In my experience, the never-ending source of technical debt in projects is all those teeny-tiny details that you decide not to do right away. Because the feature is already “good enough.” Because you are tight on time. Because the thing does not seem worth polishing right now, so you create a ticket, throw it into the backlog, and promise yourself you will pick it up later.
The thing is, this later often becomes never. There is never a good time for it, never enough developer capacity to fit it into the workload, and it is never the priority. These things are left to rot in the backlog and, more importantly, in the codebase, until they become a monster you are scared to touch.
With the new AI coding tools in place, there is much less reason to compromise and leave stuff for later in your project backlog.
Programming is about problem-solving. If you are a tradcoder like myself, used to hand-typing the core solutions to the hard problems in your project, you already know that once you get everything working, you get that dopamine hit. You have solved the puzzle.
Everything after that can feel like a burden. After an intensive, focused coding session, you often do not feel like doing what you know is the right thing to do: cleaning things up, polishing the code, and making the final pass that turns a working implementation into a solid one.
AI is like a spotter that helps you finish the last rep.
It is always there, and it never gets tired. It can help you carry the work over the line and handle the final refactoring. It is still fun to come up with a clever way to rearrange the code or extract a smart abstraction; going through the project and applying that pattern to 10 other files is not. AI can help with that final push while you still get to do the part that actually feels like engineering.
There is another angle here. We all know the programmer meme about spending eight hours automating a task that takes five minutes to do. After learning that lesson the hard way, many of us stopped caring too much about automating small routine tasks that pop up here and there in our projects.
AI can automate tasks that were not worth automating before.
Need a one-time script to populate extensive test data so your colleagues can test something on staging? AI can take care of that.
Need a new CI/CD job for one small maintenance task you only do once every two weeks? Let AI write that one too.
Right now, AI agents are not good at building robust, complex distributed systems on their own. But they are good at producing polished code under supervision. We should make the most of that. Let them take care of tedious tasks so we can spend more of our time on the parts of programming that are actually interesting.
With the tools we have now, there is less point in leaving these nice-to-haves for later and throwing them into the backlog by default. Go look at your backlog, pick one of those small tasks you have been postponing, and let AI help you do it now. It may not remove every excuse, but it removes a very big one.