I think it is worth discussing more how do we use AI in a structured way for sustainable software development? In some companies, especially the ones having the pleasure of working in a compliance-heavy domain, SDLC or software development life cycle is something important. They have to be formal and explicit about how they build and maintain their products. And it is in general a potential sign of maturity when it comes to software development. And of course, the SDLC document should preferably reflect reality.
So here’s the question: what would an AI-SDLC look like? And let’s go even further: what would a company where AI is a native component of the organization look like? What would the AI “Spotify Model” look like? I have a gut feeling that an AI-native software development company would look quite different than a non-AI-native one. Right now, many companies try to adopt AI and embed it into their current processes and structures. But what if that’s the wrong approach? What if we have to rethink everything - company structure, culture, processes - from the ground up?
There are hints that we could probably make do with smaller teams. But why stop there? Do we even need teams anymore? What would the communication channels look like? Do we even need the same roles? The same structures?
And then there are the hard problems that an AI-native company would have to solve - for which solutions still need to be streamlined:
- Quality. How do you keep quality under control when models and agents don’t inherently optimize for it?
- Security. How do you prevent agents and models from leaking precious information, or having it extracted from them by malicious actors, or even being used against their “owners”?
- Reliability. If AI is a deeply embedded dependency, it has to be up. Just look at the status pages of the biggest AI companies - are we ready to bet our businesses on that?
P.S. The image is mostly Nano Banana 2 interpretation of the simple prompt I gave it, but I think it is cool.
Originally posted on LinkedIn.