Topics covered:
We invite you to listen to another episode of our WM Talks podcast, where we discuss technology and business topics related to the IT industry.
Szymon Kubala: Hi, welcome to another episode of Webmakers Talks.
Today we are going to talk about why artificial intelligence matters in search engines. Today in the studio with me are:
Damian Maślanka: Hi, my name is Damian Maślanka, I am the CTO of Webmakers.
Karol Kozak: Hi, my name is Karol Kozak and I am the Head of Appstore at Shoper.
Szymon: Great. Listen, today at the start, at the very beginning of our podcast, I have a question for you, or more precisely for you Karol. Why do you think AI-powered search is crucial nowadays? This is actually the topic of this episode, so let's start with that.
Karol: Maybe I will start with why search in general is crucial. As our statistics show, the search engine is the first element that most users interact with on our website. If I use the search and see zero results, then about 10-12% of users leave the store immediately. That is why search in general - and AI search even more so - is something that can keep a customer in the store. So there is a very high chance that they will buy something if they use the search. And this is also confirmed by our statistics - not only ours. There is about a 7-10 times higher chance that they will convert.
Szymon: Okay, so in a way the search engine is a natural mechanism for leads, for acquiring leads, but actually more for conversions - something like that.
Karol: Maybe not exactly conversions, but speaking less statistically - if we enter a store, search for something and see that it is not there, we think "okay, I don't need to be here" and go somewhere else. But if I find the product I was looking for, there is a very high chance I will buy it, because I already had a clear intent - I was searching for something. And here a big problem often is, especially in stores with a very large assortment, that it is hard to find products because there are so many of them. But we will probably talk about this in more detail in a moment, so I will not go deeper into it now.
Szymon: Okay. Damian, we are also here because together with your company you created an application called Findnado. And what exactly does it do? How is it supposed to improve the search that Karol just talked about?
Damian: Yes. Here, above all, there are the two factors that Karol mentioned. The first one is zero results - if someone enters an imprecise keyword or simply a phrase that is too broad for the search engine to handle, they will receive zero results, and this is a factor that simply pushes the user away. The other side of the coin is when we have too many results - for a basic phrase or in a more advanced store, the search returns a long list of products. This also creates a problem, because users generally prefer more specific, narrowed-down results rather than having to browse through multiple pages. So this is a dual factor that needs to be addressed.
Szymon: Okay, great. So going back to search engines - what other problems, besides the ones you mentioned, do current search engines struggle with? And how do you think artificial intelligence can eliminate these issues?
Karol: There are many such problems - from typos to synonyms. And sometimes there are customers who roughly know what they are looking for, but they do not know the product or what it is actually called. Let me give an analogy - probably everyone has experienced this. You go into a hardware store, approach an employee and say: I need something to connect two pieces of wood, but I do not really know what I need or what the product is called. And the person replies: Ah, wood glue, right? Let's call it that. Standard search engines, when you say you are looking for something to connect two pieces of wood, will respond: There is no such product. And this is exactly where AI search comes in.
Damian: Yes, I can share a case from my own experience. At one point I needed a specific screw. I do not know much about screws, so it was hard for me to find the right type in online stores. I just took a photo of it, uploaded it into Google image search at the time, and Google found similar images. That led me to stores where I could find the exact screw and match it properly.
Szymon: Now instead of Google Lens, you can just ask ChatGPT what kind of screw it is and where to find it, right?
Karol: Exactly.
Damian: If we integrate image-based search into an online store, so the search engine works not only with text but also allows uploading an image to find similar products, then we gain an advantage. The first interaction becomes positive - the store provides value, shows whether it has the product or not, and does not force the user into a tedious search process.
Szymon: Okay, but from the customer perspective - if I want such an AI-powered search engine in my store, do I need to optimise for AI? Do I need to appear there? Is it even worth it? Or does this tool work independently within my store and learn my products internally? How does it work?
