Agentic SaaS Pricing

How AI Agents redefine Software’s Value

For the past two decades, the Software-as-a-Service (SaaS) business model has largely relied on per-user or per-seat pricing. This works well as long as software is primarily used by humans. However, with the rise of agentic AI systems, the argument “more users equals more value” is rapidly losing its impact. AI agents are increasingly replacing human interaction with software. We are entering the age of Agentic SaaS (ASaaS). This is no longer about usage, but about achieving goals and results. The economic logic therefore requires a reorientation of SaaS pricing and the value proposition. But it’s not just the user role that is fundamentally changing.

Why the user-based model no longer works

The impending farewell to the per-user model has - among others - two main reasons.

1. Humans are no longer the main actors

In the ASaaS model, humans transition from active users to passive goal-setters. Instead of clicking buttons or entering data in an e.g. ERP system, they formulate goals in natural language (Conversational AI). In the agentic world an input might be: “Create a cash flow projection for Q4 considering the last two years and planned expenses for the launch of our new product”. The AI agents take over the interaction with the software and available data sources. Therefore, the customer will hardly be willing to pay the same price as for access rights for e.g. 100 previous SaaS users when three agents do their job – even though the ASaaS provider delivers enormous efficiency gains and a significantly better user experience.

2. The ASaaS cost structure is different

Agents work around the clock, without fatigue, simultaneously worldwide, will likely be networked independently of manufacturers in the future, and scale in seconds. These processes are based on a different cost structure: ongoing training costs and model adaptation, as well as GPU/TPU time, high energy consumption due to the constant processing of large data volumes, scalable cloud resources, etc. According to OpenAI, GPT-4 Turbo costs $0.01–0.03 USD per 1,000 tokens (so-called Inference Cost, July 2025). When multiple agents are used, for example, in conjunction with those specialized in certain tasks, these costs can accumulate very quickly. AI systems are therefore more cost-intensive to operate and maintain than static SaaS solutions. The user model does not reflect this cost structure and can pose significant risks for ASaaS manufacturers.

New Pricing Models for Agentic SaaS

Against this background, providers are experimenting with new pricing approaches that better suit ASaaS. The term “experimenting” is deliberately chosen here, as much is still in development.

Outcome-Based Pricing (Value-as-a-Service)

Prices are based on concrete results – for example, per document created, per qualified lead, or as a percentage of savings achieved. Revenue sharing is also conceivable in some areas. The software vendor receives a share of the generated added value. This is particularly attractive in performance-oriented industries such as e-commerce or marketing with a clear ROI. Examples:

• Customer support: Price per resolved ticket or conversation.
• Marketing: Price based on conversion rate or lead generation.
• Productivity tools: Price measured by hours saved.

This model better aligns price and value but significantly increases the complexity in defining, attributing, and tracking results. Variants of the Value-as-a-Service pricing model can be found, among others, at Salesforce, Zendesk, Intercom (customer service), or ChargeFlow (eCommerce).

In addition, there are different consumption models:

Agent Licenses

Instead of buying a license for each human user, customers pay for the number of AI agents that can be active simultaneously or in parallel. Example:

Such a pricing model can be found at Genesys, a provider of call center and customer experience solutions. In addition to other pricing models, the company offers “Concurrent Bots” to customers. The more parallel conversations a company wants to conduct via bots, the more capacity (i.e., licenses for virtual agents) it must acquire.

Task Credits

The customer pays for the tasks successfully completed by the agents. Example:

OpenAI offers various AI models (such as GPT-3.5, GPT-4, DALL-E, …) via APIs, and billing typically occurs based on the consumption of “tokens” or the number of API calls. Customers pay for each completed task (e.g., generating text, creating an image, transcribing speech), which corresponds to the Task Credits model.

Hybrid Models

They consist of a mixture of a basic fee and usage-based components. This can be, for example, a basic tariff for access, combined with usage-dependent costs for higher-value, AI-supported functions. Additional services justify a premium price in these cases. The goal is a balance between predictable revenue for providers and value-based costs for customers. Example:

SAP SuccessFactors offers, in addition to its HR basic functions, AI functions that require more computing power or special data processing and are therefore subject to a fee. These include deep predictive analyses for employee fluctuation or highly automated applicant screenings by bots.

Consequences for Software Providers

The integration of AI into SaaS applications means not only a functional addition but a profound change in business models. Measurable value creation instead of usage is their foundation. This implies, on the one-one hand, that both providers and customers must be able to unequivocally verify, quantify, and attribute results to the software (attribution capability). Product design and analytics must be aligned with these new metrics. Nobody wants to argue with their customer about every invoice because the definition of “success” is unclear or results are not measurable. On the other hand, this can significantly shift the business risk towards the provider, while a lack of value creation in “per user” models is (at least in the short term) more the customer’s problem. Therefore, software providers must delve deeper into their customers’ business processes and develop robust analytical capabilities with regard to results. This will make them a partner of the customer and a direct participant in the business risk.

Source: GOEUROPE CONSULTING, by Dr Jürgen Müller.