Artificial Intelligence is bringing unprecedented changes, particularly in enterprise IT. According to market researcher IDC, businesses typically conservative in their technology spending are projected to allocate $4.25 trillion to tech in 2026, largely driven by AI advancements.
Recent findings from venture capital firm Madrona indicate that 74% of 150 surveyed enterprise IT professionals intend to increase their AI budgets over the next year, with others maintaining their current spending levels. However, less than half of the AI initiatives launched by these companies transition into full production.
While this marks an improvement compared to last year’s MIT report which noted a staggering 95% failure rate of enterprise AI projects regarding ROI, the statistic remains troubling. A success rate of under 50% is certainly a low threshold, but it’s a step up from just 5%.
A significant finding from Madrona’s research is the lack of long-term commitment to AI technologies, even after implementation.
Approximately 77% of enterprises review their AI vendors every six months or on a continuous basis. This trend creates a “fast in, fast out” mentality that starkly contrasts with traditional enterprise SaaS models, where long-term contracts create a certain inertia. As noted in the report, “In enterprise AI, switching costs are lower, and re-evaluation is frequent.”
This situation has important implications for the rapidly growing annual recurring revenue (ARR) figures reported by startups. Initial trial budgets from enterprises catalyzed the AI boom witnessed in 2025. This year was anticipated to be when these large organizations would commit more firmly to AI startups. In fact, the ability of many AI startups to claim swift revenue growth—such as rapidly reaching $10 million—depends on enterprise contracts.
However, for the first time, revenue from enterprises is unstable, even after a startup’s AI product moves beyond the pilot phase and secures adoption.
Part of the problem lies in how many AI startups approach pricing their services for enterprises. Research by Andreessen Horowitz, which surveyed 50 technical AI buyers, revealed that a majority prefer AI fees to be linked to the outcomes achieved rather than conventional usage metrics such as consumption of tokens.
Utilizing a pricing model based on token usage resembles a traditional SaaS approach. Once companies recognize their need for a specific software solution—such as email, HR tools, or cloud services—pricing typically aligns with employee counts or data volumes.
In the AI sphere, pricing based on tangible outcomes—in terms of processed reports, resolved tickets, or generated leads—enhances the perceived value of the product to both the provider and the customer, as noted by a16z partners Tugce Erten and Sarah Wang.
In summary, AI may be steering a new phase of enterprise experimentation. This shift opens avenues for startups, as companies are more inclined to test new technologies; however, it also indicates that enterprise contracts may no longer guarantee long-term financial stability. It remains unclear when or if companies will revert to their traditional purchasing habits.
This article may contain affiliate links, which won’t affect our editorial integrity.