Damian: This mainly works inside the store. It is not something that serves external positioning. Of course, you can use it in marketing communication - for example, promoting that your store has an intelligent search engine that helps users find products faster and more accurately. So in that sense it has some external value. But primarily, it is meant to improve processes inside the store and guide the user.
Maybe this is not the perfect comparison, but something like a sales assistant - and statistics show this - does not always work well for users. People often have an aversion to chatbots. If I contact support, I usually want to talk to a human. The same applies to sales assistants. So you need to find a balance - enable better, more interactive search where users can refine results, but without turning it into a full virtual assistant that tries to advise everything, because it is still not a complete replacement for a human.
Szymon: This is actually a question related to what Damian said - Karol, for you. You mentioned chats and some kind of consultation. Do you actually see on Shoper that this movement is shifting from chats, from suggestions, from customer service to tools that are more "no human", without involving people?
Karol: There are more and more of these tools appearing and I have to say I see value in them. However, they also need to be used properly. If I can actually shorten the time it takes to get information about certain things - like how to make a return, what the delivery policy is, and so on - then I do not really need a human here. Of course, contact with a human in a store is sometimes necessary. Our merchants often have a lot of generic inquiries, where the human - maybe "replaceable" is not the right word - but we can use them for other tasks, which gives a big advantage. However, if we wanted to rely only on a chatbot and completely eliminate the human factor, that probably would not work. We know there are situations where a human is needed, where the case is unique. So as support - yes, absolutely. As a full replacement - not yet.
Szymon: Okay. We drifted a bit away from search, but for a good reason. I also wanted to ask about Shoper as a whole. How do you approach AI integration? Implementing it directly into your platform? Then we will come back to the search engine, because that is the key topic and I am also curious about it.
Karol: Shoper is currently investing heavily in AI, in various products. I cannot talk about some of them, but I recommend watching the investor conference on YouTube. Among other things, a merchant assistant was presented there - powered by AI - which will help our merchants sell more. Segmentation, reports, and so on. It will be a very powerful tool. We are also investing in other areas - content generation, visual themes, and so on. There will be more and more of these solutions. The question is not whether Shoper will do it, but how fast.
Szymon: Okay. The market is basically validating everything that is happening. So from your perspective, Damian - how does AI influence development in software houses and the development of all these tools? You are delivering a tool for Shoper. How did you have to change for this to work?
Damian: It can be compared to the fact that the market is now mature. There are many tools that support online sales or basic IT processes. AI is now another layer - an overlay - that gives these tools a fresh wave of innovation.
Things that were previously less possible are now becoming achievable. For example, in the past we had advanced CMS systems. You had to click through many steps to achieve something - add a record, modify something. The future will be different. We will increasingly delegate tasks - instead of specifying every step, we will say what we want to achieve, and systems will adapt to that. So the market has huge potential now. The key is that every solid, stable tool will have plugins, extensions, additional layers that bring AI into it.
And it is similar with systems like Shoper. There is already a large base of plugins in the marketplace, and more AI-based plugins will continue to appear. These will add functionalities that were previously impossible because AI was not advanced enough. This will definitely improve everyday usage - both for store customers and store owners, who will be able to manage their stores in a more optimal way.
Szymon: Okay. As Webmakers, together with Shoper, you created and implemented this plugin in a store. So I would like to ask two things. First, where did the idea come from - although we already partially answered that. And second, from a seller's perspective - what problems can this plugin solve? It basically suggests to the user what they can get when they enter the site. Will I be able to talk to it in natural language? How would you convince someone who does not yet use AI? Because in the tech industry we do, but sellers are not always tech geeks - often they are not.
Damian: Right. Where did the idea come from? As soon as ChatGPT became popular - when people started talking about it, including friends - we saw the potential. If people get used to talking to AI, then naturally this interaction will start to integrate into systems.
