From Tokens to Outcomes: Rethinking AI Tool Pricing Models in Enterprise SaaS

The Tokenmaxxing Era and Its Costs

For the past two years, many enterprises ran internal campaigns to drive AI adoption, sometimes incentivizing employees through leaderboard rankings and usage targets. The implicit assumption was that more AI usage would naturally translate into better outcomes. Analysts now describe this period as “tokenmaxxing” — a no-holds-barred experimentation phase driven largely by fear of falling behind.

The consequences are becoming visible on balance sheets. Astronomical compute bills, unpredictable cost curves, and underwhelming ROI have forced a reckoning. As Liz Miller, VP and Principal Analyst at Constellation Research, put it plainly: tokenmaxxing is “probably one of the most detrimental things to actually finding success in AI that we could have gone through.”

The core problem was structural. Enterprises were deploying frontier models — the most capable and most expensive — indiscriminately across tasks that could have been handled by far simpler tools. The incentive structure rewarded volume, not precision.

Outcome-Based Pricing: The Model Taking Shape

Against this backdrop, a wave of SaaS vendors is rethinking how they charge for AI. The direction is clear: away from per-seat subscriptions and raw token consumption, toward pricing tied to measurable outcomes.

Pegasystems is among the most prominent examples. Rather than billing for the compute consumed during agentic workflows, Pega is moving toward charging per resolved case — a model that directly aligns vendor revenue with customer value. Zendesk and Intercom have introduced similar outcome-based pricing options. Intuit’s CTO has publicly signaled the company is exploring alternatives to traditional subscription models.

This is not merely a pricing experiment. It reflects a deeper shift in how enterprise buyers are evaluating AI tools — and how vendors must now justify their value propositions.

The Architecture Behind the Economics

What makes outcome-based pricing viable is not just commercial intent — it requires a fundamentally different technical architecture. Pega’s approach offers a useful model for understanding this.

AI Ops as a Discipline

The shift toward cost-conscious AI deployment is also creating demand for a new operational skill set. Miller describes this as “AI ops” — the capability to decide where, when, and how different AI models and functions are applied within a given workflow.

This includes knowing when to reach for a traditional deterministic machine learning algorithm rather than a generative model. Many enterprises already have years of investment in classical ML pipelines. Layering agentic workflows on top of those existing systems — rather than replacing them wholesale — is emerging as a more pragmatic and cost-efficient architecture.

The implication for tool selection is significant. Enterprises evaluating AI platforms will increasingly need to assess not just raw capability, but composability: how well a tool integrates into a hybrid stack of frontier models, small models, and classical ML.

Governance: The Concentrated Risk

Outcome-based pricing and design-phase reasoning are not without risk. Keith Kirkpatrick, VP and Research Director at Futurum Group, identifies a structural vulnerability in Pega’s approach: by concentrating reasoning in the design phase, the model places enormous pressure on the quality of upfront configuration and governance.

“Enterprises will scrutinize whether Pega can deliver consistent agent performance across complex, regulated workflows without introducing new bottlenecks or governance gaps,” Kirkpatrick noted in a recent research analysis.

If the design phase is well-governed, the model can deliver reliability and cost predictability at scale. If governance falters, the deterministic runtime becomes a liability rather than an asset. For enterprises in regulated industries — financial services, healthcare, insurance — this is not a theoretical concern.

What This Means for AI Tool Selection

Alignment of incentives matters. Outcome-based pricing forces vendors to stand behind the results their tools deliver. When evaluating platforms, buyers should ask whether the vendor’s commercial model is aligned with actual business outcomes — or simply with consumption.

Architecture determines cost trajectory. Tools that separate reasoning from execution, and that support hybrid model stacks, will be better positioned to deliver predictable costs at scale. This is now a legitimate evaluation criterion alongside capability benchmarks.

Governance is non-negotiable. As reasoning shifts upstream into design and configuration phases, the governance frameworks around those phases become critical. Enterprises should assess vendor maturity in this area with the same rigor applied to security and compliance.

Small models are a strategic asset. The ability to fine-tune and deploy smaller open-weights models for specific tasks is no longer a niche capability. It is becoming a core component of cost-optimized enterprise AI architecture.

The Broader Market Signal

Pega’s stock is down 46% year to date — a figure that reflects broader market anxiety about SaaS vendors whose value propositions are being disrupted by the very AI capabilities they are trying to monetize. The pressure is structural, not cyclical.

Vendors that can credibly demonstrate outcome delivery, cost predictability, and governance maturity will be better positioned to survive this transition. Those that continue to sell token consumption as a proxy for value will face increasing resistance from enterprise buyers who have learned, often expensively, that the two are not the same thing.