You Probably Aren''t Ready for AI

Continuous delivery has been the most effective way I know to shine a bright light on all of the hidden problems in an organization… until now. The light AI shines on the problems makes CD look like a glowing ember.

The industry is hitting the “FO” portion of “FAFO.” For a couple of decades, management has focused on commoditizing software development by slicing what was once a generalist role into ever-thinner specialties. Some of us have been yelling about that for a long time. The DevOps movement was meant to fix that, but too few were willing to adjust their HR and incentive models, so they created another slice called “DevOps” and declared themselves “transformed.”

Now the fun begins. Guess what skillset you need to leverage AI effectively? You need generalists who understand product, development, testing, and operations. Those who have been lucky enough to learn that instead of trying to be experts at their tiny slice are the people we need for the teams we need to build. We’ll need to help with the remedial training of the rest. Many won’t be able to hack it. They will have jobs for a while, but not the good jobs. That’s over.

Teams are going to change, too. This isn’t futurist thinking. This is what we are going to be doing in 2026. Smaller, more nimble teams delivering more with higher quality and shorter lead times. Product management will either level up, or “PM Productivity” will be the new buzzword.

Processes will need to change. When the Manifesto was written, getting feedback every 2-4 weeks was amazing. Then came CD, and we wanted that feedback daily. We talk about CI batch sizes of 1 or more commits to trunk per day for a developer. Now that will generate huge batch sizes. Time boxes no longer work. We need to figure out another constraint to mitigate risk.

The sacred cow of code review will be slaughtered. Code quality can be 100% automated. I don’t care if we have the right level of abstraction to keep things flexible anymore. It takes minutes to change it. We will also need to find other ways to satisfy compliance on “two sets of eyes on every change” if we want to keep our potential delivery blast radius under control.

It’s hilarious to see people posting about how AI is fake news, demanding to see evidence, and calling us liars. In the meantime, we’re in meetings planning for the org changes and the training needed to adjust to a world where all our normal constraints have evaporated and things we would never have done six months ago are now obvious solutions to support the business today.

Look closely at the things you believe to be true. They may not be true anymore.