And the first thing that came to my mind was that the keyword-based search we had at the time would simply become less and less desirable. Users would stop relying on short keywords - one, two or three words - and instead start writing full sentences. It would simply be easier for them. It is much easier to type "I am looking for a dress for an evening out up to 250.00" than to define it with keywords and then filter results. And that is how the idea was born - to create a search plugin that can handle more complex queries, the kind that come naturally in a casual conversation when we are looking for something. Like the example Karol mentioned earlier - you go into a store and tell the salesperson what you are looking for because you have a specific problem. So naturally, a search engine should move in that direction - becoming something like a sales assistant. Maybe not a full assistant, but a support tool where I describe my need and it gets interpreted and searched.
Szymon: Like an intern behind the counter. Something like that.
Karol: Kind of, yes. But it brings a lot of value. And if we look at younger generations - 15, 16, 17, 18-year-olds - I do not know if you have noticed, but even younger ones too - they already talk to their phones. They speak to them. So I think this will keep developing. But there is also a big factor of trust. If we look back 20–25 years, at the beginnings of a marketplace platform in Poland, people were thinking: how can I pay and not receive the product immediately? It will arrive later? People were saying: "That is impossible. I will pay when I receive it." There was no trust at all. And I think it is similar now with AI agents. Today we do not trust them fully yet, but that will probably change.
Szymon: You say it will change, and I would like to ask you, Karol - how do you see the future of e-commerce? We already see major changes. The last 2–3 years have brought a huge leap in AI, personalisation and ordering.
How do you imagine it? Because you work closely with data, dashboards and analytics, so you probably have some predictions.
Karol: Predicting the future in tech is tricky, so I hope no one quotes me in two years saying I said something stupid. But I think there are two possible scenarios, and they will probably merge. The first one is that we will have our own personal shopping assistant. The key question is trust - whether we will trust it enough to handle purchases for us. With access to user data - what we buy, when we buy, our preferences - we could say: Listen, I am out of T-shirts, order me three like last time. If we trust it to order from the right place, pay the right way and handle everything correctly, this could be one direction.
Szymon: This is already happening with fridges. You have Samsung fridges and when you run out of milk - it just orders it.
Create a modern online store with us that will make your brand stand out on the market.
Karol: Exactly. And speaking about personalisation - if we do not fully trust that scenario, then there is a second one: decision-making. Around 2019, if I remember correctly, Amazon tried to convince everyone that we would shop using voice. Hey Alexa, buy me a toothbrush. And it turned out that only a small percentage of people used it and did not come back to it, because they lacked control - how it would be delivered, how they would pay, what exact product it would be, cheaper or more expensive, which brand.
So that was a bit of a failure for Amazon. And something similar could happen here. Maybe we will still want to make final decisions ourselves. But perhaps such an assistant will say: Hey, I prepared your shopping list for this month. You bought filters two weeks ago, so they might be running out. You last bought clothes two years ago, so you may need these items. Check this list and choose what you want. Then we review it, select items, choose payment and delivery ourselves. That is the second scenario. But which direction it will take - that is hard to predict.
Szymon: So it all comes down to trust, like you said.
Karol: I think trust and decision-making will be the two key values here.
Szymon: Okay, Damian, based on what Karol said - do you think that in the future we will actually delegate purchases?
Damian: I agree that it is very important to align technology with the right moment in time. The market is much younger now. The younger generation mentioned earlier perceives trends and the internet in a completely different way. These are not the times from 20 years ago when everything was just starting. So I think trust will evolve in the same way - just like people once were afraid to shop online because they might not receive the package. That fear will gradually disappear, and we will trust these assistants more and more. At the same time, AI models will continue to improve, which means these assistants will become more reliable, understand us better, and it will be easier for us to entrust them with tasks. As we know, everyone is looking to save time on routine activities. If such an assistant can handle these tasks reliably, there will be nothing stopping us from using it.
Karol: If we allow it, it will probably know more about us than our family.
Damian: That is actually the most frightening part.
Karol: It is frightening, but also very realistic. Even with basic shopping data, there is an огром amount of information. Such a system can know what we buy, when we buy, how much we spend, how often we spend money on impulses or luxury items. So it is possible to build a very detailed customer profile and suggest what and when we should buy.
