Are we using AI to do things faster — or to rethink how we make decisions in software development?

Are we using AI to do things faster — or to rethink how we make decisions in software development?

Something has been on my mind for some time: how we are using AI in software development. Most of the initiatives I have seen focus on using AI inside the standard development process. Because of that, I keep asking myself: did AI come to speed up the processes we already know, or to break our paradigms about software development?

AI started being used to speed up existing processes, but its natural effect is to break these processes.

It is not easy to answer this question with full certainty, because when we talk about breaking patterns, we deal with predictions and uncertainty. Even so, I believe it is very unlikely that the development flow will remain the same as we know today.

There are many initiatives using AI for product discovery and design, as well as for refinement, architecture, development, code review, and finally testing and delivery. AI has been placed into the existing pipeline. The risk is that we are speeding up bad decisions instead of improving decisions. What is changing — and we still do not fully see it — is that before we followed a traditional model: discover requirements, refine, develop, test, and fix. Now, with AI, the trend is different: simulate scenarios and risks before building, create multiple solution options, and evaluate the impact before implementation, only then turning it into code. Building is no longer the center; deciding is becoming the center.

We are still very attached to the process culture we know — based on sprints, tasks, and delivery — and not on risk, impact, and system behavior. In my opinion, the direction is moving us toward using AI before coding, treating quality as an input instead of an output, and adopting executable models, where contracts, scenarios, and flows are treated as the main source, not as derived documentation.

I believe the biggest challenge we will face is no longer how to use AI, but how to prepare for these structural changes. They are happening quietly, while most people are still focused only on speeding up the SDLC as we know it.