57% of enterprise AI buyers have faced a spend-related issue with an AI vendor in the past six months, according to Levelpath's new survey of enterprise software buyers. The most common surprise is an invoice that comes in higher than budgeted (35%), followed by users pausing work after hitting usage caps (27%) and budget pulled from other priorities to cover higher AI costs (26%). A smaller group, 9%, has gone as far as terminating an AI vendor agreement over price increases.
None of these results mean that AI is too risky an investment. Spend is rising because teams are putting AI to work, and a growing bill usually reflects growing adoption. What the data reveals is a visibility gap at the heart of AI cost control. Costs that scale with usage reward the organizations that can see usage clearly, and many procurement teams are still building that muscle.
.png)
The Real Story Is Visibility, Not Volatility
Usage-based pricing is a feature of the AI era. It lets organizations start small, scale what works, and pay in proportion to the value they consume. The same structure also means costs respond to adoption in real time, so budget conversations need to happen continuously rather than once a year at renewal.
Our survey suggests most organizations are mid-transition. When asked how confident they are in their ability to control AI costs, 47% of buyers said they are somewhat confident, while 36% said they are not so confident or not at all confident. Only 16% described themselves as very or extremely confident. That last group is small for now, and it represents the standard everyone else will be chasing: teams that can forecast AI spend, watch consumption as it happens, and walk into every renewal with evidence in hand. Reaching that outcome starts with having procurement data the whole organization can trust.
.png)
Flexibility Is the First Thing Buyers Negotiate
AI contracts are evolving along with the technology they govern. When asked how their organizations have changed their approach to AI software agreements over the past 12 months, the most common answers centered on the ability to change course:
- 39% added explicit exit or transition clauses with data portability, keeping every option open in a market where a better model can arrive within a quarter.
- 36% shortened initial contract terms, trading multi-year lock-in for the freedom to renegotiate as pricing matures.
Both moves let buyers keep pace with the technology instead of guessing where it will be in three years, and strong contract management is what makes those clauses enforceable when the moment comes: renewal dates tracked, obligations surfaced, and terms ready to compare.
Roughly a third of organizations (32%) have not changed their contracting approach at all, which makes the next renewal cycle a ready-made opportunity for them.
Cost Terms Start with Transparency
When buyers do negotiate cost-specific terms, they reach for visibility before hard limits. 32% have pushed vendors for more granular usage transparency, asking for detailed reporting on what their teams consume and what it costs. Half as many, 16%, have introduced per-unit price caps or total spend caps. A further 28% have moved toward outcome or value-based pricing, where fees track the results the tool delivers rather than the volume it processes. Every other cost term depends on knowing what they actually use, so that is the term buyers are securing first.
One emerging clause deserves attention. 12% of buyers have added model efficiency clauses, which require the vendor to pass on savings when its own costs fall. The cost of running AI models has dropped steadily, and these early movers have turned that industry trend into a budget line.
The AI Cost Control Playbook for Procurement Teams
The playbook emerging from the data is practical, and it starts well before the invoice arrives:
- Make usage visible before it becomes a bill. Granular usage reporting, agreed in the contract, turns AI spend from a quarterly surprise into a weekly data point.
- Negotiate flexibility and cost terms together. Exit clauses and shorter terms protect your options, while usage transparency, caps, and efficiency clauses protect your budget. The strongest position holds both.
- Treat usage caps as a signal, not a stop sign. A team approaching its cap is a team getting value from the tool. Monitoring and alerts lets admins step in early, adjust the plan, and keep the work moving.
- Revisit your earliest AI contracts. Anything signed before your organization understood AI pricing deserves a fresh look at renewal, with the full menu of modern terms on the table.
We believe visibility is the foundation of AI cost control, and it is what we built Levelpath to deliver: every purchase, contract, and renewal in one place, so procurement teams can forecast with confidence, negotiate from evidence, and manage AI spend with confidence. The same visibility lays the groundwork for what comes next, as AI agents take on more of the purchasing process itself.
About the data: Findings come from “Are You Ready to Buy AI?”, a Levelpath benchmark survey of enterprise software buyers fielded in June 2026. Respondents work primarily in procurement and supply chain roles, mostly at organizations with more than 1,000 employees, and the AI findings reflect buyers involved in AI software purchases above $10,000 in the past year.
