SaaS vendors shift to usage-based AI pricing, forcing CIO budget rethinks

SaaS pricing models are fundamentally changing with AI

Major software-as-a-service vendors are abandoning traditional per-seat subscription pricing in favor of usage-based models as AI features become central to their platforms. The shift is forcing enterprise technology leaders to rethink budgeting strategies and cost forecasting.

GitHub moved to token-based pricing for its premium request model on June 1, charging separately for input, output, and cached tokens according to published API rates. Zendesk and Workday have similarly restructured their pricing as AI capabilities expand across their platforms.

The change reflects a fundamental business reality: per-person pricing breaks down when AI automates tasks that previously required human headcount. "Many people look at AI as a technological upgrade. It's not just a technological upgrade. It's really a fundamental shift in how SaaS companies will operate," said Marko Markov, technology, media and telecommunications senior analyst at RSM.

Dhaval Moogimane, software industry lead at West Monroe, pointed to customer service as a clear example. If AI reduces the number of support agents needed over time, vendors can't sustain revenue on a per-agent model. They must find alternative metrics, typically moving to consumption or outcome-based pricing.

Why it matters

This pricing transition creates immediate budget uncertainty for IT organizations accustomed to predictable annual software costs. CIOs must now establish baseline token consumption patterns and build forecasting models for variable AI usage—a capability most finance teams haven't needed before. The shift also affects innovation budgets, as experimental AI projects consume tokens without guaranteed returns.

CIOs face forecasting challenges with token consumption

Michael Corrigan, CIO at World Insurance Associates, told CIO Dive that AI pricing models with vendors have changed "very rapidly" over the last three to six months. The company, a heavy SaaS user, had grown accustomed to planning around predictable per-seat costs tied to organizational headcount.

Many vendors now charge a combination of seat licenses and consumption-based fees. "What we have to do is a little bit differently than in the past and establish those baselines and forecasts," Corrigan said. While seat-based costs remain straightforward to plan for, token consumption is not.

The complexity extends beyond IT departments. Organizations must track how employees across all functions burn through tokens, creating a moving target for budget planning. Corrigan's team is working to understand usage patterns that will help forecast costs under different scenarios.

Markov noted that understanding token requirements becomes critical when organizations face unexpectedly high bills or exhaust their token allotments too quickly. Better forecasting enables CIOs to adjust spending based on seasonality rather than paying flat annual fees. A hotel chain, for example, could budget for higher token usage during summer peak season and reduce consumption in slower winter months.

Innovation budgets face new constraints

The pricing shift particularly affects companies running AI-driven pilots with in-house developers. "AI adoption is an R&D process," said Alex Bakker, director and primary research lead at ISG. Some experiments will fail, making token spend appear wasteful—but successful projects could deliver outsized returns.

Organizations uncomfortable with that risk may be better served relying on commercial vendors with built-in AI features rather than developing custom solutions.

Bakker also highlighted an expectation gap: employees who use free ChatGPT accounts for basic tasks may assume enterprise AI should handle complex processes with similar ease and minimal cost. "That positive surprise probably created a bit of irrationally high expectations," he said.

More than half of technology executives expect usage-based revenue to grow by 2027, according to a Revenera report. For CIOs, that means the current transition period is just the beginning of a longer adjustment to variable AI costs.

These details were first reported by CIO Dive.