An Antidote To TokenMaxxing: Emerging Enterprise Software Pricing Models

The enterprise AI return on investment (ROI) debate rages on as fresh concerns over runaway costs emerge. In April, Uber CTO announced that the firm had burned through its AI budget in 4 months on Claude Code. This has now been followed up by an "unnamed enterprise" receiving a $500 million Claude AI bill and Microsoft pulling Claude Code licences as monthly costs exceeded expectations.

This has added fuel to the fire with buyers now increasingly nervous about ‘TokenMaxxing’: the tendency to maximize AI usage without a clear link to measurable business outcomes. This is an acute concern for software vendors relying on token-linked consumption-based models where marginal costs per interaction are low but can scale rapidly and unpredictably at enterprise volumes, especially within complex agentic workflows.

These announcements may lead some enterprises to hit the brakes. At a minimum, buyers are likely to intensify scrutiny during procurement – on ROI measurability, total cost of ownership (TCO), governance and cost predictability. This comes at a challenging time for many enterprise software vendors that are overhauling pricing models to balance customer adoption with margin protection.

Findings from the Verdantix AI SaaS pricing model report show that rather than letting this shift slow pricing transformation efforts, vendors must use it as a strategic lever to address customer concerns and build long-term trust in AI investments. Communication around this should focus on five key areas:

• Customer engagement and alignment: early and deep collaboration with select customers through workshops, pilots and trials can help vendors understand customer perspectives and ensure alignment before introducing new pricing methodologies. During this process vendors should discuss expected usage and deploy ‘custom’ packages to ensure alignment with goals and expected outcomes.
• Enterprise governance and controls: clear guardrails around AI feature usage that go beyond firm-level aggregation and enable customers to implement role-based controls (such as max tokens consumed per workflow) across humans and agents. During deployment, vendors should work with buyers to deploy these controls from day 1 with appropriate controls and alerts.
• Usage tracking and support: vendors should actively monitor customer usage patterns to identify when organizations are approaching cost or consumption red zones. Proactively alerting customers and recommending optimizations when unusual activity is detected builds trust and demonstrates a commitment to cost transparency.
• Product: vendors should prioritize simple, transparent and navigable pricing models wherever possible, ensuring that customers clearly understand what sits within fixed versus variable costs. Where feasible, introducing caps, tiers or hybrid pricing structures can reduce cost volatility. Additionally, communicating the underlying product architecture (for example, LLM abstraction and switching) and cost drivers can improve long-term confidence.
• Process: vendors must continuously stress-test pricing methodologies to ensure they remain aligned with customer expectations.

Enterprise customers will understandably be spooked by reports of soaring AI costs. However, software vendors should not abandon pricing transformation efforts. Instead, pricing will become a defining battleground for enterprise AI adoption.

For more information on AI-driven enterprise software pricing transformation and actionable frameworks for commercial strategy decisions, read Market Insight: Rethinking SaaS Pricing Models in An Age.