Essay
AI strategy is org design in disguise
The bottleneck to AI value is rarely the model. It's who decides, who's accountable and how work flows.
Every few months a new model arrives that is meaningfully better than the last. Yet most companies’ AI results don’t improve at the same rate. Why?
The model isn’t the constraint
When an AI initiative stalls, the post-mortem rarely blames the model. It blames unclear ownership, workflows that were never redesigned, or incentives that punish experimentation.
Three questions to ask instead
- Who owns the outcome? Not the tool, the outcome.
- What decision changes? If no decision changes, nothing changes.
- What gets stopped? Capacity is finite, so new work needs old work to end.
Replace this sample with your own thinking.