A New Era with E-Commerce, Content, and Artificial Intelligence

Revolutionizing e-commerce product descriptions, category copywriting, and automated user journeys with AI.

Speakers: Can Mutioğlu, Tahir Yıldız, Yiğit Konur

26 Mar 2024

Video Transcript

⏱ 38 min read• 6,792 words

Read the full speech or switch languages

— Hey everyone. I’m hosting today. Tahir, welcome. Thanks for joining us. You look younger since we last met. Your energy is different. You look great. I hope everything is going well.

— I’m good. Thanks. How about you?

— I’m doing great. I moved to a new country. I’m still adjusting. The AI agenda is intense. It’s a fun time to be alive. Keeping up with all the updates is stressful though. How are things with you? Is the e-commerce scene busy? How much does AI impact your day?

— It impacts all of us. It’s not just e-commerce. Everyone’s job is changing. It feels like the pandemic. We didn't know the impact at first. This is similar. Things keep happening. The pace isn't slowing down. New stuff drops every day. There’s anxiety, but people want to find opportunities. In my experience, integrating AI into large corporations is tough. It’s tough due to processes. We don't know where to integrate AI. Everyone is experimenting. How and where should AI be integrated? There’s a lot of uncertainty. It’s not like launching a simple startup.

These processes have existed for years. People are involved in them. It impacts their roles. They wonder. "What happens to me if we use this?" I see this everywhere. Not just with us. There’s a lot of chaos right now. It’s a global issue, not just our sector. I’m part of the Euronics group. We met with 17 or 18 countries last week. Everyone says the same thing. Organic traffic dropped after LLMs emerged. My peers report losses between 15% and 30%. That’s a massive hit. It’s really affecting things.

— And it will likely continue. This isn't a temporary dip. That traffic isn't coming back. It has moved elsewhere.

— Exactly. The chaos comes from constant innovation. New models appear every day. Brands are paralyzed by choice. They don't know which AI to integrate. Especially big corporate brands. You start using something because it’s good. A week later, something better drops. You think you can do better. Then they do it way better. That’s the biggest problem.

— Model capacities are evolving fast. The quality of what they can do is rising. With the latest Opus jump, everyone is reaching a developer level quickly. It’s really impressive. I agree, Tahir. The pandemic analogy is spot on. It’s a period with a glimmer of hope. It’s not all negative. But the overall energy is definitely stressful. How did you adapt personally? I’m sure your brand is doing a lot. Balancing team motivation is tough. As a manager, it’s even harder. People have real concerns. It affects their work.

If any profession feels safe from AI, drop it in the chat. We’ll start taking their courses now. It’s a tense time. How did your daily workflow change? I’m curious about that. You manage a big brand. There are regulations, sure. It’s not as heavy as finance. But you’re in a big holding company. How did it affect your daily life practically? I'm really curious about that too.

— Like everyone else, I try to keep up with what's happening. But there's that FOMO thing, right? We don't actually have to follow every single thing. I follow it because of my job and because I love it. If I didn't love it, I wouldn't want to keep up. It only impacts specific areas. E-commerce is clear about what it does. I always say this. It’s no different than running a small business. E-commerce covers everything. Marketing, accounting, supply chain, logistics, and shipping. It’s a mini company. It’s just traditional trade done digitally. People think e-commerce is something else.

— It’s basically a digital bazaar.

— Exactly. The name sounds fancy. Electronic commerce. But at the end of the day, we just buy and sell products. It’s no different from normal trade. We just do it online. The fundamentals never change. They just get easier. AI is now huge for product entry. Tons of startups are building tools for this. We call them attributes. They enrich product details and photos. Now they even create videos. It’s never been easier to make custom product videos. But how many websites actually use them?

Probably less than 1%.

— Yeah, adoption is lagging. With so many products, users expect better experiences everywhere. But there's that "AI slop" factor. It’s often low quality and lacks consistency. You might have hundreds of thousands of SKUs. Even with a great video model, you can't just go live. You have to check everything. Hallucinations are a real risk. That’s why it hasn't hit the mainstream yet.

