AI agents can rack up huge unbudgeted bills — even for cost-management vendors

The useful part

ZDNET's key takeaways AI agents risk spinning out of control and running up charges. AI agents are proliferating, often beyond IT managers' control, and the costs of unseen agentic activity can add up quickly, a recent analysis shows. StackGen's latest State of Reliability Report found that unregulated or semi-supervised agents are proving to be problematic on many levels.

How it works

• This potential issue outlines how organizations must manage the AI agents they deploy and the risks of their unchecked proliferation.
• The issues arise because AI sessions are budgeted similarly to SaaS sessions, with a flat per-seat or per-token cost, tracked as an average.
• The median cost for agent-based work was $2.24 across 557 code-implementation tasks over 90 days, which looks obviously very sustainable.
• However, tracking averages hid the potential for runaway AI agent costs, as shown with the $3,762 session cost mentioned above.
• "The spread reflects different work and different patterns of AI use," the team members related.

What to take from it

"If you're managing an AI bill against an average, you have no visibility into what could happen tomorrow morning," the engineers cautioned. Look for high-end AI use case costs, not averages. Their efforts were invisible to standard monitoring until the damage appeared.

Example or evidence

• Among 10,005 interactive agentic sessions studied, the bill came to $109,118.
• He has authored numerous research reports in partnership with Forbes Insights and Unisphere Research, a division of Information Today, Inc.
• By the time anyone noticed, the agent had made 4,819 calls at a total cost of $3,762, without anyone's knowledge.
• Still, these are upfront costs, and pricing does not account for runaway costs with under-supervised agents, which, of course, can quash ROI.

Details worth keeping

Even an AI cost-management vendor can lose control of its agent spending. One provider recounted its issues with uncontrolled agents. The study, which analyzed 109,000+ incidents, also found that as failure modes multiply, recovery isn't speeding up: median resolution times have been roughly flat since 2023, and the single most common fix is waiting for another company's engineers.

Related coverage

• Saastr: Almost every pre-AI B2B product we use has now told us they're raising prices for agent access.
• Ombulabs: Originally appeared on OmbuLabs.ai.
• Financial Planning: Token prices have collapsed significantly. Enterprise bills have grown anyway. Here's what advisors should know before scaling up their AI usage too quickly.