"AI can absorb that layer" is not a flattening strategy
It's a hope.
Flattening is trendy right now. Fewer layers, faster decisions, leaner teams, it's an easy pitch to a leadership team looking to cut costs and call it strategy instead. And increasingly, the justification has a new shape: AI will pick up what the layer used to do.
Here's the problem. AI absorbing work isn't automatic. It takes planning. It takes investment. And it takes time, the exact three things most flattening decisions skip in the rush to announce them. You can't remove a layer of judgment, coordination, and quality control on the assumption that a tool will fill the gap before the tool has actually been built into how the team works.
Flattening isn't the problem. Flattening without a plan is. Before you touch the org chart, consider this:
- Audit your team's skillsets honestly. Are they built for one lane, or can they operate across several? Most teams, right before a flattening, are still built for one
- Map decision rights explicitly. Someone has to own what the removed layer used to own, on paper, not by default
- If AI is part of the plan, treat it like a real workstream. What's it actually doing today, not eventually. Who's training it, refining it, and owning its output
- If you can afford it, build in redundancy for the first two to three quarters. Don't remove the safety net and the backup plan at the same time
- If you can't afford it, that has to be planned for too. Look 18 months out and set realistic expectations for what your team can actually deliver with the headcount you have
- Set a review checkpoint at 90 and 180 days to catch what's slipping before it becomes normal
Flattening should be earned through a plan, not declared because it's trending or because a tool promised to cover the gap.
If you're already past the point of prevention, I've got thoughts on what to do next. More on that soon.
Sources
More on the AI angle driving a lot of these decisions: