In early 2026, Uber exhausted its AI budget within four months, prompting monthly spending limits and an internal dashboard as STOPGAP measures. However, the larger issue remained: justifying the ROI on AI spending. Many organizations grapple with the accelerated AI adoption outpacing financial oversight, leading to significant costs becoming embedded in operations before they're fully recognized. As Mary Carmichael wrote in her recent ISACA Journal article, it’s crucial for leaders to understand current expenditures, anticipate new costs in production stages, and decide on worthwhile investments. Establishing an enterprise AI budget unifies expenses across projects, helping inform funding decisions.
AI budgets fail due to scattered costs, variable consumption, ongoing commitments post-pilot and activity that doesn’t indicate value. Licenses, tokens, APIs, computation, data, agents, and human oversight costs are often dispersed across cloud accounts and business units. Additionally, autonomous systems may trigger unexpected expenditures. Enterprise AI budgeting is thus a continuous governance process rather than a yearly forecast.
Carmichael outlined five key steps for developing a sustainable AI budget: (confirm!)
- Align AI spending with strategy: Investments should address specific business problems. Metrics for success and stakes in case of failure must be identified. AI initiatives should compete with other priorities for resources.
- Determine the current run rate: Before new investments, assess current AI expenditures. Identify visible tools (like Microsoft 365 Copilot), embedded AI costs, and corresponding business goals, ownership, and cost drivers.
- Forecast operations and projects: Use existing spending to predict how future projects will alter costs. Factor in both implementation and operating expenses, differentiating between one-time and recurring costs. Consider varied future scenarios supported by clear assumptions.
- Develop and monitor the budget: Categorize costs into access/contracts, usage/infrastructure, projects/updates, people/operations, and governance/resilience. Implement spending controls, assign ownership, and regularly review and adjust budgets.
- Evaluate value and adjust funding: Assess the effectiveness of services against their cost. Examine usage quality and connect expenditures to business results. Validate benefits with evidence, and decide whether to continue, expand, or retire services.
Uber's experience underscores the risk of rapid AI adoption surpassing financial transparency. Deploying spending controls without value assessment is inefficient. An enterprise AI budget is more than a forecast. It provides a structured approach to manage and justify AI investments, aligning them with strategic outcomes.
Read more on this topic in Mary Carmichael’s article here.