Spending on autonomous AI agents will increase fivefold by 2028: Gartner warns
Making tokens cheaper will not save the situation when complex workflows consume them many times more, warns the analytical company Gartner.
Making tokens cheaper will not save the situation when complex workflows consume them many times more, warns the analytical company Gartner.
According to Gartner's forecast, by the end of 2028, the cost of workflows using autonomous AI (Agentic AI) will increase more than fivefold due to the transition of users to more complex scenarios for using this technology, The Register writes .
While Nvidia and other tech giants are promoting inference and AI agents as the next wave of artificial intelligence, Gartner warns that the cost of implementing such systems will increase even as the base models become cheaper.
Analysts have studied the young market for AI agents — systems capable of acting autonomously to achieve a set goal — and see many challenges ahead.
Even leaving aside the significant security risks, such software agents are significantly more complex than conventional chatbots.
Gartner believes that falling model prices are driving the development of more complex workflows. However, the increased token consumption in such processes offsets the savings from lower fees and increases the overall cost of inference.
In other words, tokens are becoming more cost-effective, but these savings are not keeping up with the increase in the cost of advanced AI capabilities. The pace of innovation is outpacing the cost reduction curve.
“The harsh economics of the ‘inference paradox’ are clearly evident in the difference between a simple chatbot and an AI agent. While a simple chatbot only has to read and interpret a query and quickly produce a likely correct answer, an AI agent is forced to constantly reason, negotiate, and question its own actions,” says Will Sommer, senior research director at Gartner.
All these processes accumulate: redirecting the task to an agent-based reasoning model increases the cost of inference by at least five times, and even more as the task becomes more complex.
To make such tools pay for themselves, they require either significantly higher performance than the base models, or better optimization of inference, routing, and orchestration. This involves assigning each task to the most cost-effective and capable model.
The transition of some providers from a fixed subscription to pay-as-you-go has only complicated the situation. High token consumption scenarios under this model lead to uncontrolled bill growth.
It's no wonder that earlier this year, Gartner predicted that about 40% of organizations would scale down or decommission their AI agents due to problems with this much-hyped technology.
Recall that in June, Gartner predicted that by 2028, AI coding tools could cost more than the average developer salary .
Overspending by 860% of the original budget. Amazon loses millions of dollars due to lack of control over AI costs
Overspending by 860% of the original budget. Amazon loses millions of dollars due to lack of control over AI costs
Startup Accidentally Spent $30,000 on AI Tokens in a Month
Startup Accidentally Spent $30,000 on AI Tokens in a Month
Ukrainian IT guy created a CLI tool that simplifies front-end development with AI and saves tokens
Ukrainian IT guy created a CLI tool that simplifies front-end development with AI and saves tokens
Claude appears to be significantly more “greedy” for tokens than OpenAI’s GPT-5.x — study
Claude appears to be significantly more “greedy” for tokens than OpenAI’s GPT-5.x — study
Read the country's main IT news in our Telegram
Read the country's main IT news in our Telegram
