Despite falling AI costs, corporate spending on AI continues to rise, creating a leadership challenge in managing economic returns. Organizations that treat AI investments as capital allocation decisions tied to measurable business outcomes generate the most value. Simply controlling usage or budgets is insufficient; the focus should be on economic returns.
IBM Consulting's Global Managing Partner Neil Dhar emphasizes that AI investments should be approached like any other capital allocation, with clear expectations of revenue, profit, or cash flow benefits. Many companies, however, treat AI as a technology deployment rather than a strategic investment, leading to uncontrolled cost growth before value is established.
Effective AI management requires aligning investments with specific business processes and expecting tangible results within three to six months. While capping usage may control costs, it can also limit innovation. A better approach is to focus on economics, such as token yield, which measures the value of successful outcomes relative to costs.
Control over the AI operating layer is crucial for optimizing cost, performance, governance, and value. This control allows organizations to choose models, set usage rules, and track costs, rather than relying on external providers. IBM demonstrated this capability by running 3,000 AI agents simultaneously on its own infrastructure.