Organizations are messy. The Bitter Lesson says AI can bypass that mess and focus only on outputs. The Garbage Can Model says process matters. Which perspective wins will define the future of work.
Most companies are chaos. Unwritten rules. Informal networks. Undocumented workflows. Bespoke knowledge everywhere. Can AI work with that, or do you need to clean it up first?
This question sits at the heart of a clash between two perspectives: the Bitter Lesson and the Garbage Can Model.
The Garbage Can Model says organizations are messy by nature. If this is right, traditional automation needs clear processes. You'd spend years mapping workflows before AI could help. Enterprise AI adoption stays slow and expensive.
The Bitter Lesson says encoding human expertise into AI doesn't work as well as just letting AI learn through brute computational force. If this applies to organizations, you can skip the process mapping. Just define good outputs. AI finds its own path through the chaos, even if that path is opaque.
In other words, one says: process matters, understand the complexity first while the other says: outputs matter, let AI figure out the rest.
For now, I'm leaning more to The Bitter Lesson side. The models just want to learn, and humans are the bottleneck.