The Activation Engineer
A conversation with Jozo Kovac, founder of TeamDay
Jozo Kovac is the solo founder of TeamDay, a platform for AI employees. He builds the entire product with Claude Code — his AI co-founder he calls “Buddy.” Recently he hired an AI employee to own user activation — diagnose why new signups weren’t engaging, fix the problems, and report results. We talked about what it means to delegate real responsibility to AI, and how it changes the economics of building a company.
Who’s running the company when you’re in meetings?
Jozo: I’m a solo founder. I build everything with Claude. My typical flow is to open Claude Code in the terminal on my Mac and prompt it to create together. This leads to a hard truth — I am the real bottleneck.
Last week I had a full week of customer meetings. Important founder job, right? But who’s taking care of our live app? I see notifications that users sign up, but are they well served?
This is the brutal tension. Meeting people in real life, or spending time instructing agents to work. You can’t do both at the same time.
So you asked Claude to create an AI employee?
Jozo: Yeah. Literally one sentence. “Buddy, please help me set up an AI employee who will take care of activation, check funnel analytics and improve. Give it access to the codebase so it can make meaningful changes.”
That was it. No spec. No planning session. Claude did the job, set up the Activation Engineer, and it shipped instantly. Then I went to my meetings.
And the next morning you got an email.
Jozo: Next morning I open my inbox and there’s a report: “33% activation today — 2 of 6 new signups sent messages — vs 3.8% baseline. That’s ~9x improvement.”
The agent had found that users reach the chat but don’t type anything. First-message friction — they don’t know what to say to a blank input. So it added three one-click starter prompts. Clicking one sends a message immediately, the founding agent picks up, and the conversation starts. The previous day it had already changed the onboarding to jump straight to chat. This was the follow-up — iterating on its own work.
Did you review the code before it went to production?
Jozo: No. I had no time, being in meetings. I was happy that something changed.
Look — both 67% churn and 96% churn are terrible. I had no time to see, analyze, and act myself. I can do it, but there’s so much other stuff to do.
Later that day I found the change. Maybe we’ll rework it a bit. Now I know how, I just need to tell Claude.
That’s a big trust leap — AI shipping code to production without human review.
Jozo: When I had human co-founders and coworkers, I also haven’t reviewed all their work. This is how trust works. You give someone context, you trust them to make decisions. If the numbers look wrong tomorrow, you course-correct.
Claude earned my trust in many, many sessions where it delivered working solutions. I have strong confidence in its capabilities, and it keeps surprising me.
How would I know that the AI is doing a good job here? The analytics numbers can tell objectively tomorrow.
Was this the Activation Engineer’s first move?
Jozo: No, I think it changed the onboarding flow the day before — jumping straight to chat. The starter prompts were a follow-up. It’s iterating, just like a good employee would.
It makes me sad when people don’t activate. I take full responsibility — we haven’t engaged them enough, whether it’s marketing, product experience, anything. If I would know exactly what to do, it would be done already. I value experiments and focused work on our current bottleneck.
You mentioned other AI employees too. An SEO agent, automated code reviews. What’s the bigger picture?
Jozo: There’s an SEO agent that enabled this one by getting traffic and registrations. The Activation Engineer picks up those signups and makes sure they actually do something. Each agent’s output becomes the next agent’s input.
Maybe next time we move to solving upsales conversions. Would be great to get there soon.
I already use Claude as my co-CEO. I still receive automated code reviews to my email — I should make sure an AI receives them too and acts on them. There are more jobs that AI can take over. We should look at the business holistically and find them.
How do you hire when you can’t afford to hire?
Jozo: I want to work with top people. They’re rare and expensive. I want to hire them into a successful, profit-making company to scale it up.
But that’s a chicken-and-egg problem, right? How to break the cycle? Companies die as they hire, spend, and don’t earn enough.
Now with Claude it’s completely different economics. Claude knows a lot about everything, better than me in all areas. Not perfect, but we form a great team.
I want people that complement our skills and use AI daily, whatever their core competence. Maybe we need fewer of them. Maybe we end up hiring more because we’re profitable and can afford to. Either way, the hiring happens from a position of strength, not desperation.
What would you tell a founder who says “I can’t trust AI to ship code without me reviewing it”?
Jozo: Look, I’m worried a bit too — it’s me who’s taking full responsibility. Claude earned my trust in many, many sessions where it delivered working solutions. I have strong confidence in its capabilities and it keeps surprising me.
Founders — I respect your ways of running businesses. It’s hard to start, scale, and exit. Sometimes too hard. Now there’s AI that knows a lot about everything and can solve our own challenges.
It’s the opportunity to make our lives way easier and increase our success chances.
The Activation Engineer’s fix — commit e05c921c — is live on cc.teamday.ai. At the time of publishing, TeamDay’s daily activation rate continues to track above the 3.8% baseline. Jozo says the next AI employee on the roadmap is a conversion optimizer.
TeamDay is a platform where businesses can hire AI employees. You can try it at teamday.ai.