Why software subscription sales do not prove AI value

Subscription stability leaves AI returns unresolved

A software subscription can survive renewal while the value of its AI features remains unclear. Software providers and enterprise customers are both seeking returns on AI investment, even as the technology and its pricing change quickly. For chief information officers, or CIOs, continued subscription sales leave an immediate purchasing question: which capabilities are being used, what each vendor charges for them, and whether the business outcomes justify another commitment.

Answering that question gets harder when charges are difficult to forecast. Some providers are moving toward usage- or outcome-based pricing, while many CIOs still struggle to connect AI spending with business results. Understanding these purchasing decisions starts with separating the changes providers have already experienced from those they expect.

Subscription losses and expected seat reductions measure different changes

Current revenue patterns show how much providers still depend on established software pricing. A survey report found that subscription pricing remains the leading revenue model for 66% of software providers. One-time perpetual payments, which give customers an ongoing software license for an upfront payment, rank second: 44% of companies report significant revenue from them. These percentages measure different things: one identifies the leading model, while the other identifies a significant revenue contributor, so they cannot be added together as shares of a market total.

Alongside those established models, providers report usage- and outcome-based arrangements. Prepaid usage means customers pay before consuming a service, while postpaid usage means they pay afterward; outcome-based pricing links the charge to a result. Some providers are also moving toward token-use pricing, which charges for units of text or other data processed by an AI model. Each approach changes how the bill is calculated, but their presence alone does not show that subscriptions are being replaced across the market.

Evidence of subscription losses is narrower than the disruption debate might suggest. Newly released AI models prompted fears of a “SaaSpocalypse”: a disruption to the use and pricing of software as a service, or SaaS, which customers access as an ongoing service. Yet only 7% of surveyed providers reported selling fewer software subscriptions because of AI. The measure captures losses providers already attribute to AI.

Providers’ expectations for AI agents concern a different measure and a future change. Nearly one-quarter of surveyed providers anticipate that wider adoption will soon reduce demand for subscription seats, the user licenses customers buy. AI agents can carry out tasks, so their adoption could change how many licensed users an enterprise needs. But an expected reduction in seats and a reported reduction in subscription sales cannot be treated as successive stages of an established trend.

For buyers weighing those expectations, the most visible AI pricing changes may give an incomplete picture of everyday enterprise software. Segerer said most providers that enterprises buy from are changing prices cautiously and considering their CIO customers as they do so. So changes around frontier models, the most advanced AI models, need to be distinguished from pricing decisions by vendors supplying routine business applications.

Existing vendor relationships already shape how employees encounter AI. Most employees use AI features supplied by their current software producers in their daily work, Segerer said. CIOs must therefore assess AI within purchases they already manage, often across several vendors. A subscription can remain in place while the buyer is still working out whether particular capabilities justify their cost.

Underused features and vendor-specific AI pricing complicate the assessment

That assessment starts with whether newly introduced capabilities get enough use. “Many technology companies have rushed to implement AI capabilities, but as AI develops and changes quickly, some of those capabilities are not getting used enough,” Segerer said. Rapid releases give customers more features to evaluate, while technological change can alter which ones remain useful. Buyers therefore need to understand what each available feature contributes to employees’ work.

Understanding that contribution helps CIOs decide which services to keep buying. Technology leaders still learning how employees use AI internally have less clarity about which parts of an offering are essential. Access lets employees use a capability; examining its use shows where it fits into their work. This gives buyers a basis for assessing the charges attached to it.

Charges need their own examination because vendors apply consumption-based pricing differently. “Understanding what exactly a consumption-based model is per vendor does take a little bit of work,” Segerer said. Consumption-based pricing makes payment depend on use, but buyers must still interpret how each supplier calculates the bill. Even the same level of employee activity can require separate pricing assessments across suppliers.

The work grows with the number of vendors a CIO manages. “When I’m a CIO, that problem compounds to me because every vendor might have a different way of dealing with it,” Segerer added, describing the buyer’s position hypothetically. Learning one supplier’s approach still leaves the others to assess. These differences help explain why forecasting spending and connecting it to business outcomes can remain difficult while subscriptions continue to dominate providers’ revenue.

Use and cost should guide customization and renewal

To turn these assessments into purchasing decisions, CIOs can compare essential uses with the vendor’s charges and the business outcomes the organization seeks. Frequent use shows that a feature has a place in employees’ work, but its contribution must still justify its cost. Building that view takes time while capabilities and prices continue to change.

Where a supplier allows customization, buyers can use this comparison to decide which services to retain and which to change. Customization can limit unnecessary spending, although the available plans determine whether a suitable adjustment is possible. Buyers must compare the proposed offering and its charges with the uses they have identified as valuable.

Renewal offers another opportunity to act on that comparison. Segerer said CIOs retain some negotiating power to seek lower prices during contract renewals, even as pricing models change. Buyers can draw on actual use to identify which services to retain and which charges to discuss. If agents change demand for user licenses, buyers may also need to revisit seat counts alongside the terms attached to AI capabilities. Negotiation can secure a lower price or a better-fitting offering, but renewal still depends on whether the capability’s business benefit justifies the vendor’s terms.

Key takeaways for decision-makers

Separate current sales from future seat demand: Only 7% of surveyed providers report fewer subscriptions because of AI, while nearly one-quarter expect agents to reduce seat demand. CIOs evaluating future license needs need to distinguish reported losses from forecasts.
Assess AI use and pricing vendor by vendor: Underused features and differing consumption charges complicate spending forecasts. CIOs can map employee use to each supplier’s billing terms and assess whether business outcomes justify the cost.
Use renewals to align spending with value: CIOs retain room to negotiate prices and, where available, customize services. Usage evidence can inform which capabilities to retain, which charges to challenge and whether seat counts still fit business needs.