— We might be biased. I know I am. We’re deep in the weeds, so we assume everyone manages it this way. But most people are still far behind. There are plenty of solutions, though. Adding products is easy. Making product videos is easy. But this is still not mainstream. It’s growing as Gemini’s UI evolves. We saw it with Nano Banana and DALL·E 3. Few people used DALL·E back then. Hardly anyone knew it. Nano Banana is doing something better now. It’s finally reaching the masses. Most people in Turkey have the free ChatGPT on their phones.

They’re learning what it can do. That’s why AI images and videos are blowing up on Instagram and TikTok. As access gets easier, adoption will speed up. It’s easier to use a pre-built agent than to build one.

— Exactly. We see that clearly with visuals. As model quality improves, it reaches more people. Go ahead, Tahir.

— I agree on the adoption part. But big companies have rigid workflows. Changing how they work isn't easy. If you're a startup, you can just use Nano Banana or Gemini. But big companies have used the same product workflows for 15 years. The dashboard hasn't changed in 15 years. The staff has been there for a decade. You can't expect them to be as tech-savvy as us. They aren't. I’m speaking generally, not about us. But the younger generation... The younger crowd is different. But not everyone is equally open to innovation. If they see Nano Banana in their system, they might try it. They think they can just pull data back and forth.

That doesn't really work. Then there's the model side. You integrate Nano Banana today. A month later, a better model drops. Now your system is already outdated. Big holding companies have to be responsible. We deal with consumer complaints and arbitration boards. We can't have AI generating weird product photos or videos. Customers will return items if they don't match the images. That makes using AI risky. Nobody wants that headache. Online return rates are already climbing. In textiles, we're talking about 30 percent. If I sell sweaters and use AI, the pattern might be wrong. We all know that happens. The customer will just return it. We end up hurting our business while trying to improve it.

That’s why companies still see it as risky. I get why they feel that way.

— Honestly, before coming here, Can and I checked our research agent. We wanted to see how the conversation flows. We looked at some global examples to see what’s happening out there. Like you said, there’s great movement in content, product listings, and photography. After software, support services are where AI adds the most value. Live support and voice agents are seeing very active use. I’m expecting some interesting shifts there too. I’m curious about your take on this. Recommendation engines usually feed off browser behavior.

They use classic search algorithms to create a shortlist. Then they prioritize the results. I know you’re well-versed in these recommendation engines and product suggestion tools. You’ve invested in these tools for years. As someone who knows LLM structures well, do you see a big win coming? Where do you think generative AI can really disrupt e-commerce? Have you noticed anything interesting with recommendation engines lately?

— Yiğit, recommendation engines changed a long time ago. Even before the ChatGPT wave, top players were using them. Israeli firms were already using LLMs. Or at least some form of machine learning.

— They used embeddings. Algorithms based on natural language similarity. Exactly.

— Right. Here is my point. You are right about browsing behavior. If a user buys an iPhone and then a case, the pattern grows. I can show cases to other iPhone users. That is how it worked. But we moved past that three or four years ago. Many Turkish companies are doing this now. They come to us pitching their products. They use AI to track everything. From mouse clicks to behavior analysis. They model if a user is ready to buy. They claim they show AirPods to me and cheaper headphones to someone else.

I say "claim" because everything is AI assisted now. It scares business owners like us. We cannot verify the tech. How are you actually using AI? Everyone is saying AI. It’s unclear if it’s actually there. It’s all a bit messy. Machine learning gets mixed up with AI too. I don’t know how much is generative. There’s a behavioral side and a learning side. It’s really inside the website. It complicates everything. Plus, there’s just so much of it.

— There’s the search side too. It’s where people go to type things directly.

— Lately, we’ve had the cookie issues and Apple’s blocking. We can’t track cross-device usage. We can’t see or learn from it. In our world, it’s not just a simple desktop purchase. Especially for high-ticket items. We always see cross-device behavior. They start on mobile, switch to the app, then move to desktop. Tracking that is tough. It’s the cookie problem.

— Tracking AI trafiğini is also really hard. The world of measurement keeps getting more challenging.

