How to manage AI investments in the agentic era

OpenAI published a guide on July 14, 2026, outlining five steps for enterprise leaders to manage AI investments as teams shift toward longer-running agentic workflows. The company said token prices have fallen 97 percent from GPT-4 to GPT-5.4 and that GPT-5.6 delivers better coding performance with 54 percent fewer output tokens and 57 percent less time per task. OpenAI said leaders should measure useful work per dollar rather than token cost alone. The guide recommends sharpening visibility into usage and spend through updated Admin Console analytics that show adoption, credit usage, and trends by user, product, and model. It advises evaluating model efficiency by outcome ROI using evals that reflect real tasks and tracking cost per accepted outcome paired with business value. Governance should be treated as an operating layer that defines approved context, tools, actions, and approval paths before workflows scale, with centralized controls for access and spend limits. Investments should be managed as a portfolio across broad access, function-specific workflows, and strategic bets built on proprietary context, with funding tied to maturity stages. Finally, capacity should match proven demand using commercial structures such as Guaranteed Capacity for production agents, Scale Tier for high-volume API workloads, and Batch API or Prompt Caching for asynchronous work.