Szymon: This is already happening in banking. We have summaries of all our purchases. Banks know a lot about us - whether we are sick, whether we buy medicine or supplements, and so on. It is a bit like a Black Mirror episode.
Damian: But this is no longer just a vision - it is becoming reality. Who knows, maybe one day a courier will knock on our door with a package we never ordered, but containing something we actually wanted.
Karol: Exactly. With the option to pay if we want it. Interesting vision.
Szymon: Okay, we are speculating, but to wrap this up - if you had to give one piece of advice to a seller or anyone watching this episode, something they should remember - what would it be? Times are changing, we are in tech, but they are not necessarily. If you are not growing, you are falling behind - so how would you push them in the right direction?
Damian: When it comes to search, the first question is: are we analysing search behaviour in our store? Do we check what our users are typing? Do they always receive a list of products? If yes, does that list meet their expectations? Or are we losing customers because they entered a keyword that was not handled? This is the first step we should take to improve search. Based on this analysis, we should consider whether it is worth implementing an AI-powered search engine that can handle these unserved queries, enable contextual search, or even image-based search. In the case of Findnado, we also have an advanced analytics panel that shows even more - trends, what users are searching for, and what they actually want. So from a search perspective, this is the path I would follow. And as I always say, it is also important to observe what competitors are doing. We may run a store in the middle of this AI revolution, but we should constantly check what AI plugins our competitors are implementing, whether they offer functionalities that our store does not have.
We are living in a time where a period of stability has ended. Now we are in a phase of rapid adoption of AI solutions, and it is important to keep up with it.
Szymon: And what about you, Karol - do you agree with that, or do you see it differently?
Karol: Definitely. I will keep it short - if search is a key element in your e-commerce, you should invest in an AI-powered search engine, for example Findnado, because it does a great job. No traditional search engine - especially in stores with a large assortment - can deliver the same results, and conversion can significantly increase. I will also add that we often see cases in stores with very diverse assortments, where they say: "Okay, some things are easy to search, but others are not, because our search engine is configured to target specific products." AI-based search engines solve these problems. So anyone struggling with search should definitely look into this type of solution.
Szymon: Great. Finally, where can listeners find more information about Shoper and Webmakers?
Damian: webmakers.expert and the Findnado plugin itself at findnado.com.
Szymon: Great. Thanks a lot for today's episode of Webmakers Talks and see you in the next one. Take care. Bye.
Thank you for listening to this episode. For more interesting content, visit our blog at www.webmakers.expert.
FAQ
The search engine is often the first element a user interacts with in a store. If they cannot find a product, some users leave immediately. On the other hand, users who use search are much more likely to make a purchase.
When a user enters a query and receives no results, some of them leave the store. This means losing potential customers and directly impacts sales.
A very broad list of results makes it harder for users to find a specific product. Customers prefer receiving relevant and narrowed-down options instead of browsing through multiple pages.
Classic search engines struggle with typos, synonyms and descriptive queries. They also fail when users do not know the product name but can describe its use.
AI search can interpret more complex queries, understand context and match products even when users do not use precise keywords. It can also support image-based search.
Yes, thanks to AI it is possible to search for products based on an image. The user can upload a photo and the system will find similar products.
No. It can support users in finding products and handling simple needs, but it does not fully replace human contact, which is still necessary in more complex situations.
No. AI search works primarily inside the store and learns from its data, helping users find products faster and more accurately.
Users are moving away from single keywords toward full sentences and natural language, which requires more advanced search systems.
AI acts as a new layer for existing systems, enabling process automation, better result matching and the development of new features that were previously difficult or impossible to implement.
It is important to analyse what users search for, whether they find products and where issues occur. Based on this, AI search can be implemented to better handle queries and improve user experience.
Systems can analyse purchase history, spending frequency and user preferences to better match product recommendations.
If search is a key part of the store, investing in AI solutions can significantly improve conversion and help customers find products more easily, especially in stores with large assortments.