— That’s a whole other headache. It deserves its own show. Everyone I talk to is in a different world. Especially post-cookies. E-commerce hasn't really found common ground yet. We tried server-side tracking. We actually did it. We pull everything first-party from our own server. Some take risks, some don't. There are so many methods. AI methods are in there too. Google said they’d provide AI-powered visitor estimates. But visitor count is a key KPI for us. I don't want to leave that to AI. I want to see the real, accurate numbers.

AI is everywhere these days. It’s definitely being used there. But how much is it actually working? After we get our first-party data, it doesn't seem very consistent. AI'ın impact is changing things. Many e-commerce pros complain about GA4. We’ve sailed into a complex world. I’ve been in e-commerce for years. Even I get lost in GA4 sometimes. We were so used to UA for years. The event-based model and API integrations are totally different. Everything needs a custom report now. It will improve over time. Nothing happens overnight. People will adjust.

— Things are getting complex. But complexity for us isn't complexity for AI. In a way, it makes our jobs more secure. It ensures our professional continuity. As long as there's more context, AI performs better. Can and I are working on an MCP server right now. It lets AI easily access SEO, analytics, and Search Console data. It’s highly token optimized. It doesn't dump raw data. It returns a preview instead. You can filter and sort it. Then you pull the full data into your coding editor. We’re trying to innovate in these specific areas. One thing really caught my attention here.

When we talk about doing things with AI, there’s no clear metric for quality. Take content creation. You need no hallucinations. It must showcase the product clearly. It should match the brand voice. And you don't want trouble with Google. On paper, ChatGPT makes content creation look easy. But once you dive in, it gets nuanced. Where do the agents get their data? Are you using ChatGPT’s search or your own custom sources? These distinctions really matter now. Quality work requires serious testing. You have to evaluate the output. We’ve reached a point of saturation. There are too many products now. Software users are feeling real burnout.

Remember 5 or 10 years ago? When a new tool hit the digital marketing stack, managers were hyped. Everyone was curious. They couldn't wait to tinker with it. I don't know if it's just us getting older. I don't see that same spark in younger people. I blame AI for this. There’s a genuine sense of fatigue. People just aren't that curious anymore. But the industry pushes us that way. Anyone can build products now. I’m sure you’re coding your own daily shortcuts. Software is becoming deeply personalized.

It’s a chaotic time. I want to ask Can about this. Can has been working on content creation for a long time. Let’s make this concrete. Think about a content creator versus an AI agent. For an e-commerce owner, what’s the difference? Is it better to build an agent or just use ChatGPT? What are the major practical gaps? We discussed a different meeting context today. We looked at sources. Let's make that concrete. People think big products like cloud or ChatGPT do it all. With code execution, agentic capabilities, and MCP, they think one interface handles everything.

What are the practical differences, Can? I’m curious about your experience. Where does bringing in the context actually make a difference?

— We always talk about this. Everything is context. The better you define your needs, the better the output. This applies to more than just content creation. It applies to everything you do. It’s true for vibe coding and SaaS development. We’ve spent a year experimenting with content creation. We tested many approaches. We started back in the GPT-4 Turbo era. Models had narrower context and less reasoning then. I look at the content creation process through that lens now. Sure, you can use the ChatGPT interface for product descriptions. But the depth of that content depends entirely on you. It all comes down to the context you provide.

Google always pushed for value-added content. Even before GEO or AI. You need to stick to that strategy. Think about MacBook M1, M2, or M3 content. There is already so much out there. What makes your content different? What value do you add? How do you stand out from competitors? What are you providing in your informational content? Using ChatGPT with AI is faster than a human writer. But you still have to research the info manually. Or you have to keep pushing data to the AI. You need to chat first. Remind the AI of the context. Then write the content.

We start our content process with research. We don't focus on writing first. We ask what we know about this topic. What did the WHO say? AI models know this, but we research it first. We force the AI to focus on our core goal. Then we check what competitors wrote. We analyze their content. We filter it all out. Writing is the very last step.

