Agentic AI costs set to balloon fivefold by 2028

By 2028, the expenses associated with agentic AI workflows are projected to surge more than fivefold, according to a forecast from research firm Gartner. Despite Nvidia and other technology leaders promoting inference and agentic AI as the next phase of the AI trend, Gartner cautions that the cost of deploying these systems will rise even as foundation models become more affordable.

The analyst firm has examined the emerging realm of AI agents – systems capable of operating autonomously to achieve a specific objective – and identifies numerous challenges ahead. Excluding significant security concerns, these software agents prove considerably more intricate than chatbots. Gartner asserts that the decreasing prices of models are enticing users to construct more complex workflows.

However, these workflows consume more tokens, and these savings may not keep pace with the escalating expenses of more sophisticated AI functionalities. Gartner asserts that the rate of innovation outpaces the cost curve. Senior director analyst Will Sommer of Gartner explains that while a simple chatbot must swiftly process and reply to a query, an AI agent must continuously reason, negotiate, and scrutinize itself.

These processes contribute to a fivefold increase in inference costs, and potentially even more as tasks become more intricate. Securing a return on investment from such advanced AI tools necessitates either substantially higher returns than basic models provide or improved optimization of inference, routing, and orchestration. Gartner warns that attaining a positive return on advanced AI investments demands either significantly better returns than basic models offer or more efficient management of inference, routing, and orchestration.

One potential solution includes assigning each task to the most cost-effective model capable of handling it. The shift from flat-rate subscriptions to usage-based billing by certain AI providers, as reported by The Register, may not have alleviated the issue. Token-intensive workflows can result in escalating costs under the new billing model.

It is no surprise that Gartner had earlier predicted that 40 percent of organizations would demote or discontinue AI agents due to problems with the highly publicized technology. The analyst company has also forecasted that at least half of all generative AI projects would exceed their budgets due to suboptimal architectural choices and a scarcity of expertise, while most attempts to construct custom models would be abandoned.