the understanding of what these particular products are and in this case it can better predict what are the best next products for this particular person. It solves completely the problem of the cold start. You know when you have to train model for a long period and have statistics on every item sold so it will be recommended. Now it's not necessarily the

product algorithms improved and improved not only in my imagination it's uh actually uh confirmed by these numbers. We implemented this new technology across different industry. In electronics we we saw 40% uplift in conversions. In tools and home and gardening, it was 21% improved. In fashion, we had tremendous results because complete look when you purchase jeans, jacket and t-shirt, it has to be completed and in this way

it works even better. Uh we had almost over three times better conversions with these new algorithms. And let's now move to the initial question. So who actually improved those results? Are those algorithms? The team of Neprom which is a manufacturer and seller of tools uh number one uh brand like this uh in my domestic country but also present across the Europe. uh they improved CTR by 105%

basically twice improved the CTR and uh grew the sale of uh share of sales twice not because the algorithm is so good but because their team was smart enough to implement it another major uh solution that is already in the market for a few years it's not something completely new is uh when AI knows who are the customers that want to buy from you. Basically, it helps

to understand from all your customers who are the most interested in this particular offering you have for today and helps you as a marketer to pick the proper audience. It works in a similar way. It analyzes the historical data. What people were looking at, what were they purchasing, what was um what was their experience? Any type of data could be used. And then as an uh output,

we have predictions. The level and probability of conversion of each particular person you have in your database. And why it works better? Because we all here marketers, we know how to do segmentation. Yes. Who Let's play the game again. who does uh segmenting for Okay, we have uh guys who are know about it. For others, I will explain. As a marketer, you usually have an opportunity to

narrow down your audience. So, for instance, for Apple watch, I will take the I will take the audience for people who use Apple devices when browsing my website. They also purchased some products from Apple before. um maybe those who have uh were engaged in emails when where I advertised Apple products. Uh I could also understand your gender and target in this way. But uh predictive AI algorithms

take into account a lot more. It also sees patterns in in the way how people purchase. It sees uh all all the browsing history of the users and it also sees similarities between the different groups of users and in this way it can improve your segment. Just one nice example. Otaji is a Ukrainian clothing brand. Uh they offer closes for women and they had the Viber campaigns.

So I know Viber is quite popular here in Bulgaria as well as in in my home country. So this example is very valuable for you. Before that they used manual segmentation. Marketer was actually also trying to narrow down the audience so they would get the most effect from the var campaign. Uh but with predictive segmentation they managed to eliminate all the users who will not likely buy

and also add to the segment people who were not taken to into account by manual filtering but adding more people who are ready to buy and they were increasing the number of conversions eventually. So the simple number 310% increase in return on marketing investment means that we managed to get four time more sales spending four times less money. One of this the next example where we could

as marketers benefit from using AI is continuous AB testing. Well, typically AB tests, everybody speak about AB testing. Um, but honestly, it's hard and it's not very popular. So, just to for statistics, who of you made an AB test in the past months? Okay, one, two, two people, three, four. Okay, it's even less than uh I expected but there is a reason for that. Uh there is

I don't know if you see it very well but there is an example how to do a multivariant AB testing in any marketing automation tool. It's hard to do. Uh so that's why uh the new approach one from many block is a block that automatically iterates different variants of the message. Uh quite often we see that once implemented a very well done uh abundant card campaign or

abundant view campaign it stays like this for for months or years because you have to take care about a lot of other stuff. You have your seasonal campaigns, you have something broken, you have data missing. And AB testing is something that we put for for future. But now we can do AB test immediately and continuously. Even if you don't want to do them yourself, you can delegate

it to AI and just confirm the suggested changes. This is an example of from UA. This is the number one marketplace in Ukraine. It has about 52 million monthly visitors on the website and they have about about 2.5 orders monthly. So this is huge scale and the team implemented this sequence of automatically optimized campaigns which started running for two months. Only in two months with this minor

change that didn't cost them anything. Additionally, they managed to increase 25 uh 26% more clicks and in general it brought five more% of sales. Zero investment just new approach to the way of doing those campaigns and the agentic AI is something that we expect to come. uh we understand that the way we interact with chat GPT when he runs uh Python script to show us a diagram

or analyzes the data uh is something that is quite familiar already and in this way we're going to see how all the tools marketing automation will work in future. So marketer sets the goal, the system tries to offer the best ways to achieve those goals and marketer have to oversee them, accept and let the system uh execute. We have an example of how it works with email

uh email campaign creation. Uh for instance, for webinars, we already have this pilot solution. When you provide only link photo for the webinar, the AI generates the con the sequence of emails that have to be sent. It prepares a content and it prepares all the copyrightiting for this uh campaign and then it runs AB tests iterating from one type of message to another sending email and text

message or maybe email and another email if it was not opened. And in this way we have the continuous improvement process running daily. So there is a plenty of benefits of using AI solutions. They really improve the numbers. And uh I want to show you how as an illustration the results of analysis uh from British airwaves the on average up to 70 uh% of pilots have an

ability to have an app during the plane flight because there's a lot of automations and of course this is a joke. I don't mean that we will sleep at our works in future. We will not sleep at our workspaces. But what will change is that we will preserve our energy from tedious and repetitive work for creative strategic thinking and u that's where AI will help us and

quite common phrase it's not uh AI that will replace the humans it's humans with AI will replace humans without AI and if you get back to initial question. So who actually increases the revenue of the company? Is this a marketer or is this an AI? My answer is it's you. You are the people who can implement the changes in your businesses so that the benefits of AI

will be available for your business. So it's you. Thanks. Thank you.