— We can make this concrete, Can. Let's use one of our MCP server setups. This might also answer Eda's question. I can give an example. It fits your point about bringing context together. Before the presentation, I’ll try to mirror what you just said. Take this prompt. It’s pretty standard. Nothing special here. What concrete examples can you give me on using e-commerce and AI together? Can you search Reddit and make 20 tool calls for me? This is actually a key feature of the MCP we built with Can. If you’re researching a topic, Reddit’s context is usually very reliable.

Like Can said, if we were doing health content, we’d pull data from Reddit. Our MCP server grabs keywords generated by AI from Reddit. AI suggested about 20 keywords here. We told it to only show results after a certain date. Then we get our search results. It’s JSON, so it might look a bit messy. But it’s aggregating 20 different search results. It filters out the recurring results. AI isn’t satisfied, so it makes another tool call. The agent researches again. It mentioned things like abandoned carts and e-commerce conversion rate optimization with AI.

To get more context, AI picked URLs from Reddit. These came from the MCP. The result is a very strong text. It covers different areas. What do we have here? "How crucial is AI for e-commerce right now?" How critical is AI for e-commerce right now? It scanned tons of Reddit threads across various niches. This MCP server returns the full content. Normally, you can't get this depth from ChatGPT. Even with screenshots, it’s really tough. What do we get in the end? We see what’s trending in e-commerce. It even cited a voice AI example. We get grounded outputs. Everything is based on real user discussions.

Prompt engineering is shifting toward context engineering. But explaining this to someone new to MCP server is tricky. It’s a bit of a hurdle. I wanted to make this concrete. You mentioned product comparisons. That triggered a thought. Take the new M4 or M5 Macs. If you ask ChatGPT for user reviews, it struggles. The product is too new. ChatGPT can’t easily crawl blogs. Its Reddit access is limited. I wanted to show how we inject that context. Sorry for stealing the spotlight. I went on a bit long.

— Actually, I’m glad you showed that. It ties back to our earlier talks. We used to focus on few-shot prompting. Then we moved to chain of thought. That was the gold standard. Then we saw AI models start reasoning on their own. They do their own chain of thought.

— Yeah, nobody even talks about that anymore. That type of prompting is really gone.

— I cover it in my training. But people do it subconsciously now. LLMs handle it, so we don't discuss it. We focus on the core issue. You say it best. If you don't get the right answer from AI, you're asking the wrong question. Exactly. To get the right answer, you must explain your needs clearly to the AI. It all comes down to context. What are you explaining first? If I can't explain something well, I use my go-to method. I ask the AI how to explain it. I let the AI guide me. Then I ask the AI again. I saw a tweet saying stop writing prompts. Let the AI write them. It's true to an extent.

We still need to step in for context. But you're taking it further. You're talking about research.

— Notice how the question-asking pattern is exploding in agents lately? Cloud Code added a question tool. Codex and Copilot did the same. It’s all about gathering context from the user in chunks again. Starting and iterating by talking with AI is super logical. It’s one of the most practical techniques right now. I’m tracking the questions. There’s a great one here. It’s perfect for Tahir. I’m curious about his vision for the future. Will AI agents go beyond taking orders? Can they compare prices, pick the best option, and wait for discounts? You can actually do this today with a few MCP server setups.

Just grab an MCP server for a site like Akakçe or Cimri. You could even write your own MCP server without touching search. You’d find the best price instantly. Let’s be real about where this is going. It feels like it’s getting harder for small shops to compete with giants like Teknosa. Where do you think this takes e-commerce, Tahir? When a user finds the best price for an E-ink tablet instantly. How does that change the market? Will users focus only on price, or will experience still matter?

I’m curious about your take. This actually scares me a bit. Are we heading toward massive monopolization?

— Let’s start at the end. It doesn’t scare me at all. I don’t think much will change. Look at Google. They’ll likely integrate a tool to handle everything. You’ll buy directly through them. No more shopping links. Just pay with your credit card right there. That’s the talk. We pay Google a commission. They take their cut. E-commerce evolves. It was different ten years ago. We started with our own sites. Then marketplaces took over. Then social media hit. Brands went from ignoring TikTok to chasing Gen Z there. Things are always shifting. These are just innovations. I remember this from my camera days.

