
The per-seat subscription is no longer the default unit of SaaS pricing. According to Metronome's 2025 State of Usage-Based Pricing report, 85% of SaaS companies have adopted some form of consumption-based pricing.
The shift is driven by AI. Features that involve tokens, credits, and compute minutes cannot be sensibly priced per user. So vendors changed the model. Atlassian now charges $0.30 per conversation when AI usage exceeds included limits.
The result: 78% of IT leaders have received an unexpected charge on a SaaS bill due to consumption-based or AI pricing.
Why procurement isn't keeping up
Most enterprise procurement processes were built for fixed costs. Annual contracts, seat counts, renewal dates, these map cleanly to budgets and approval workflows. Consumption pricing does not.
The problem is structural, not a matter of effort:
The forecasting tools most enterprises use were designed for fixed costs. They cannot account for usage spikes driven by agentic workflows, new AI features vendors activate by default, or token consumption that scales with model capability rather than user count.
What good looks like
The organizations managing consumption pricing effectively share one characteristic: they have visibility at the usage layer, not just the invoice layer. They know which teams are consuming what, against which models, before the bill arrives.
That requires per-user, per-model, per-application cost attribution, something traditional SaaS management platforms were not built to provide.
CloudEagle.ai tracks AI token consumption and AI usage, giving IT and finance teams the granularity they need to forecast consumption-based costs, set thresholds, and act before spending exceeds budget.
Let the headlines come straight to you with Access Granted — a monthly edition of Okta announcements, expert perspectives, analysis, and more.
