I read this Microsoft post on AI-native software development, and couldn’t help but smile at the irony that it takes something like “Waterfall” for a technology company to become a Frontier Firm.
For two decades, Agile taught us BDUF, Big Design Up Front, almost as an anti-pattern. Don’t try to specify everything upfront. Build something, learn, iterate. And rightly so. When software was expensive and slow to build, it made far more sense to learn as you built than pretend you knew everything upfront.
And to be fair, it worked. And it worked well. Heck, I built a successful company around building products for startups this way.
But AI changes the economics of software development. When writing code becomes cheaper and faster, perhaps the bottleneck moves from building the thing to being precise about what we want built.
The Microsoft approach of doing AI-native software development talks about defining the business problem, creating the technical implementation plan, breaking it into tasks and validating requirements before implementation. If you read that sequence again and you may understand why I smiled.
Which makes me wonder if BDUF is making a comeback, with one small change:
Better Design Up Front.
Not bigger specifications. Not 200-page requirements documents. Not needing “42 signatures” before anything can move to the “development phase.” Just clarity of intent, architecture, constraints and expected outcomes before we ask AI to build.
Perhaps BDUF is back. Just with less Big, and a lot more Better.
Lead Better
I write about the things that make me stop, think, and occasionally change my mind. If that sounds like your kind of thing, come along.