When cameras first shot video, everyone hated it. They said cameras are for photos only. Same with washer-dryer combos. People always want separate machines. Humans always have a defensive reflex. We saw it with ChatGPT too. We’ve been through this. We’ll keep going through it. But local shops will always find a way to survive. This happens in inflationary markets like Turkey. Holding onto inventory is often more profitable than selling. It’s a unique situation. That’s why some massive global giants haven't fully scaled here yet.

They do it easily elsewhere. In other countries, pricing is stable. Everyone follows the same price points. Here, finding the cheapest price for a product is already easy. We can already find it. Sure, our click volume is high. Searching for the cheapest phone gives instant results. We’re past that stage. But price isn't the only factor. We have issues with fake or replica products. We see these problems more in developing markets. Trust is a huge factor. Review counts matter.

— Support is key. Having someone to talk to for returns is vital.

— Easy returns are essential. Sellers often refuse them or skip invoices. Anything can happen. Everything will evolve. We will evolve too. But trust me, the rest of the world isn't like us. Most people still act in very primitive ways. It’s all very standard. We’re in a bubble. My Twitter timeline is nothing but AI talk. It’s honestly unreal. But that’s just our circle. The algorithm feeds us what we engage with. The real world isn't like that. I go out and see the same old people. That won't change.

Remember when they said e-commerce would kill retail? That physical stores were dead? Look at the data. Even in China, it’s only 50%. In Europe, it’s around 30%. People still love shopping in person. And they’ll keep doing it. Not much will change. Sure, percentages fluctuate. Pandemics happen. That’s a different story. Maybe AI makes Google shopping faster. But most people crave human interaction. They like talking to people. They enjoy the social aspect.

— I hope this transition takes time. I’m glad you’re an optimist. I’m so detached from it all. I was even surprised by the e-commerce boom during the pandemic. I thought everyone already knew and used it. We’re stuck in our own bubble. I saw so many people back then. People making their first card payment ever. It popped up in 2020. It was a total shock to me. I assumed everyone started e-commerce in the early 2010s. Maybe not the 2000s, but definitely by 2010. I thought everyone had cards by then. Like you said, we need to look past our own bubble. Let’s move on to some questions. I don’t want to miss the ones sent in earlier.

We built SEO around user intent for years. If agents like ChatGPT start deciding and acting, do we optimize for humans or agents? Did we really focus on user intent out of respect for the user? Or were we just trying to satisfy an algorithm built on user intent? This question really makes me think.

— I don’t even have to think about it. We can just say yes. At the end of the day, in SEO, we did what Google said. Google kept pushing user experience, especially as AI started entering the picture. It was always user experience. Then came Core Web Vitals. They kept adding new metrics. We did all that for Google to rank better. But yeah, it helped the user experience too. We built features for Googlebot. But great titles and descriptions improve UX too. As people use ChatGPT or Gemini, this experience will only get better.

Providers are clear about this. It’s all about UX, personalized ads, and custom answers. ChatGPT now uses memory. It tailors answers and product suggestions based on that. The flow is moving toward UX. But you can't personalize everything for everyone. For SEO, the core process will stay stable. It’s already built around user experience.

— I agree. It’s about maximizing brand awareness. Small players have it tough. Search results used to be a list of ten sites. That was the standard. It was more open by nature. Now, I see ChatGPT ads. I’m testing the free version of ChatGPT. OpenAI has deals with companies like Best Buy in the US. Ads are appearing slowly. But it’s only one or two slots. Users want direct answers. I’m not optimistic there. But giants like Trendyol will still find opportunities. Like Tahir said, huge e-commerce giants entered Turkey years ago. Everyone expected them to dominate the market instantly. But every market has its own unique dynamics.

So, maybe we shouldn't be too pessimistic.

