Why AI Agents Expose Hidden Enterprise Gaps

ServiceNow's Chief Innovation Officer Dave Wright makes a case that agent failures aren't bugs - they're a map of where enterprises assumed connections that never existed. Clinical coordinators reconciling patient data across three unconnected systems, procurement managers emailing for missing approvals, junior analysts re-entering identity data between incompatible platforms. Agents hit these walls and stop. BCG data backs the argument: 70% of AI scaling barriers trace to people-and-process issues, only 10% to algorithms. Google DeepMind and MIT research found that coordination overhead in multi-agent environments can actually exceed the productivity gains. Wright's CAP framework - catalog stalls, assess business impact, prioritize integration investment - is simple, and the companies treating agent failures as diagnostic tools rather than product shortcomings are the ones making actual progress on enterprise automation.