Why Are AI Costs Rising Faster Than AI Savings?
AI costs are rising faster than AI savings because the investment required to operate, govern, and scale AI programs grows with every deployment while the savings those deployments produce remain below target, and in most organizations the gap between the two is widening rather than closing. The cost side of the equation is compounding, the return side is not.
- The cost base expands with every deployment: Each new AI deployment adds licensing fees, integration costs, data maintenance requirements, and governance overhead. These costs are additive and largely fixed once the deployment is live. The savings each deployment was meant to produce are variable, frequently delayed, and in approximately forty percent of cases land below the target range the business case projected.
- Savings are being measured against optimistic baselines: Most AI business cases are built on cost reduction targets in the eleven to twenty percent range. The actual savings that materialize are closer to single digits, and in many cases the gross saving is partially or fully offset by the operational costs of running the AI program itself, leaving a net position that is materially worse than the approved business case.
- Budget increases are compounding the problem rather than solving it: Approximately ninety percent of companies whose programs have underdelivered are increasing their AI budgets again, which means the cost base is growing while the return gap remains unresolved. The additional investment is not being directed at the structural causes of underperformance. It is being directed at more deployment, which reproduces the same conditions at larger scale.
- The governance cost is the fastest-growing line: As AI programs scale, the compliance, audit, oversight, and risk management functions required to operate them responsibly are growing faster than the operational savings those programs produce. This is a cost structure problem that does not appear in the original business case and is not resolved by deploying more AI.
Organizations that are managing this gap have separated the cost accounting of AI operations from the cost accounting of AI outcomes, which allows them to see where the program is consuming more than it is returning and make deployment decisions on the basis of net financial position rather than gross activity.
Source: City Shift Finance
