Enterprise spending on AI continues to rise, but new research suggests companies are becoming less committed to individual vendors even after products move beyond pilot programs. IDC expects global technology spending to reach $4.25 trillion in 2026, while a Madrona survey found that 74% of enterprise decision-makers plan to increase their AI budgets over the next 12 months.
At the same time, 83% of the 150 enterprises surveyed by Madrona said fewer than half of their AI pilots became scaled deployments during the past year. Integration complexity, security and compliance, and difficulty proving financial value ranked ahead of product performance as reasons pilots failed.
AI Vendors Face Frequent Reevaluation
Even successful deployments may not translate into durable customer relationships. Madrona found that 77% of enterprises reevaluate their AI vendors at least every six months or on a rolling basis, a much faster cadence than the multi-year relationships traditionally associated with enterprise software.
The pattern creates what Madrona describes as a “fast in, fast out” market. AI startups may be able to enter large companies more quickly, but they must repeatedly demonstrate value to keep those accounts.
That dynamic also complicates how investors and startups interpret rapidly growing annual recurring revenue. Enterprise contracts can produce substantial revenue quickly, but frequent vendor reviews mean that revenue may be less predictable than similar contracts in conventional SaaS.
Buyers Want Pricing Tied More Closely to Work
Pricing is another area where enterprise AI is still changing. Research from Andreessen Horowitz found that 27 of 50 technical AI buyers preferred credits tied to recognizable units of work, while only 14 preferred token-based pricing.
The firm argues that application companies should price around the value they deliver rather than simply passing through the cost structure of underlying AI models. Examples include charging for completed reports, resolved support conversations, qualified leads, processed claims, or other measurable work.
Madrona’s own survey showed a similar mismatch. Thirty-three percent of enterprises preferred outcome-based pricing, while usage-based pricing was the model they encountered most often, at 45%.
The research suggests enterprise AI purchasing is becoming more experimental even as overall budgets expand. Companies are increasingly willing to test new vendors, but production deployment no longer guarantees the long-term customer relationships that have traditionally supported enterprise software businesses.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.