— I have a take on this. I’m not exactly an optimist here. But mobile app usage in Turkey is massive. For e-commerce, users prefer marketplaces. I don't think they'll ditch apps for ChatGPT to buy things. Here is what might happen. Eventually, marketplaces must integrate with these UIs. They will need to build their own MCP server and connect. That is the future. It could evolve this way. Users will search for products in ChatGPT. They'll check features, like a sequined dress. Then they might jump to the mobile app to buy.

Or maybe they’ll link ChatGPT directly to the mobile app for checkout. But I don't see most users abandoning mobile apps anytime soon.

— We’re lucky. Apple exists. They’ve lived without AI for a long time. They’ll likely stay that way for a while. I recently bought an Android to test it out. Used it for a month or two. Maybe it’s just my inner boomer. I couldn’t get used to it. I switched back. But Gemini on Pixel is flying. The AI features are incredible. The gap between Apple and Android is night and day. There’s this new thing called WebMCP. It’s not well-known yet. It’ll change mobile behavior a lot. But you’re right. Think about my sister. She doesn’t work in digital.

She’s not in tech. She just clicks an app to shop. Amazon dominates the US. You pay 20 bucks for Prime. Shopping elsewhere feels pointless. Returns are easy. It’s a total monopoly. But this mobile app stuff gives me hope. What do you think, Tahir? You know traffic well. You talk to other e-commerce managers. As long as these mobile habits stick, AI adoption feels like it’ll be slow.

— I think I’m more traditional than you guys. I don’t see an easy transition here. People in Turkey won’t easily buy through ChatGPT. At least not at first. Plus, ChatGPT’s future is still unclear. Usage rates are dropping, you see it too. It’s number one because it’s on everyone’s phone. But it will work like Google Ads initially. You’ll click through. I’m with Can on this. It will likely redirect you to the site’s app. If there’s no app, it’ll send you to the App Store or Google Play. We might use these spots for CPI. Like dedicated app install zones.

Honestly, I wouldn’t buy a product directly through ChatGPT. Because of the past.

— Maybe you would. You might get to the point of purchase. Researching there is fine. But actually buying? That feels weird to me.

— You need to visit the site and build trust. You often need info from the site itself. We don’t just look for product specs. We check shipping details too. When will it ship? Which carrier do they use? Sure, ChatGPT can integrate that. But the product isn't the only thing that matters. How does it ship? When will it arrive? Can I return it? There are so many questions. I even check things myself. Yesterday, I was buying swim trunks for my son. You can't find those everywhere. If you search on Trendyol or Hepsiburada, options are limited. But there are niche swimming sites out there.

Sites that specialize in this. You visit them and want to reach out. Is this shop legit? Where are they? You get what I mean? These processes exist. Not everyone is buying from big players.

— They could be ghost sites. Maybe they opened years ago and no one checks orders.

— I deal with this too. What if ChatGPT loves a site and starts recommending Ahmet.com for everything? Things get messy. There are legal sides to this. How will they hack that? It’s tough. E-commerce pros have been stressing over legal stuff lately. So many new regulations. More are coming regarding AI.

— I follow the regulations in every country. Everyone is going through a regulatory process. It’s the same on our end. KVKK is becoming a real issue. We don't just look at this from a legal perspective. We focus on e-commerce, SEO, or AI. But KVKK is a factor. It will hit the purchasing stage. Will credit card info be entered? I’m thinking about common ground. It will show the product and redirect. I think it will evolve. Maybe I’m thinking too far ahead. I have a gut feeling about where this is going. I bought a perfume. My agent will reorder it when it’s about to run out.

Or it will ask me. It will say, "Your perfume is low, buy it." We can predict this now. But it will eventually make decisions for me. How long will that take? Will people actually use it? Trust is the issue. You still want to check things yourself. You don't just trust the site. You check reviews on the marketplace. We all have this trust issue. Is the product legit or not? People will want to verify the supplier at first. It will create a lot of doubt. There is a negative sentiment toward AI. Despite its benefits, being anti-AI is becoming popular.

People might not want AI to handle big purchases yet. But I saw something cool at a recent hackathon. They’re building wild new protocols. Take X402. It’s a protocol letting agents pay other agents directly. It’s token-based. Coinbase is backing it, if I’m not mistaken. They’ve really nailed the infrastructure. It looks like the new standard. Trade volume is already starting to pick up. It’s small, but it’s growing. This is a trend worth watching. I also want to talk about WebMCP. We’re building a product for it. Check out WebMCP.net.

