Insights
Notes on enterprise AI adoption: methods, case studies, and engineering practice, from workflow diagnosis to systems that actually run.
When agents absorb repetitive work, action-based KPIs lose meaning. Why firms must redefine individual value, leverage, and system contribution.
Meituan gave 100,000 staff 80,000 AI agents, then went back to workflows. Why the unit of enterprise AI adoption is the process, not the tool.
Forward-deployed engineering is not about sitting with clients. Without an ontology and a platform feedback loop, it is just IT services.
Task-level AI gains rarely move the whole process. Why companies should reshape workflows the way Ford reengineered accounts payable.
Most AI projects pick tools before defining outcomes. A goal-first path across cost, efficiency, and growth to decide what AI should change.
Result metrics can't explain how AI changed a process. Why process data from calls, meetings, and agents becomes the basis for enterprise AI.
Strong tools inside an unchanged process rarely produce business results. Here are three common reasons AI projects stall, and how we start from workflow diagnosis.