By Nila Kartika Wati 
Artificial intelligence has fundamentally altered the landscape of enterprise IT, shifting the sector from a period of predictable, long-term infrastructure investment to a high-velocity, high-stakes era of relentless experimentation. While market analysts project that global technology spending will reach an unprecedented $4.25 trillion by 2026—driven almost entirely by AI—the relationship between the world’s largest corporations and the startups fueling this innovation has become increasingly precarious.
The era of "set it and forget it" software contracts is effectively over. In its place, a "fast in, fast out" dynamic has emerged, creating a volatile environment where even the most successful AI startups find their recurring revenue under constant, rigorous scrutiny.
Main Facts: The AI Spending Paradox
The narrative of AI’s ascent has been one of exponential growth, but beneath the surface of the $4.25 trillion spending projection lies a fundamental tension. Venture capital firm Madrona recently surveyed 150 enterprise IT professionals, revealing that while 74% plan to increase their AI budgets over the next 12 months, the actual realization of value remains elusive.
Crucially, these enterprises report that fewer than 50% of their AI pilot projects successfully transition into full-scale production. While this figure represents a significant improvement over the dismal 5% success rate reported by MIT in 2025, it remains a sobering metric. Despite the influx of capital, the "graduation rate" for AI tools remains low, suggesting that companies are willing to experiment but are struggling to find meaningful, sustainable integration for the technology.
Chronology of a Shifting Landscape
To understand the current state of enterprise AI, one must look at the rapid evolution of the market over the past 24 months:
- 2025: The Year of the Pilot. The initial AI boom was characterized by "trial budgets." Enterprises scrambled to allocate capital toward generative AI tools, driven by FOMO (fear of missing out) and the desire to remain competitive. During this period, startups saw astronomical growth, with some hitting $10 million in Annual Recurring Revenue (ARR) in as little as three months.
- Late 2025: The Expectation Gap. As the year closed, market analysts and venture capitalists predicted a period of consolidation. The prevailing theory was that 2026 would be the year enterprises transitioned from experimental pilot programs to long-term, multi-year commitments with a select group of preferred vendors.
- 2026: The Era of Continuous Re-evaluation. The reality has diverged from those predictions. Instead of settling into long-term contracts, enterprises have doubled down on their agility. Rather than cementing relationships, they are treating AI vendors as plug-and-play modules that can be swapped out at a moment’s notice.
Supporting Data: The Volatility of Revenue
The most damning evidence regarding the instability of the current AI economy comes from the Madrona research report. It highlights that 77% of enterprises now reevaluate their AI vendors every six months, or even on a rolling, continuous basis.
This frequency is a departure from traditional Enterprise SaaS, where multi-year contracts acted as a "moat of inertia." In the SaaS era, once a company integrated an email platform or HR software, the cost and friction of switching vendors ensured that the provider could rely on stable, multi-year revenue. In the AI sector, however, the barrier to entry—and the barrier to exit—has been lowered.
As Madrona notes, "In enterprise AI, switching costs are lower and the re-evaluation cadence is relentless." This creates a scenario where a startup’s ARR is no longer a reliable indicator of its long-term financial health. Even after a product graduates from a pilot phase, the contract remains insecure, subject to the quarterly or even monthly whims of the enterprise buyer.
Official Responses and Pricing Disconnects
The instability of enterprise AI contracts is exacerbated by a fundamental disconnect in how these services are priced. For years, the software industry relied on usage-based pricing—charging per seat, per gigabyte, or per API token. However, this model is failing to satisfy the modern enterprise buyer.
Recent research from venture capital firm Andreessen Horowitz (a16z), which surveyed 50 technical AI buyers, indicates that more than half of enterprises are rejecting token-based or usage-based pricing. These buyers are pushing for "outcome-based" pricing, where fees are tied to the actual work produced—such as tickets closed, reports generated, or leads converted.
According to a16z partners Tugce Erten and Sarah Wang, pricing around "recognizable work" is the only way for startups to prove their worth in a skeptical market. When a vendor charges for tokens, the enterprise views it as an operational expense that must be minimized. When a vendor charges for outcomes, the service becomes a direct contributor to the company’s bottom line, making the "buy" decision easier to justify during the next re-evaluation cycle.
Implications for the Tech Ecosystem
1. The Death of the "Moat"
The traditional enterprise sales strategy—securing a three-year contract to guarantee revenue—is currently non-viable for many AI firms. Startups that built their valuations on the assumption of long-term renewals are now facing a "churn cliff." If they cannot demonstrate measurable, high-value outcomes within six months, they risk being replaced by a competitor with a better model or a more effective interface.
2. A New Era of Experimentation
The silver lining for the broader ecosystem is that the barrier to entry for new startups has effectively collapsed. Because enterprises are constantly re-evaluating their stack, they are more open to trying new, niche AI solutions than they were during the peak of the SaaS era. This has created a vibrant, albeit chaotic, marketplace where small, agile teams can win enterprise business, provided they can quickly move from the pilot phase to delivering tangible, outcome-based value.
3. The Investor Pivot
Venture capital firms, which previously prioritized raw ARR growth, are now shifting their focus toward "retention and utility." Investors are increasingly scrutinizing whether a startup’s enterprise customers are merely "testing" the product or if it has become an indispensable part of their workflow. Startups that cannot prove they are "sticky"—that they are essential to the daily operations of their clients—are finding it significantly harder to secure follow-on funding, regardless of how impressive their top-line revenue numbers appear.
4. The Future of Procurement
As we look toward the remainder of 2026 and beyond, it remains to be seen if enterprises will eventually revert to their historical, long-term buying habits. The current climate suggests that "agility" has become the primary organizational goal. Enterprises have learned that locking themselves into long-term AI contracts with unproven or rapidly evolving models is a strategic liability.
Consequently, the future of enterprise software is likely to be characterized by modularity. Companies will likely maintain a "core" set of stable infrastructure while treating their AI layer as a dynamic, constantly evolving suite of tools.
Conclusion: Adapting to the Velocity
The AI-driven transformation of enterprise IT has ushered in a "new normal" where revenue is ephemeral and competitive pressure is constant. Startups that thrive in this environment will be those that move past the "usage-based" pricing models of the SaaS era and embrace outcome-based, value-driven partnerships.
For the enterprise, the benefits of this shift are clear: increased flexibility, reduced reliance on individual vendors, and a sharper focus on ROI. However, for the technology providers, the challenge is existential. In a market where 77% of buyers are ready to switch vendors on a whim, the only way to secure long-term revenue is to move beyond the pilot—and to keep proving, every single day, that the value of the AI product outweighs the cost of the switch.
As the industry continues to mature, the startups that survive will not necessarily be the ones with the most funding, but the ones that have successfully integrated themselves into the essential workflows of the modern enterprise.
Neng Nana 
Suro Senen 
Layla Zulfa 