There’s a hyphen in there. It launched about a week and a half ago. You just add specific schema tags to your site. It wraps the site’s search bar into a Search Product tool. Now, an agent can search your site directly. No need to write a custom MCP server. It can pull product details by ID. It can browse categories or add items to the cart. It can apply coupons. It can even track orders. These protocols solve so many of our concerns. Seeing this kind of progress actually makes me a bit nervous.

I hope things slow down a bit. I track every AI development. I love it. But I bet you feel the same. It’s good if things slow down. People need time to adapt. Companies can't innovate too fast due to regulations. Or Zuckerberg spending millions to poach researchers from OpenAI. That feels positive to me. Power shouldn't be concentrated. It’s healthier if it’s spread out.

— Here’s the thing. New models keep dropping. Vibe Coding is getting faster. When Gemini 2.5 Pro launched, we all said it was amazing. That was only back in July. It actually launched in July 2025.

— Then 3.0 arrived. You aren't using the old one, right? It became obsolete within a month.

— We’re used to that in the AI space. Those of us building with AI have a fast-consumption mindset. Once a new tool drops, we’re done with the old one. If there’s a better version, we ditch the old one immediately. I think of it like this. We do the same thing. We compare Claude Sonnet 3.5 with Codex. Codex wins. Let’s switch. Let’s check out Windsurf. Let’s move there. We pivot to whatever is best. New stuff drops constantly.

— AI'ın has a brutal consumer base. Power users switch the second something is slightly better. They jump ship instantly. It’s the kind of customer you really don’t want to deal with.

— Exactly. We think everything is coming fast. Maybe that’s just us. I don’t know how others think. We expect shopping on ChatGPT with credit cards to arrive soon. I still think it’s coming fast. It’s not about adoption. I think this integration will happen very quickly.

— Let’s mark this date. Will you make your first purchase on ChatGPT by 2027? At least in the US. Europe might take longer. Tahir, I know you have to leave in 10 minutes. One last question for you. I’m curious about your take. We talked about how fast AI'ın is moving. Everyone went into a frenzy. They said live support would vanish in 3 to 6 months. They hyped up voice agents. When I talk to AI engineers or founders about voice agents, they’re hyped. They thought people would adopt voice agents instantly. They assumed everyone would love them.

But look where we are now. Even with lower latency, I doubt business owners will trust them with high-value leads. What’s your take on voice agents? Can they work in e-commerce support in the short term? Support usually means dealing with angry customers. Where do you think voice agents will actually fit in?

— We’ve started using them. Many companies probably are too. People hated standard text-based chatbots. I never liked that tech either. I’d ask about A, and it would ramble about Z. Voice agents are much more consistent. Sure, there’s still some latency. I speak, it processes, then it answers. But I have friends in call centers who are deep into this. It’s directly impacting their work. I heard a demo recently. It processes while I’m still talking. They’ve really sped it up. It’s not quite human yet, but it’s like a person with a lower IQ.

It can keep the conversation going.

— How’s the vibe? It tried Turkish. Is the tone right?

— Eleven Labs just integrated their voice agent tool. They killed a bunch of startups. That’s a real problem right now. Businesses are dying left and right. But companies keep using it. If you’re asking if I’d use it for a specific lead? No, I wouldn’t. But if I’m running a call center? If 95% of calls are for the same thing, there’s no real risk. It depends on your business model. But everyone agrees this is where AI’ın is making the biggest impact. It’s an easy transition. The cost savings are massive. I’m hearing 80% to 90% cuts. Those are huge numbers. Big companies might even trade a little customer satisfaction for that.

I think it will be used. I’m on the same page as you. That transition will be fast. But I don't think the other one will be.

