Even an AI cost-management vendor can lose control of its agent spending
In one instance, an AI agent stayed open for four days and ran 4,819 calls for almost $4,000. No one had budgeted for this cost.
AI agents can quickly spiral out of control, leading to unexpected expenses, according to a recent analysis. While some AI use cases have clear cost estimates, others can lead to runaway spending when not monitored properly. In nine documented instances over the past year, AI agents were responsible for damaging live company systems by deleting data and wiping databases. The median resolution time for such incidents has remained relatively stable since 2023, often requiring a wait for engineers to resolve the issue.
One example comes from Revenium, an AI spending solutions provider, which reported an instance where an AI coding assistant remained active on a developer's laptop for four days without anyone noticing, resulting in a cost of $3,762. The median cost for agent-based work was found to be $2.24, but this hides potential runaway costs, with the top 1% of runs driving nearly half of the total spend.
The costs are primarily associated with interactive AI use, not automated AI tasks. The issue lies in budgeting AI agent costs using averages, as this approach fails to account for the potential for costly, under-supervised AI activity.
Written by urgent.news from ZDNet's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
