Why Programmers and Process are More Important Than Ever in the AI Era: Part 2

So, can we have rapid development with software development processes to manage the code development in place? That was the question I posed in the previous post. In the words of every Top Gun instructor I ever flew with back in my aviator days, “It depends.” While this may seem non-committal, it’s true.

I have a relative, former video game lead for a large science fiction IP’s MMO, and eventual CEO of the game studio, who put it this way: Software development is walking a tight rope. You have a safety net that is your process. Without it you could possibly die. However, there is also a ceiling above the tightrope that limits your top performers when they run against it. The more you raise your net to decrease the fall (increase in process) the lower the ceiling becomes, limiting the ability of your top performers to excel. So, it’s a balancing act. Enough process to prevent catastrophe and extreme inefficiency, but not so much it holds back innovation. A lot of this depends on your team, and thus, “it depends.”

However, getting back to the crux of this discussion, we also face the rapid increase in AI usage in development. It’s one thing dealing with developers by trade who have been building large software projects, but now we are faced with those not familiar with the process given the ability to generate a lot of code, and fast. All who at the same time are unfamiliar with the software development process. They may throw out terms such as agile, scrum, etc, but really by agile they mean just vibe code as fast as they can.

Unfortunately, with a team all working on a single codebase this can get ugly quick. That minimum amount of process, where seasoned developers managed their merge requests, did regular pulls, and followed good coding standards (even when using AI) is no longer enough. More process is necessary to keep the team from falling off the tight rope into the abyss, which will inevitably cause complaints and frustration from top performers.

So, I can hear you asking, what do I do? And to that I say, “it depends.” In some cases, leadership, non-technical business types, will stay out of it and you can manage it gradually, increasing the process, while monitoring the ceiling. Other times, the lack of awareness of what AI can and cannot do from the business leaders ends up in a “must go fast or we will be left behind” mentality. Here I see two options: 1. Minimal to no process and allow it to eventually devolve into chaos, or 2. Put process in place and then when the leaders come and say why all the process, just ask “Do you like the results?” If they must answer yes, even if the non-developers leveraging AI are complaining about the constraints, then your method will be hard to argue against. If they say no, either because they have unrealistic expectations of speed, or maybe you overburdened the team with process, then be willing to make a change. Be honest and ask “What would you like to see moving forward?” Then rinse and repeat that process based on feedback.

Now, for those of you stuck in a place where its AI speed or nothing, damn the process, I feel for you. This seems to be becoming increasingly prevalent, and while it may be sustainable for a time, eventually no one will understand all the bad code, and in mission critical systems that is a truly terrifying proposition.

So, long story short, I don’t have a great answer. You must manage your team; you must constantly assess the ceiling and net height while you walk the tightrope.

At a minimum, pick up a scrum book, maybe take a three-day online course. Understand the product manager, scrum master, and developer roles. Understand the basic processes used in Jira, Github, or Gitlab for project management. For a team, build out tasks in your planning board. Ensure you have a robust test pipeline, and no merge requests are approved without passing those tests. I would say at a minimum knowing how to use a collaborative platform’s software development tools and using git smartly are low hanging fruit to get you started.

There will be pushback, but if you can bring success, people will get on board. If they are so set in their ways they cannot handle the minimum amount of process required to make this work, then they will move on. Either way, you must find that balance. Remember there is no one size fits all, so “it depends…..”