According to UBS on June 15, enterprise AI adoption faces rising token and compute costs, but the increase stems primarily from usage growth rather than unit price inflation. The bank surveyed approximately 13 enterprise IT executives and found that about 60% of organizations view AI token and compute expenses as a material issue, especially after transitioning from simple chatbots to autonomous agents.
Most enterprises have deployed or plan to implement guardrails including token pooling, model downgrading, waste alerts, and user restrictions to eliminate inefficiencies without curtailing adoption. Executives emphasized they aim to accelerate employee AI usage while optimizing costs through measures like reducing external IT services and consolidating cloud spending.