For most of SaaS history, the dominant value proposition was straightforward. Software helped people do work. The buyer paid per seat, per month, because value scaled with the number of humans using the tool. In 2025 and 2026, that assumption is breaking. AI is pushing enterprise software from “helping users” to “performing tasks.” That is not a feature upgrade. It is a structural change in how software is packaged, priced, adopted, defended, and valued.

If that happens, the question is not “is SaaS dead?”. The real question is: what survives when the UI stops being the product, when value moves to data, orchestration, and outcomes, and when pricing shifts from seats to consumption or per-agent economics. McKinsey’s analysis of AI-era monetization makes the same point from the business-model angle: as software starts performing work (not just supporting it), vendors are pulled toward usage-aligned pricing and new meters that map more directly to value delivered.

Satya Nadella has been blunt about where this is going. In recent interviews, he argues that many “classic” business apps are basically CRUD systems wrapped in UI, and that as agents become the primary interface, a big part of the “business logic” will migrate out of individual SaaS products into an agentic layer that orchestrates work across tools and data sources.

McKinsey’s framing is blunt: AI is transforming software from a tool that enables work into a platform that actively performs and orchestrates work. That one sentence implies a downstream chain reaction that touches pricing metrics, procurement behavior, product strategy, GTM, and even how investors interpret growth signals. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

Why this is different from every previous SaaS “innovation cycle”

SaaS has absorbed major waves before: mobile, cloud, APIs, collaboration, analytics. Most of those waves improved usability or expanded distribution, but they did not alter the fundamental unit of value. AI does. The reason is simple. When software starts completing work that previously required human time, value is no longer tied to the number of seats. It is tied to the number of tasks executed and the quality of outcomes.

That creates a tension that many teams are already feeling: buyers want predictable spend, but AI cost and value often scale with usage. McKinsey highlights “price predictability” as a core friction, quoting a CFO who cannot forecast what the organization will spend on AI in the quarter because usage is spread across many vendors and is hard to predict. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

This is why the story of AI in SaaS is not primarily about chatbots. It is about economic alignment.

The hidden engine of disruption is falling inference cost

The most underappreciated driver of the AI SaaS transition is cost collapse. When inference becomes dramatically cheaper, entire categories of “AI as a premium add on” become difficult to defend. If the marginal cost of generating a useful output drops fast enough, the market tends to treat that capability as expected, then bundled, then commoditized.

Stanford’s AI Index shows how extreme this has been. The report notes that the inference cost for a system performing at the level of GPT 3.5 dropped by more than 280 fold between November 2022 and October 2024. It also reports hardware costs declining by about 30 percent annually and energy efficiency improving by about 40 percent each year. These are not small deltas. They are structural forces that reshape pricing power. https://hai.stanford.edu/ai-index/2025-ai-index-report

When costs drop that quickly, two things happen at once. First, more vendors can ship AI features because the compute bill becomes less terrifying. Second, differentiation shifts away from “having AI” toward distribution, workflow depth, proprietary context, and switching costs.

AI becomes a labor budget competitor, not an IT line item

Traditional enterprise SaaS competed within IT budgets. AI driven software increasingly competes against labor budgets, because its pitch is not “better software.” It is “less manual work” or “faster throughput.” McKinsey explicitly argues that AI plus SaaS expands the addressable market beyond IT budgets to include labor. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

That shift changes who buys, how value is justified, and what gets measured. It also changes why pricing becomes politically sensitive inside enterprises. A per seat model is easy for procurement. A per action model touches operational throughput. Once the buyer starts evaluating software based on work completed, internal stakeholders begin debating whether the “work” is defined correctly and whether outcomes are auditable. This is where many agentic products will struggle, not because the tech fails, but because the organization cannot agree on what “done” means.

The real battleground is pricing design, not model quality

In early SaaS, the dominant question was “how do we acquire users cheaply.” In AI SaaS, a central question is “what is the meter that aligns with customer value while keeping spend predictable.”

McKinsey describes the emerging design space: hybrid models, credits, pay per action, buckets, platform fees untethered to user counts, and various consumption structures. It also observes that global enterprise spending on AI applications increased eightfold over the prior year to nearly 5 billion dollars, yet still represented less than 1 percent of total software application spend, which helps explain why monetization remains early and contested. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

This is the paradox of 2025 and 2026. Adoption narratives are loud, but monetization is uneven, and the pricing architecture is still being invented in public.

