Why enterprise AI agents disappoint (and why the fix is not “better agents”)

The author argues that many enterprises are failing at AI implementation because they are attempting to jump directly from basic chat interfaces ("crawling") to autonomous agents ("running") while skipping the essential "walking" phase of building context, skills, and judgment. Instead of racing toward full autonomy, companies should focus on developing a middle layer of structured knowledge and workflows to ensure that AI tools—whether simple or agentic—are used efficiently and effectively for each specific task.