I started in sales, which surprises people. I was selling carpets door to door in California and around the US, back when Skype was just appearing. I was always really into tech, always had the latest laptop, and my friends would laugh because they knew I'd have the latest gadgets. When I moved to London I realized that selling people things they didn't really need wasn't very ethical, and that's when I discovered marketing, then got into growth.

From there I became a growth consultant in Barcelona, mostly for B2B SaaS, but a lot of it was deep tech. I was actually working in computer vision and AI about nine or ten years ago, when AI wasn't cool at all. I loved deep tech, but I'm not a developer and I'm an extroverted person, so I couldn't be the coder sitting at home all day. That was tough for a while.

Then ChatGPT came along. At the time I was the CEO of a 110-person sales outsourcing company, a big role and a good salary, and I left all of it. I quit because I saw the potential in these large language models. It was pretty crap back then, honestly, but I could see where it was going, and I'm glad I jumped. I founded AI Operator about a year ago and it's been one of the fastest growth journeys I've ever been part of. On the side I'm also a sound engineer, music producer, and DJ. That's kind of me.

I call it a mindset on purpose, because it's a complete change, and I honestly think it's a bigger change than electricity. Think about London a long time ago, where people would walk the streets at night and light the torches so we could see. That job disappeared, and we learned new ones: electrical engineering, all of that. It's the same shift now. We have to relearn how we work.

So it's not just here's ChatGPT, here's prompt engineering, off you go. It's asking yourself, every time you do a process, how can I do this with AI? How do I find a document to feed it? How do I speak to these machines? Prompt engineering matters, context engineering is becoming even bigger, and there's so much that goes around it.

That's why when we train people, we train the mindset first and get into the tools after. The tools will keep changing, you'll always get a new one, but the mindset is the thing that lasts. And the goal is that people stop fearing this technology and start to love it, because once you love it, you use it for the benefit of the people around you instead of bracing against it.

The mistake I see most often is a leader playing with ChatGPT over the weekend, getting excited, and then giving the whole team a login and saying, here you go, be super productive. Then they come to us, we run a survey, and we find only 20 or 30 percent of the team actually uses it properly. They're shocked. They paid for the teams plan, so why isn't everyone flying? Because this isn't HubSpot, it's not just another CRM. It's a completely different way of working.

So you can't just hand out access. You need a bit of training, some change management, often a third party to come in and help. And the way we approach it is to make everyone an AI champion, not to appoint one chief evangelist. We want every team member to stop fearing it, to see where they're saving time today and where they could save it tomorrow, so they all help each other and discover new things together.

The principle underneath all of it comes from something Dharmesh at HubSpot said that really stuck with me: try everything with AI first. Every task, every project, every question, attempt it with AI. If it fails, fine, you learn why it failed and what the tool is good and bad at. While you're waiting for perfection, other people are already running with it.

I fell in love with all of these tools and lost a lot of sleep trying every one of them. On their own, they're not powerful enough yet for incredibly stable, high-stakes products. But they're fantastic for low-stake internal tools. Why buy a giant CRM and use twenty percent of its features when you can take a weekend and build your own, custom to exactly how you work? If it breaks, it's not the end of the world, it's an internal tool. I've built a client portal, a training portal, an AI ROI calculator, an AI readiness assessment. Great lead magnets, and low risk.

Where it gets really good is the combination. The lesson isn't the right tool, it's the right set of tools. I'm not a developer, so I'll start a project in Replit because it sets up the folders, the backend, the frontend, and the auth, all the structure I'd struggle with otherwise. Then, inside its terminal, I run Claude Code, which is the best thing I've seen for refactoring, breaking pages down, and cleaning up the code. Replit and Lovable can't do that part, and Claude Code can't do the scaffolding part as easily, so together they're unstoppable.

The deeper shift is that building isn't the hard part anymore. What matters now is creativity and the ability to articulate your vision so the model can understand it. I had an idea over a weekend for a habit app for my four-year-old daughter, posters of unicorns and whatever she likes, running on a tablet on the fridge so she builds good habits. I built the whole thing and called it Routine Magic. Two years ago I'd never have touched an idea like that.