Key takeaways

  • Monitor useful work per dollar, not just token prices.
  • Implement clear visibility into AI usage and spend.
  • Evaluate models based on full cost and business value.

What happened

OpenAI has made significant strides in reducing the cost of AI usage, with token prices falling by 97% from GPT-4 to GPT-5.4. The latest iteration, GPT-5.6, further optimizes performance, reducing output tokens by 54% and time per task by 57%. However, the cost savings alone do not guarantee value creation. Leaders must focus on the work accomplished per dollar spent, including tasks completed, time saved, and improved decision-making. As AI transitions from chat-based interactions to longer-running workflows, administrators need enhanced visibility to manage demand, spend, and risk effectively.

Why it matters

With AI becoming more integral to business operations, the ability to track and optimize its usage is crucial. Without clear visibility, rising costs can be difficult to interpret, potentially reflecting waste, productive experimentation, or critical business processes. OpenAI’s tools, such as updated usage analytics and spend controls in the Admin Console, provide a shared view of demand across platforms, helping administrators understand and manage AI usage more effectively. This is particularly important as teams adopt more advanced capabilities, such as plugins and connectors, which can operate across enterprise systems.

What to watch

Enterprise leaders should focus on evaluating models based on their full cost and business value, not just token prices. This involves considering factors like model and tool usage, attempts, completion rates, latency, and human review. For priority workflows, tracking cost per accepted outcome is essential. Additionally, clear instructions, focused tools, and explicit stopping conditions can reduce wasted spend. Governance should be treated as the operating layer that determines which AI work can scale, defining context, tools, actions, and approvals. OpenAI’s AI Deployment Engineers can provide tailored support for high-value deployments, ensuring both performance and cost efficiency.