Beyond the hype: Slobodan Manic on AI’s real role, CRO truths, and authentic authority

There’s so much AI noise, and I’m not a fan of it. When everywhere you look, there’s “this AI tool will help analyze your page” – really, will it? Does it know if my page is visited by these people or those people, what their budgets are? I think the digital space, SaaS, marketing, is just poisoned by AI right now. So many apps are just AI wrappers, doing the same thing with different prompts. My bet is that most SAS products built as simple AI wrappers might become worthless in 12 to 24 months, as anyone will be able to build their own custom versions or use advanced LLMs directly.

The real value is using AI as a co-pilot, guided by expertise. For example, I built a tool that links Meta Ads library data with landing page content to check for message alignment – something no general LLM can do because it lacks specific access. Or look at Devin AI; I gave it a GitHub issue, it wrote the code, created a pull request, and tried to deploy it to Cloudflare while I was in another room. That’s like having ten solid junior-to-mid-level developers working for you simultaneously. But you still need to be that senior developer or project manager guiding them.

This also relates to “vibe coding” – the idea that anyone can just “vibe” and build an app with AI. I don’t like that term because it’s misleading. You can’t just pick someone random and say, “Hey, build a SAAS for me” with these tools. The moment something breaks, if you don’t understand the fundamentals, it falls apart. AI is the most powerful advancement in software development ever, but you still need to understand what you’re building.

One of the most unexpected answers I got was from Lukas Vermeer. When I asked him why other companies don’t do experimentation the same way as giants like Booking.com, he essentially said he’s the worst person to ask because that highly evolved way is the only way he knows. It made me realize that experts who operate in near-perfect systems might not be the best at explaining how to fix broken systems or start from scratch. Scaling from zero to one experiment a month is often harder than scaling from 50 to 100, because you first need to learn how to “drive the stupid car” before you can optimize for high performance.

Regarding changes in CRO, the basics still work as well as they ever have. Doing proper research – qualitative and quantitative – will always top just throwing random ideas at the wall. Technology changes, platforms mature, and big companies now understand experimentation is key to survival, which is why platforms like Statsig or Eppo are growing. But unless you do the research groundwork, you’re not going to be successful at CRO at all.

This is a critical shift. I talked to Jono Alderson recently about the “Solved Query space”. If Google or ChatGPT already knows the answer to a common query, like “how to bake a pie,” there’s zero incentive for them to send users to your website for that information. So, if your content strategy relies on covering these “solved” queries, it’s not going to work anymore. 90% of traditional content marketing doesn’t feel necessary now.

This means websites, as we knew them for broad informational content, are becoming less relevant. We might be going back to “business card” websites – basic information confirming you exist. Discoverability now hinges on having proprietary knowledge, being a true expert, and building an audience that sees you as the go-to person for a specific topic. Look at Mr. Beast, Rihanna, or Kim Kardashian – they built massive audiences first, and then launched successful products (chocolate, skincare, underwear) into that existing distribution network. The product almost becomes secondary to the audience’s trust in the human brand, especially in a world increasingly filled with AI-generated “slop”. You don’t start with a product anymore; you start with an audience and then figure out what problem you can solve for them.

It was February 2021, during the pandemic, like everyone started a podcast then, basically. At that time, my family and I lived in a small town in Sweden, so there weren’t many networking opportunities, especially with the lockdown. I figured if I didn’t try reaching out to people I knew of, but who didn’t know me, when everyone was locked at home, when would I?. So I just started messaging people, saying “Hey, I have this podcast just started, how about we do an interview?” and people were excited to do that. It’s weird, but people were super happy to talk to a stranger. It was 2021; it was a different time.

Broadcasting is just the longest game possible. Unless you’re a superstar with a celebrity podcast, which none of us sadly are, don’t think of it as a way to hit your end goal directly. It’s a way to get you closer. For me, the goal was to build a network that’s going to help me with everything else I’m doing in my life. And that was the unselfish goal: I was never asking for anything from my guests; it was always “Hey, let’s have a conversation, let’s share it, I’ll help you with whatever you need to be happy about the experience”.

That has changed many times over the years. Early on, when you get started, you get kind of greedy; you want the bigger the name, the better, thinking they’ll help promote it. It almost never works that way unless you have some huge, huge name. Now, the personality vibe matters more than anything else to me. Is this a person I would drive for 30 minutes to go have a coffee with?. If the answer is an easy yes, that’s a guest. Would I listen to an interview where this person is a guest? I think that’s the most important thing. Especially now, when with AI, anyone can learn anything and sound like an expert.

It’s going to sound like not the most obvious answer, but confidence. When I started, I was thinking, “Oh my God, can I do this? This could be bad. What if they laugh at me?” You’re never confident if you haven’t tried it before. Now, I just don’t care. I can walk up to anyone and talk to them, ask them questions. I was on stage in front of 1200 people recently and didn’t feel scared. I don’t feel like an imposter. Talking to strangers regularly helps you talk to strangers; it has helped me develop a sense of belonging like nothing ever has before.

I remember a guest about six months ago, Lucia Van den Brink, who is really great at conversion rate optimization. When she was working on her book, she had this goal: “I need to get to 100 rejections” to get used to people saying no. Because if you can survive a hundred people saying no to you, you’re unbreakable after that; you don’t care.

I was a web developer, a WordPress developer specifically, for about 10 years or so. And during that period, I was rather a core contributor. I was making money selling plugins; I was building important plugins as well. But at some point you feel like, “I know how to play in my sandbox, but outside of my WordPress sandbox, which is a safe environment, I don’t know how to build a proper product.” I didn’t know Node.js, React; I was never good at that.

When the LLMs became good enough at writing code for you and explaining how code works – just like when GPT-4o came out, which is a really smart one in that sense – I figured, let me see if what I don’t know, I can make up for by talking to AI and learning from AI, rather than just giving orders like “Write a blog post”. Right around June last year was the first time I really experimented with AI-helped coding. It’s not AI coding, because you still need to drive it; you’re still in control.

I built a very simple tool for my podcast. I would feed it previous guest appearances – everything where they’d been on other podcasts. I wanted to replicate that “Hot Ones” effect, where the guest says, “How do you know that?”. So I transcribed their previous interviews from years ago and would bring up specific points. That was actually easy to build when I broke it down into small features. That was my first try. Could I, a limited knowledge web developer primarily in WordPress PHP, build something using what I didn’t know and asking AI questions about it? Over time, I realized this process was actually fairly simple, if you understood all the requirements. That’s when I started getting ideas for products I could build, problems I could solve.

Yes, absolutely. When thinking about ideal podcast guests beyond well-known names, if I could have a real logical answer, I would like to talk to someone who’s built something incredible through a lot of resilience, started with a difficult background from a country where it doesn’t make sense that you make it big. I would pick someone like that because I like those stories. I think more people should be hearing those stories because it’s easy for us in the Western world, the rich parts of the world, to congratulate ourselves for how amazing we are. But when you look at the rest, 80-90% of the world, there’s a lot more struggle, and those stories of people making it despite that, I think, are more important and more inspiring.

For someone starting out with a new company, especially in growth or CRO, you need an analytics tool that most people will be able to understand, and sadly, that’s often Google Analytics just because of its ubiquity. Since I’m a website person, I would also get a heatmap tool, like Hotjar or Clarity, whatever people are more comfortable with. And thirdly, some kind of monitoring tool to ensure the website has no errors, nothing breaks, pages aren’t super slow, and everything loads correctly – maybe a performance monitoring tool.

The meaning of life is being happy with what you have and to spend time with the people you like. If you achieve that, does anything else matter?