CIO Dive reported that businesses continue to grapple with managing AI costs as usage accelerates and that OpenAI published a Tuesday blog post outlining five steps to control spend. The guidance is particularly relevant for CTOs, CIOs, FinOps leaders, and AI platform owners moving from pilot projects toward scaled deployments.
What changed
CIO Dive said roughly three-fifths of IT professionals, citing Flexera, reported increased AI overspend. The outlet also said more than two-thirds lacked visibility into AI software usage.
Why B2B teams should care
The reported issue is not just model list pricing. Per CIO Dive, OpenAI said cheaper models may fail and retry, which can drive more token usage and undermine apparent savings.
OpenAI’s guidance, as relayed by CIO Dive, was broader than price selection alone:
- Increase visibility into AI use and spend.
- Track model outcomes.
- Incorporate governance.
- Manage AI investments as a broader portfolio.
- Match the product to demand after proving value.
Who is affected
The guidance primarily targets enterprise IT and finance leaders — including CIOs, CTOs, FinOps teams, and AI platform owners — that are scaling AI deployments from pilots toward production.
CIO Dive reported that Shutterstock CTO and CISO Courtney Totten said understanding AI costs is foundational to business strategy.
The issue is especially relevant for teams moving from pilots to scaled deployments. CIO Dive reported OpenAI’s view that the strongest candidates for scale are workflows that repeat at meaningful scale and have clear ownership.
What teams should check now
Based on the reported guidance, enterprise teams should review four areas immediately:
- Audit who is using AI, which products and models they use, what capacity is consumed, and what work the usage supports.
- Track both spend and model outcomes rather than treating usage volume alone as the success metric — for example, cost per resolved support case or per tested code change that passes review.
- Define governance for what context, data, applications, and tools LLMs can access.
- After proving value, match product, capacity, and support model to demand.
As analytical framing for readers, not a quantified claim from any one vendor, the main spend drivers in agentic systems typically include:
- Model tokens
- Tool calls
- Retrieval and vector search
- Orchestration layers
- Repeated agent loops
Beyond the steps OpenAI recommends, teams across the industry also lean on cost controls that sit above any single vendor’s guidance:
- Model tiering and routing
- Prompt and response caching
- Batching
- Rate limits
- Usage budgets
- Observability and FinOps dashboards
For background on OpenAI pricing and admin controls, see our OpenAI Pricing Guide: ChatGPT Plans and API Usage Controls.
What remains unclear
- Not yet confirmed: the specific OpenAI blog post title cited by CIO Dive.
- Not yet confirmed: the exact publication date of the OpenAI post, beyond CIO Dive’s description of it as a Tuesday blog post.
- Not yet confirmed: the Flexera study details and sample size behind the overspend and visibility figures.
- Not yet confirmed: the Protiviti study details and sample size referenced in the coverage.
- Not yet confirmed: any quantified savings or cost impact from the recommended steps.
What to watch next
Enterprise teams should monitor whether vendors begin exposing richer AI usage telemetry and governance controls as agentic deployments move into production.
They should also watch whether enterprises adopt the portfolio-style management of AI investments that OpenAI recommends.
While unrelated to OpenAI’s guidance, Oracle separately announced an AI-native builder experience for Fusion Agentic Applications with logged actions and inherited security and governance controls — a move that reflects the same broader industry emphasis on governance for production AI systems.
