I'm on my own again

In the last essay, I wrote about learning that customers wanted results rather than a system, and starting an agency service to build a product. We decided to do the work ourselves to learn what an AI employee should take on.

Along the way, my cofounder and I decided to go our separate ways.

We saw the same agency work differently

As we were about to start the agency service, he said he did not want to do agency work. He was not drawn to working around customer requests, and he felt it would be difficult to scale the business.

Building a larger agency was not my goal either. But I believed we needed to do the work ourselves before we could build a product to take it on. We first had to understand what information customers provided, what they wanted to hand over, and where human judgment was needed.

I agreed with his concerns about scalability. If every new customer meant adding more people, we would be moving away from the company I wanted to build. Calling agency work part of product development would not change that on its own. We had to identify repeat tasks and actually move them into a product.

I wanted to work through that process. He was more interested in using AI to change how work was done in another field. We talked, but could not resolve our differences. In the end, we accepted each other's choices and worked through the handover and wrap-up of what we had done together.

I was deeply disappointed. Still, I felt it would be difficult to keep working together by continually persuading him to do work he did not want to do.

I decided to stay with this direction

Before, I had looked for opportunities by trying many different things. This time, I wanted to explore one direction more deeply.

Libera aims to build digital employees that understand a user's situation and carry out real work. For now, I am testing that possibility through Meta ad operations for beauty brands.

The approach had shifted from selling a system to doing the work directly, but the problem I wanted to solve had not changed. I felt I had gained a clearer understanding of what AI would need to do to deliver the results customers wanted.

For me, a vision mattered at moments like this. Plans could change, but I wanted to know what the work I was doing now would contribute to.

Of course, conviction alone is not enough. I need to see whether customers' problems are being solved, whether they keep paying, and whether a system that can take on the work is actually being built. I felt I needed to trust my instincts enough to act, while gathering evidence I could explain to others.

What I noticed after I was on my own

The strongest feeling after becoming a solo founder again was loneliness.

It had been much easier to stay enthusiastic when we worked together. On my own, my productivity varied widely. Some days I was deeply absorbed in the work; on others, I could barely get anything done.

My strength is taking action. I tend to try things quickly once I have an idea. Planning and attending carefully to details are weaker areas. I had talked about the importance of vision and planning, but believing they mattered was different from being good at them.

While we were parting ways, I thought about our differences. Once I was on my own, I could see more clearly what I needed to improve in myself.

In the first essay, I wrote that I needed a structure that would keep me acting and learning. Only later did I realize that the person I worked with had been part of that structure. Choosing a direction did not, by itself, make me move toward it every day.

An AI team we use ourselves first

Having people to work with still matters. But I also felt I needed to build an internal AI team quickly to help with my work now.

I needed help turning my thoughts into actionable plans, reviewing what I had missed, and following up on work already underway. Good answers to my questions were not enough. It needed to know what I was working on and help until the work was finished.

Libera is a company trying to build those digital employees. That meant we had to make them good enough to entrust with our own work every day.

AI cannot fill every part of the gap left by a cofounder. I still need the strength that comes from having someone to think through our direction with and share the enthusiasm. But I could not leave planning, review, and follow-through entirely to my own motivation and memory.

I thought client agency work could be a starting point: keeping a connected record of what a customer requested, why a decision was made, and what changed after we acted. I want to start by asking AI to use that record to plan the next task and check for anything missed.

Something that works well internally does not automatically become a product that works for customers. Still, if I entrust it with work every day, I can see how much it can do and where I have to step back in.

The goal of a $100M exit remains the same. So does the direction: building a product by solving customers' problems directly.

But now, when I think about the AI employees customers need, I also think about the work I want to hand over myself. Next, I plan to choose one recurring task from our agency work and give it to our internal AI. The first thing I want to find out is whether the work gets finished with less follow-up from me.