Agentic AI could make the cost of running AI workflows more than five times higher by 2028, according to Gartner. The research firm says the rise will come as organizations move from simple AI interactions to more complex agentic workflows that can perform multiple steps and actions. Gartner describes this as an “inference paradox.” AI inference can become cheaper on a per-unit basis, while the total amount of computing used by each workflow grows. As AI agents take on more complex tasks, they may generate more requests and consume more compute during each process. Gartner says this could make the economics of AI harder for organizations to predict. The forecast comes as businesses increasingly explore AI agents for work that goes beyond generating text or answering questions. The report highlights the need to consider total workflow costs, rather than focusing only on the price of individual AI tokens.
Agentic AI could make the cost of running AI workflows more than five times higher by 2028, according to Gartner. The research firm says the rise will come as organizations move from simple AI interactions to more complex agentic workflows that can perform multiple steps and actions. Gartner describes this as an “inference paradox.” AI inference can become cheaper on a per-unit basis, while the total amount of computing used by each workflow grows. As AI agents take on more complex tasks, they may generate more requests and consume more compute during each process. Gartner says this could make the economics of AI harder for organizations to predict. The forecast comes as businesses increasingly explore AI agents for work that goes beyond generating text or answering questions. The report highlights the need to consider total workflow costs, rather than focusing only on the price of individual AI tokens.