Why enterprise AI fails at adoption, not technology
Models are rarely the bottleneck. Workflows, incentives and trust are. Adoption has to be designed as deliberately as the system itself.
On AI adoption, building products, and leading meaningful change.
Bridging the GenAI Chasm: From Sandbox to 90% User Adoption
Change Management Is the Real AI Infrastructure
From Buyer Conversations to Signals: Building an LLM Product in Production
Perspectives
Models are rarely the bottleneck. Workflows, incentives and trust are. Adoption has to be designed as deliberately as the system itself.
If you can't say what changes in the P&L, the pilot will become a demo. Quantify first, then build.
Having shipped an LLM product end-to-end, I know where AI breaks — and how to set expectations that survive go-live.