I’ve been thinking about what AI-assisted development does to the old “just enough, just in time” approach to requirements. Because I’m not sure just enough is enough anymore.
As a BA you obviously did the analysis up front, but you didn’t need to have every single thing nailed down perfectly before development started. You got it to the point where a developer could pick it up, and then you worked through the rest together.
They’d start building, questions would come back, they’d challenge something, you’d realise something didn’t quite work, they’d suggest another way and you course corrected as you went. Some really good analysis happened in those conversations.
What changes when AI is in the loop
Now put AI into that process. The developer is still at the wheel, but what used to take a few days to build might come back incredibly quickly. That window where the requirement could breathe a little and get refined during development has suddenly got a lot smaller.
So does “just enough, just in time” now need to become “just enough, just in time, but you’d better be bloody right”?
Obviously you can change the code quickly. That’s one of the benefits of AI. But if we genuinely have humans in the loop, someone still needs to review that regenerated code. Then review it again when the BA remembers the edge case they probably should have thought about yesterday. I can imagine the joy.
Requirements matter more, not less
Maybe that’s one of the less talked-about impacts of AI-assisted development. It hasn’t made requirements less important because code is easier to produce. It might actually be doing the opposite.
The cheaper and faster code becomes to generate, the more expensive poor thinking becomes in comparison. And maybe “just enough” needs a bit of a rethink.
Need business analysts who get the thinking right up front? Talk to Talexa.
