PegaSystems Inc. has announced a new approach to enterprise artificial intelligence with the launch of Pega Infinity 26, introducing a token-free pricing model designed to reduce AI operating costs while improving reliability and scalability for agentic workflows.
The company unveiled the platform at PegaWorld 2026, positioning the new Pega Predictable AI architecture as an alternative to conventional token-metered AI models that have become increasingly expensive for enterprises deploying AI agents at scale.
According to Pega, organizations experimenting with large language model (LLM)-based agents are facing growing concerns over rising token consumption costs and inconsistent outcomes. As AI deployments move from pilot projects to production environments, many enterprises are encountering unpredictable expenses linked to token-based pricing structures.
To address these challenges, Pega has redesigned how AI is applied throughout the workflow lifecycle.
Shifting AI Reasoning to Design Time
At the core of the new architecture is a design-time approach to AI reasoning. Instead of repeatedly relying on complex AI reasoning during runtime, Pega applies intensive AI processing during the workflow creation stage using Pega Blueprint AI and Pega Infinity Studio.
These tools help organizations design, optimize, and build agentic workflows for business-critical operations such as customer service requests, loan approvals, insurance claims processing, and patient experience management.
Once workflows are created and deployed, runtime agents use a lighter semantic AI model to understand user intent, identify the appropriate workflow, and execute tasks according to predefined business logic. More advanced AI processing is only applied when required for specific tasks such as document analysis or interaction summarization.
Focus on Predictability and Cost Control
Pega says this approach delivers two major benefits for enterprises.
The first is predictable outcomes. By following pre-approved workflows rather than re-evaluating processes each time, AI agents can provide more consistent and governed results, particularly important for regulated industries.
The second benefit is predictable costs. Since the system avoids repeated reasoning at runtime, enterprises can significantly reduce AI processing requirements and associated expenses.
To help organizations evaluate potential savings, Pega has introduced an AI Token Cost Calculator, allowing businesses to compare traditional token-based AI costs with Pega’s workflow-driven model. The company estimates that some enterprises could achieve savings exceeding twenty times, depending on workflow complexity and deployment scale.
New Pricing Model Based on Business Outcomes
Rather than charging customers based on seats or token usage, Pega is introducing an outcomes-based pricing structure.
Under this model, organizations pay for completed business cases rather than the number of AI interactions performed behind the scenes. For example, an AI-assisted customer order modification would be counted as a single completed case regardless of the AI resources required to process it.
The new pricing framework is expected to become available to Pega Infinity 26 customers during the third quarter of 2026.
Industry Perspective
Commenting on the launch, Alan Trefler, Founder and CEO of Pega, said enterprises are increasingly recognizing the limitations of token-based AI models.
He noted that organizations derive value from AI when it delivers reliable outcomes at scale and emphasized that Pega’s approach aligns costs with completed business work rather than token consumption.
Industry analysts also see growing pressure on organizations to demonstrate measurable returns from AI investments. As enterprises move beyond experimentation, efficiency, accountability, and cost optimization are becoming central considerations in AI adoption strategies.
With Infinity 26, Pega aims to provide enterprises with a framework for scaling AI-powered workflows while maintaining greater control over both operational outcomes and technology spending.
