Not all companies are ready to adopt AI effectively. This post explores four types of organisations based on their technical maturity, from those overwhelmed by tech debt to agile startups born in the cloud. It argues that without solid engineering foundations, AI alone won't deliver value and may even backfire.

Tech Debt, AI Adoption, and the Maturity Gap

Over the past few weeks, I had a lot the opportunity to have various chats with different tech companies. A pattern has started to emerge. It feels like, in my opionion, there are four distinct types of organisations when it comes to how ready they are to adopt AI, particularly generative AI.

Here is how I'd group them.

🧱 The Unprepared

These companies haven't started addressing systematically their technical debt. Basic engineering hygiene is missing. They are still dealing with fragile legacy systems, manual testing, and scattered deployment practices. There is very little headspace to embrace AI. These teams are still fighting fires and plugging gaps. AI can help, especially to reduce tech debt, lack of tests, and automation, but it's hard to create momentum to introduce in the product space.

🚧 The Work-in-Progress

Some companies have started the journey. They might be 30% or 70% through. They are adopting some good practices, embracing automated testing and starting to refactor, re-design older systems. But some of them might be bogged down by rigid governance, engineering dogmas or obsolete processes. There's often a reluctance to try new tools, including AI, either out of caution or ideology. Ironically, they might have the right people and the right direction, but still struggle to move fast enough because they are too focused on doing things “properly”.

⚙️ The Mature and Measured

These are companies that have invested in strong engineering culture for years. Good testing practices, clean architecture, high deployment frequency. They are in a position to use AI effectively because their teams are already operating with discipline. They experiment without chaos. They are fast movers, and when they do adopt AI, they do it with purpose.

🏎️ The Born-Modern

These are the startups or scaleups that don't carry baggage. They use modern tooling, have no legacy systems to worry about, and are built to move fast. AI is just another part of the toolkit for them. They can integrate AI into engineering, design, product, customer support, or whatever needs it. These companies tend to outpace everyone else. What slows them down eventually isn't tech, it's scale.

There is a caveat. If these companies don't adopt some of the discipline seen in the “mature and measured” group, such as good testing, sound architecture, and thoughtful decision making, they risk burning out quickly. Speed without structure doesn't last.

🤖 Why AI Alone Won't Fix Much

AI isn't a fix for weak foundations. It doesn't replace good teams, solid delivery processes, or thoughtful leadership. I have heard all kinds of inflated promises lately: AI will cut engineering costs by 30 percent over three years AI will let us ship five times as many features These expectations usually come from non-technical executives who don't understand the actual work involved in building software, the complexity of exhisting architectures and integrations.

There is also a cultural issue. Companies tend to attract people who reflect their mindset. If you treat AI with suspicion, you will end up with teams that avoid it. If you treat it like hype, you will get shallow adoption. But if your team sees AI as a tool to improve how they work, not replace them, then you are in a good place.

🌄 Still Early Days

We are in a pivotal moment. Very few companies actually know how to use AI well in a real engineering context. The noise on social media doesn't help. Lots of hot takes, very little signal. Even some big players are getting it wrong. It is still early days. We are all figuring it out.

What is clear is that the companies who already have a strong DevOps/Agile culture and relative practises in place are the ones who will benefit first and most. The rest will need a double effort before seeing a meaningful ROI from integrating their systems with AI.

Thanks to Giordano Scalzo, Luca Grulla, Uberto Barbini, and Luca Minudel for the conversation that sparked this post. The opinions here are mine, but the thinking came from many heads.