The article reports that venture capital firm Madrona’s research indicates a significant shift in enterprise IT spending driven by AI, with 74% of surveyed IT professionals planning to expand their AI budgets in the next year. However, the research also reveals that fewer than half of AI pilots progress to full production, representing an improvement from a previously reported 5% success rate. Crucially, the study finds that even when AI technology is deployed, enterprises frequently reevaluate their AI vendors every six months, creating a “fast in, fast out” dynamic that contrasts sharply with the long-term commitments typical of traditional enterprise SaaS contracts.
This rapid reevaluation cycle has profound implications for startup annual recurring revenue (ARR), as enterprise trial budgets that initially fueled the AI boom in 2025 are no longer translating into long-term commitments. The article suggests that part of the issue stems from AI startups’ struggle to establish effective pricing models; research from Andreessen Horowitz indicates that over half of technical AI buyers prefer fees tied to outcomes rather than usage metrics like token consumption. This preference for outcome-based pricing is seen as a way for startups to demonstrate economic value to customers, but it also contributes to the insecurity of enterprise revenue for AI startups even after successful pilot phases. The article concludes by noting that while increased enterprise experimentation opens opportunities for startups, it also means that securing long-term revenue through enterprise contracts is becoming increasingly uncertain.
Source: Startup ARR is less secure than ever, new research shows