This episode’s guest was Mika Majakorpi, CTO of Nitor, a digital engineering company based in Helsinki. I wanted his view from inside a business whose entire model is writing custom software for clients, at exactly the moment AI is changing what writing software means.
Does the custom software model even hold?
We started with the uncomfortable question first: if AI can generate code, does custom software development still make sense as a business? Mika’s answer was more nuanced than “yes” or “no”: the value shifts rather than disappears, moving from typing code to the judgment of what to build and how it fits the client’s operating model.
Efficiency is not the same as leverage
He draws a distinction between AI-assisted coding as an efficiency play versus a leverage play. Efficiency saves you time on the task you were already doing. Leverage changes what your team is capable of taking on in the first place, including things like conversational commerce and AI-driven product discovery, which came up as live examples.
The bottleneck is data plumbing, not the model
The most practical part of the conversation, for me, was Mika’s tip on turning AI experimentation into actual business results: most stalled pilots are not a model problem, they are a data plumbing problem. Get the data flowing cleanly to the use case first, and the AI part gets much easier. We also touched on where this goes next: beyond chat interfaces, into autonomous AI agents that act rather than just answer.
Hear Mika’s three concrete recommendations for business leaders navigating this shift in the full episode.