Why Procurement Leaders are Finding AI Buying a Struggle
AI procurement platform Levelpath has published its 2026 report, Are You Ready to Buy AI?, a procurement benchmark survey.
Of the 300 responses from procurement and supply chain leaders at organisations actively purchasing AI software, the survey found that AI buying is slower and more expensive than most organisations anticipated and many are struggling to keep up.
"The data reflects a procurement system that wasn't necessarily designed for the speed and complexity of AI buying and organisations that don't adapt will lack control," says Stan Garber, Co-Founder and President at Levelpath.
"The organisations that will be most successful in this transition will move faster, negotiate better and build vendor relationships that actually support long-term AI strategy."
Few organisations have AI spending under control
For many of these companies, keeping AI spending in check is a constant battle. In fact 57% have already run into cost-related issues with an AI vendor, while a mere 16% feel highly confident in their ability to manage these budgets.
“The data reflects a procurement system that wasn't necessarily designed for the speed and complexity of AI buying and organisations that don't adapt will lack control. ”
- 35% say AI bills have been higher than budgeted
- 27% say users have had to stop work because they've hit usage caps
- 26% of organisations have pulled budget from elsewhere to cover higher AI costs
- 9% of organisations have terminated an AI vendor due to price increases
Despite these rising challenges, most buyers aren't adjusting their contracts to protect their bottom line. While 39% have added data portability clauses and 36% have shortened contract terms, a mere 16% have successfully pushed for cost caps.
AI deals take longer and involve more stakeholders
The data shows that AI procurement takes significantly longer and requires far more stakeholders than traditional enterprise software.
For standard software purchases more than US$10K, the typical buying cycle is 7 to 10 weeks (reported by 21% of buyers). For AI, that timeline nearly doubles, with the most common cycle stretching to 16 to 20 weeks (reported by 25%).
AI also brings a crowded house to the decision-making table. While 43% of standard software purchases involve seven or more stakeholders, that number climbs to 58% for AI, with 28% requiring 11 or more reviewers.
Every added voice introduces another potential bottleneck:
- 58% of buyers cite security reviews as a top cause of delay
- 57% say vendor evaluations delay or derail purchases
- 52% cite contract negotiations as a cause of delay
When deals stall, vendors are often left in the dark. Only slightly more than half of organisations (54%) proactively update vendors during a delay.
Autonomous procurement is close
The survey also gauged enterprise appetite for autonomous procurement, revealing a surprising willingness to let AI take the wheel on purchasing decisions.
Only 15% of respondents completely rule out AI-driven purchases, with an additional 10% held back strictly by legal or regulatory hurdles. Meanwhile roughly 40% are already open to some degree of AI autonomy today.
- 20% of buyers would allow an AI agent to execute transactions under US$10,000 with appropriate controls in place
- 14% of buyers would allow an AI agent to execute purchases under US$1,000
- 6% of buyers would allow an AI agent to make any purchase within pre-approved budget parameters
The takeaway is clear: autonomous buying cannot happen in the dark. It requires total visibility into every purchase, vendor and contract.
The businesses laying this groundwork now will be the first ones ready to delegate decision-making tomorrow as their comfort levels catch up with the tech.
Company portals
Executives
Stan Garber
Co-Founder & President
