The illusion of precision

Organizations today have access to more data, more signals, more customer insights, and more analytical power than ever before. Yet many leaders are discovering that greater visibility creates new commercial challenges. 

AI has dramatically improved visibility into how products are used, what outcomes they influence, and how customers engage. What it has not done is answer the more fundamental question: what exactly should customers pay for?

Nowhere is that question more visible than in the pricing of AI. 

AI agents are reaching the market faster than companies have fully worked out how to commercialize them. The category remains young, and many organizations are still experimenting with how value should be priced, packaged, and communicated.

Imagine a software CEO 18 months from now reviewing the performance of a newly launched AI agent. Adoption is growing. Usage is measurable. The cost to serve through token speed is real. And customer outcomes are increasingly visible.

Yet the commercial question remains unresolved: should value be captured through seats, usage, outputs, outcomes, or some combination of all four? The leadership team is not short on options. The challenge is deciding which model best reflects the value customers actually recognize and are willing to pay for.

More visibility, harder choices 

This is one of the defining pricing questions emerging around AI agents. Organizations now have unprecedented visibility into how customers use AI. The harder task is deciding how that value should be captured in a pricing model.

AI agents do not fit neatly into traditional commercial categories. At times they behave like software. At others they look more like services. Increasingly, they function somewhere in between, contributing directly to business outcomes while remaining embedded within broader workflows.

Traditional SaaS pricing models, built around user access and feature tiers, were never designed with these dynamics in mind.

Buyers feel the mismatch too. In recent market research, 86% of enterprise software buyers said they prefer usage- or outcome-based pricing over traditional seat-based structures when evaluating AI solutions.¹

The appetite for new approaches is clear. What remains less clear is whether a single pricing model will emerge as the standard. Every additional pricing dimension that AI makes measurable creates another decision leaders must evaluate and explain.

Definition before measurement

Pricing AI agents is not primarily a measurement problem. The harder question is not what can be measured. It is what should be monetized. Before organizations decide how to track usage, outputs, resources required, or outcomes, they must decide which form of value customers perceive and reward.

What truly reflects value in a way customers recognize? What supports adoption rather than slowing it? What protects margin as compute costs fluctuate? 

These are questions about value. The analytics can help answer them, but they cannot replace them. 

In many ways, AI is exposing assumptions that have existed in software pricing for years. Models that worked well when software was sold by the seat may not translate as easily when software begins performing work on behalf of the user. The metrics have become more sophisticated. Commercial thinking is still evolving alongside them. 
 

Three questions before the model 

Before selecting a pricing metric, leadership teams should answer three strategic questions. The answers determine whether a pricing model should lean toward access, usage, outputs, outcomes, or a combination of approaches.

The first is autonomy.

How independently does the agent operate? The greater the agent's autonomy, the more likely cutomers are to perceive value in outcomes delivered rather than access provided.

The second is attribution.

How defensible is attribution? Outcome-based pricing is attractive because it aligns payment with results, but it only works when both buyer and seller can confidently agree on what drove the outcome.

The third is specialization.

How specialized is the use case? Specialized agents often command premium pricing because their value is tied to domain expertise, accuracy, or measurable business impact.

Together, these answers provide a practical decision filter for pricing design. They help determine which pricing structure best aligns value creation with value capture.

Why hybrid models are gaining traction 

Simon-Kucher’s recent Global Software Study found that 45% of companies plan to use two or more pricing metrics for their AI offerings². At first glance, this can look like uncertainty. It may also reflect the reality that AI creates value in multiple ways simultaneously. 

A platform fee paired with usage credits. Or a user fee combined with output fees. Or a base subscription supported by performance incentives. 

These structures look like compromises only when measured against the elegance of a single metric. In practice, they are the most honest way to price something whose value is partly predictable, partly contingent, and partly outcome driven.

Hybrid pricing is not a substitute for strategy. Done well, it reflects how value is created and captured. Done poorly, it simply transfers uncertainty from the seller to the buyer. The goal is not to combine metrics for the sake of flexibility, but to create a pricing structure whose logic customers can understand and whose economics the provider can sustain.

Customers are often willing to accept complexity when the underlying logic feels coherent and aligned with value. The challenge is not avoiding complexity. It is making complexity understandable and creating space to evolve in step with the technology. 

The decision the data cannot make 

What’s next? For the CEO evaluating an AI agent's commercial performance 18 months from now, the challenge is not a lack of information. What is missing is the prior step: a clear answer to what the agent is worth, who receives that value, and how that value should be reflected in the commercial model. 
AI will continue to improve the visibility companies have into usage, performance, and outcomes. But pricing decisions will still require judgment. 

The winners will not be the companies with the most data; they will be the ones that can define value clearly, translate it into a commercial model customers understand, and evolve that model as the technology matures.