There’s been plenty of debate about whether AI will replace software developers. I’m more interested in a different question: what is AI doing to the way developers work with each other?
I’ve been writing software for 17 years, and AI is one of the biggest changes I’ve seen. I use it every day and wouldn’t go back. But I’ve noticed a shift in how my recent teams work, and I wonder how much of it is down to AI.
Are we becoming more productive as individuals at the expense of collaboration?
Fewer whiteboards, thinner reviews
Working through a hard problem used to mean getting a few developers around a whiteboard. Someone suggested an approach, someone disagreed, someone spotted what everyone had missed. You might spend an hour on something that took 20 minutes to build, but everyone understood how you got there.
Now it’s tempting to open Claude, describe the problem and solve it alone. It’s quick and often excellent. But nobody else was part of the process.
Code reviews are going the same way. They used to be one of the best ways to understand how colleagues thought. Now AI can review the code, find the bugs and even make the fixes. The risk is AI writing the code, AI reviewing it and developers approving changes they haven’t fully understood.
A review isn’t just about catching bugs. It’s about knowing what your colleagues are building and why.
The questions we no longer ask
“Have you worked with this library before?” “Why was this service built this way?” “Can you have a quick look at something?”
Those questions often went well beyond the original problem. You learned why decisions were made and picked up techniques from people with more experience. Now we ask AI instead.
Not every question needs a colleague. But some of those interruptions were how knowledge spread around a team. That matters most for junior developers. AI is an incredible learning tool, but learning from a senior developer means watching how they think, not just getting the right answer. And some knowledge isn’t in the docs or the code. It lives with the people who built the system.
Are we still working as teams?
Imagine five developers on the same application. Each picks up a ticket, works through it with Claude, gets an AI review and moves on. Tickets move. Features ship.
But how much do they know about what each other is building? Could they maintain each other’s code? Or have we created five developers working independently with five AI assistants?
Stand-ups and sprint planning coordinate the work. They don’t build a shared understanding of the system. And they don’t replace a colleague who knows the history, knows the business and might tell you that you shouldn’t be building the thing at all.
Productivity numbers won’t show this. What happens when a senior developer leaves, or something breaks in production and the person who built it isn’t around? You can’t measure shared knowledge and trust by counting merged pull requests.
So, what’s the answer?
Not to stop using AI. It has removed a lot of tedious work, and not every question needs a team discussion. But we should be more deliberate about the conversations we keep.
- Use AI to explore an architecture, then talk it through with your colleagues.
- Let AI review pull requests, but keep reviewing each other’s work.
- Encourage juniors to use AI, but don’t let it replace mentorship.
- When you solve something interesting, share what you learned.
The goal isn’t more code. It’s better software built by better teams.
AI is making us better at working independently. Whether it’s making us better engineering teams is another question entirely.