AI agents: High effort, low return

AI agents are generating significant interest, but the business case is still weak: McKinsey finds most companies are experimenting with them, yet only a minority see positive EBIT impact, while agent workflows can consume 5 to 30 times more compute than chatbots and frequently blow through AI budgets. For CIOs, the strategic implication is that agent adoption is not just a software decision but an enterprise architecture and operating-model change that can raise cost, complexity, and risk unless tied to measurable outcomes. IT organizations should expect to rework infrastructure, governance, and development practices before agents deliver meaningful productivity gains at scale.