What's next for companies that build products for others

What does a company that builds products for others become when execution is no longer the expensive part of the work?

Whether execution ever reaches zero does not really matter. What matters is that the cost of turning an idea into working software is falling faster than any other part of building a product, and that gap keeps widening.

What I see with founders is not new, it is just much stronger now. They want everything immediately. There is a second thing happening, where founders forget what a product is beyond the software, but that existed before too and is just a bigger problem now. AI raised what people expect faster than it raised what anyone can actually deliver.

The way we used to work does not match that speed anymore. Spending days planning, writing epics, breaking them into subtasks and assigning them one by one to developers is the first gear. Giving a whole feature to a developer who works with coding agents is the second gear. What founders expect now is a third gear that neither of those reaches.

I saw that clearly once I had ten to fifteen agent sessions running at the same time. Writing the code stopped being the problem, and holding all of it in one head became the problem: what each session was doing, what had already been reviewed, what was waiting on something else, what had quietly gone wrong an hour earlier. The worst of it came at merge time, when all those branches had to come back together and it turned out that two sessions had been changing the same files without knowing about each other.

So the whole workflow was broken, and I started thinking that something had to follow what the agents were doing, review what came back, send work out again with feedback, and surface the decisions that still needed a human. That is what we have been building for the last few months, and internally we call it rcli. How it works is a longer story than this post.

The shape of what the market actually needs is getting clearer to me. The work used to run from the partner to requirements, then planning, then developers, then software. Understanding what someone was actually building was always the first step, it just took a small part of the time. Now most of the work sits there, in going back and forth with the partner until what we are building is clear, and then in designing the system that the execution runs inside.

So I think companies that do not work out how much the system around the work matters will struggle to find a position in this market. I think that is where companies like ours either find their place again, or slowly become replaceable.