— You have to separate inbound and outbound calls. Outbound will switch to this very quickly. Think of marketing a product or selling a course. You can automate those outbound calls. I’ve already seen a few examples. Leads likely to convert go to a human. The AI hands it off or captures the lead. It’s all about saving time. The biggest problem right now is latency. It’s starting to get solved. Once latency drops, call centers will switch to voice agents. The delay is still super annoying.

Turn detection is also a pain. Did you actually start talking or not? It’s like a Zoom call where everyone talks over each other. It feels awkward with AI. You talk, AI stops. You keep talking, AI starts again. It feels totally inorganic. I wonder when the first big voice agent scam will hit. I have high hopes for Turkey on that front. I’m expecting something big. We’re a brand when it comes to phone scams. We have creative methods. They’ve become cliché and rigid. I think there’s something good there. It’s doable even at the current level.

Maybe those folks don’t have much Vibe Coding experience yet. If you hear anything soon, send it my way. I’m curious how that side evolves. Let’s wrap up with one last question. Tahir has to head out. We shouldn't take up his time. Melike Kalkan asks how omnichannel marketing and personalization tech evolve with AI. What are the constraints and opportunities? This one goes straight to Tahir. I can’t think about omnichannel from that broad perspective. You’re already working in the perfect spot for this.

— I hear things from friends in the industry. Omnichannel is very advanced in the Dubai and Arab regions. I know a store there that does this perfectly. Maybe it exists in Turkey too, but I know a great one there. They connect the web and the store seamlessly. There are reasons it’s not done in Europe or the US. Mostly security concerns. But they use store heatmaps to rearrange layouts. They track how long people stand in front of each display. It’s not just about customers. We also have promoters on the sales floor. For those who don't know, brands like Samsung send their own staff. They send people to sell their specific products.

Apple does the same thing. We track their productivity too. Cameras now measure if shoppers are male or female. They even track average age. We get the same insights as Google Analytics, but for physical stores. You can even segment the data. I’ve seen their dashboards. They can tell you how many males aged 20-25 are looking at phones. It’s actually pretty impressive.

— That’s wild. It detects age and gender from video. It links that data directly to specific product displays. That is really powerful.

— Stores are becoming super data-driven. They know exactly what to stock and sell. It connects to omnichannel strategies too. They can even track this for web advertising. They can track if a store visitor uses an Android device. It’s advanced device-to-device tracking. But these investments are very expensive. These camera systems cost a fortune. Not everyone can do this. Are you calling it AI?

— Is this real-time, Tahir? Or is it end-of-day?

— It’s not real-time. Maybe they improved it recently. But top executives have these dashboards. They just watch them. It’s like how we use BI tools. It’s a killer feature. I loved it when I saw it. They did great things in e-commerce. They built these lines. I’d love to see more of that. Think of a number line on the left. The product moves like an animation. It hits my warehouse. Then it travels to the user’s city. It’s truly impressive. Products that wow people will be the future.

It could be the next hype. We all use the same tools. No matter how advanced they get. Omnichannel marketing and personalization depend on trust. We still can’t break those security barriers. Especially with EU regulations. We can't track age via camera here. We can't even identify returning customers. You can't recognize people. You shouldn't, anyway. We can't save credit cards either. There's so much we can't do. These limitations hold us back. Other regions are more open. They move faster. They're much more data-driven.

— For sure. I can't even imagine the EU. Regulations in Germany or France are on another level.

— You said Germany, I get it. It's impossible there. Totally impossible.

— I don't know if AI technologies will force them to loosen up. Regulations are starting to hurt Europe too. China is the perfect counter-example here. Training data is completely unrestricted there. It’s a surveillance state with cameras everywhere. That massive scale helps them dominate in AI and video processing. That was a great hour and a half. We covered a lot of ground. I hope my amateur moderating kept the flow natural. Thanks for joining us, Tahir. We really enjoyed the chat. Let’s wrap it up. Thanks to everyone for tuning in.

— Thanks, Tahir. Great having you here. Appreciate it.

— Thank you. See you all later.

— See ya. Take care.

This transcript was generated using AI speech recognition and verified against subtitle style rules.

Burak Pehlivan

Content author

Burak Pehlivan

both

Contact us

How we handle your data

Open as a full page