AI can write code faster than any human alive, so why doesn’t it feel like there’s a bigger impact on velocity?
This is the question most engineering leaders are quietly mulling right now. They bought into the promise. They rolled out the tools. Some teams are genuinely more productive, but it’s spotty. The promised step-change in throughput hasn’t materialised, at least not at the scale expected.
Here’s why: everyone assumed that the mechanical act of writing code was the bottleneck. Turns out it wasn’t.
Think about where development time actually goes. Yes, there’s the implementation: writing functions, wiring components, handling edge cases. AI is genuinely brilliant at this.
But surrounding it is everything else: understanding requirements well enough to build the right thing, making design decisions that won’t haunt you in six months, coordinating work across a team without treading on each other’s toes, ensuring quality before shipping, and maintaining the shared understanding that keeps a codebase coherent over time.
AI accelerates exactly none of that.
When implementation friction disappears, everything else comes into sharp relief. The vague requirements. The inconsistent patterns. The coordination overhead. The knowledge that lives only in people’s heads.
These problems existed before AI. We just had less time to notice them.
The teams genuinely winning with AI aren’t the ones with the best prompts or the most aggressive adoption. They’re the ones who have recognised that the bottleneck has shifted and have adapted accordingly.
Less time on implementation, more time on specification. Less time on writing, more time on thinking. More rigour upfront, less firefighting downstream.
The question isn’t whether AI can write code. It clearly can. The question is whether you have the process around it to turn that capability into consistent, quality output.
Without that, you’re not increasing delivery velocity. You’re just generating faster chaos.
What shifted for your team when you introduced AI, and what stayed stubbornly the same?