How to Manage AI Spend in the Agentic Era

OpenAI's guidance comes as AI moves from simple chat interactions to longer-running, multi-step workflows. The 97% price drop per million tokens from GPT-4 to GPT-5.4 shows a broader trend of falling AI costs, but OpenAI argues that low token prices don't guarantee value. Instead, enterprises should evaluate models by outcome ROI, considering the full cost of reaching a quality result, including attempts, completion rates, and human review.

The article emphasizes governance as an operating layer, with the need to define what context ChatGPT can use, which tools it can access, and who approves higher-risk actions. This becomes crucial as teams adopt plugins, connectors, and Computer Use. OpenAI positions ChatGPT Work as a central control hub, while offering Frontier and Deployment Company for larger strategic deployments.

The stakes hinge on whether enterprises can avoid scaling AI workflows that look promising but don't deliver measurable value. The advice implies a portfolio approach, funding exploration, validation, and production separately. The real test is whether companies can match capacity to proven demand, scaling only workflows that show clear, repeatable value.