Open weight models compress differentiation and accelerate bundling

Another major force is the narrowing gap between open weight and closed models. Stanford’s AI Index notes that open weight models have been closing the performance gap with closed models, reducing the difference from 8 percent to 1.7 percent on some benchmarks in a single year. https://hai.stanford.edu/ai-index/2025-ai-index-report

This matters because it reduces the defensibility of “we have better AI.” More vendors can access strong base capabilities. Competitive advantage migrates to data access, workflow integration, trust, safety governance, and distribution.

As base models become more substitutable, customers become less willing to pay premiums for generic AI capabilities. That pushes vendors toward bundling and toward charging for higher order outcomes, not raw model access. In practice, many AI features will be treated like search, export, or dashboards. Expected, not premium.

The market is simultaneously exploding and polarizing

The investment signal is extreme. PitchBook has reported AI capturing a very large share of venture capital dollars in 2025. One PitchBook analysis noted AI’s share rising to a record 63.3 percent in a period it covered, illustrating the level of capital concentration flowing into AI related categories. https://pitchbook.com/news/articles/investors-are-plowing-more-money-into-ai-startups-than-they-have-in-any-other-hype-cycle

At the same time, the “AI label” is spreading across everything, which creates a credibility problem. PitchBook itself has warned that AI’s hottest metrics and definitional boundaries are getting harder to trust as more companies rebrand and as reporting becomes inconsistent. https://pitchbook.com/news/articles/ais-hottest-metric-is-getting-harder-to-trust

This is a classic dynamic in platform shifts. Capital floods in early, narratives outrun measurable outcomes, and the eventual winners are decided by distribution and repeatable economics, not by novelty.

If you want to understand disruption, watch three things

First, watch whether customers can forecast AI spend without creating bureaucratic friction. If they cannot, adoption will stall outside of pockets of enthusiastic teams, and procurement will force pricing simplification.

Second, watch whether vendors can publish quantifiable ROI with credible measurement. McKinsey notes that only a minority of vendors publish quantifiable ROI, and that change management often dominates the cost of scaling AI beyond pilots. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

Third, watch the cost curve and the bundling response. When inference keeps getting cheaper, the market will punish vendors that try to charge premium pricing for commodity outputs. Differentiation will increasingly come from context and orchestration, not from raw generation.

Conclusion

AI is forcing SaaS to confront a deeper truth. In the subscription era, the industry monetized access. In the AI era, the industry must monetize work. That transition is messy because it collides with procurement norms, forecasting expectations, internal politics, and the uncomfortable question of how organizations measure productivity.

The disruption is already underway, but it will not be decided by who demos the best agent. It will be decided by who aligns pricing to value, reduces adoption friction, and builds trustable systems that enterprises can govern, audit, and forecast.

References and further reading (full URLs)

McKinsey, “Upgrading software business models to thrive in the AI era”
https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/upgrading-software-business-models-to-thrive-in-the-ai-era

Stanford HAI, “AI Index Report 2025” (page and report hub)
https://hai.stanford.edu/ai-index/2025-ai-index-report

Menlo Ventures, “2024: The State of Generative AI in the Enterprise” (methodology and survey context)
https://menlovc.com/2024-the-state-of-generative-ai-in-the-enterprise/

S&P Global Market Intelligence press release, generative AI software market forecast to 2028
https://press.spglobal.com/2024-06-06-S-P-Global-Market-Intelligence-Foresees-Rapid-Expansion-of-Generative-AI-Software-Market-by-2028-to-52-2-Billion

PitchBook, “Investors are plowing more money into AI startups than they have in any other hype cycle”
https://pitchbook.com/news/articles/investors-are-plowing-more-money-into-ai-startups-than-they-have-in-any-other-hype-cycle

PitchBook, “AI’s hottest metric is getting harder to trust”
https://pitchbook.com/news/articles/ais-hottest-metric-is-getting-harder-to-trust

Qrvey, “AI in SaaS in 2026: Current State, Adoption, Use Cases & More” (useful as a market oriented overview, with vendor framing)
https://qrvey.com/blog/ai-in-saas/