As AI agents spread, usage-based plans that charge for what is used, and so-called performance-based pricing models that charge only when AI handles tasks properly, are rapidly expanding in the enterprise software market.

Software pricing has long followed a flat-rate subscription model, with fees charged per user per month. But the view is growing that the model is not well suited to AI, and the industry is moving to overhaul pricing plans in earnest.

More recently, more companies are adopting performance-based models on the view that volume-based models also have limits in an AI environment.

Beyond startups, major software companies are also rolling out performance-based pricing models. Adobe, which leads the global creative software market, is shifting to charging based on actual results rather than on AI tool usage.

According to The Information, Adobe President Anil Chakravarthy (아닐 차크라바티) said it will apply a performance-based pricing plan to CX Enterprise, an AI agent platform unveiled this week.

CX Enterprise includes AI agents that let companies automate digital marketing processes. Adobe CX Enterprise Coworker can coordinate multiple AI agents and collect related business data to set up and execute marketing plans.

It can also analyse why hotel bookings in southern France are weak by combining data from external software such as Adobe apps and AWS databases. Chakravarthy pointed to the problem of customers paying too much when they use AI inefficiently, saying AI suppliers are shifting costs onto customers without creating value for them.

Adobe did not disclose details of its performance-based pricing plan, but The Information reported it cited as an example a method that links fees to the number of advertising campaigns completed by AI agents for travel and lodging companies.

Adobe is not abandoning usage-based pricing entirely. Performance-based pricing will not apply to existing AI tools such as photo and video editing, and the company said subscription- and usage-based pricing would remain for some tools.

A number of enterprise software companies besides Adobe have already introduced performance-based pricing for AI.

Customer service automation companies Zendesk and Intercom introduced performance-based pricing several years ago, and AI-based customer service startup Sierra AI charges based on the number of tasks AI agents actually complete.

Salesforce also created a new metric called an Agentic Work Unit, measuring the number of tasks completed by an agent instead of token throughput in February, The Information reported.

CRM company HubSpot also shifted pricing for its AI agents, Breeze Customer Agent and Breeze Prospecting Agent, from a usage basis to a performance basis. Both agents are charged only when they successfully complete assigned tasks. HubSpot introduced a model that moves away from per-seat and per-interaction billing common in the existing AI market, directly linking costs to outcomes.

HubSpot said performance-based billing had been difficult because many AI systems lacked sufficient consistency to guarantee results. It said embedding agents in its Smart CRM and letting them access customer data, relationship history and work context could deliver more predictable results.