How to keep AI costs in check as agentic drives usage
Summary
OpenAI, a major large language model provider, published a blog post on July 17, 2026, detailing five steps for businesses to manage AI costs and understand value generation amidst increasing "agentic" AI usage. This guidance comes as businesses grapple with AI sprawl, with a Flexera study indicating three-fifths of IT professionals report increased AI overspend and over two-thirds lack visibility into AI software usage. OpenAI's recommendations include increasing visibility into AI use and spend, tracking model outcomes against business value, incorporating robust governance, managing AI investments as a portfolio, and matching product capacity to proven demand. The article notes that while token costs are decreasing, managing AI investments remains a challenge, with C-suite perceptions of AI value often misaligned. Companies like Prudential Financial and Shutterstock are already implementing stringent AI cost management strategies.
Key takeaway
For CIOs overseeing AI investments, you must prioritize establishing comprehensive visibility into AI demand, spend, and risk. Implement OpenAI's five-step framework to track model outcomes against business value, not just token costs. This approach helps align AI initiatives with strategic goals, ensuring resources go to workflows with clear, measurable returns. Proactive governance is crucial to scale AI safely and cost-effectively, preventing overspend and maximizing value.
Key insights
Effective AI cost management requires clear visibility into usage, spend, and value generation, especially with agentic AI driving demand.
Principles
- Evaluate AI models by full outcome cost, not just token price.
- Implement governance for LLM access, actions, and approvals.
- Prioritize AI investments for measurable, repeatable workflows.
Method
OpenAI proposes five steps: increase AI use visibility, track model outcomes against business value, incorporate governance, manage AI as a portfolio, and match product to proven demand.
In practice
- Monitor AI usage by user, model, capacity, and work supported.
- Quantify AI value via time saved or improved decision-making.
- Establish approval processes for high-risk LLM actions.
Topics
- AI Cost Management
- Agentic AI
- OpenAI
- IT Governance
- AI Investment Strategy
- LLM Operations
Best for: VP of Engineering/Data, Executive, Director of AI/ML, CTO, IT Professional
Related on AIssential
See Counsel's argued verdicts on the open AI decisions leaders are weighing →
Editorial summary, takeaway, and curation by AIssential. Original article published by Information and Enterprise Technology News | CIO Dive - Www.ciodive.com.