Fewer than one in five companies are using the technology to its best advantage, experts say, while almost one in three are only experimenting with it.
Gartner researchers issued the warnings on Tuesday at the company's Data and Analytics Summit in Sydney, where they also predicted AI agents could take over some tasks but were not ready to replace human employees.
The predictions come after tech firms, including Amazon and Atlassian, announced large-scale lay-offs due to AI adoption, while others struggled with the cost of accessing AI tokens.
Uber, for example, revealed it had exhausted its 2026 AI coding budget within four months after a company-wide push to use the technology.
The example highlighted the challenge for businesses, Gartner AI analyst Arun Chandrasekaran said, between encouraging employees to test the technology and ensuring they did not overuse it.
"Some of this is less about technology and more about how do you strike that right balance between creating the incentives but … also creating thresholds and guardrails," he said.
"I don't want a user to run a $50,000 bill on a single day but, at the same time, I don't want to block AI."
The so-called 'token-maxxing' trend of encouraging unfettered AI use would change, Mr Chandrasekaran said, as more companies put limits on AI use and developed better ways to predict their spending.
"The token-based pricing model is hard for enterprises to fathom," he said.
"They don't have the tools, processes and means to accurately (forecast costs)."
Only a small number of companies were using artificial intelligence technology effectively in all their operations, Gartner research vice-president Dr Pieter den Hamer said.
"Across industries, about 17 per cent of companies is at what we call a high level of maturity, meaning those companies have managed to scale AI and are using it pervasively," he said.
"There is another 51 per cent who are at medium-level maturity … using AI but only in isolated pockets."
The remaining third of companies were still experimenting with AI tools, Dr den Hamer said.
The technology had evolved enough to tackle some work processes autonomously, he said, but could not replace human workers, and many job cuts had been blamed on AI unfairly.
"We found that the vast majority of lay-offs announced in the media actually have nothing to do with AI, but are using AI as a scapegoat," he said.
"There is only a small percentage of lay-offs that actually can directly be attributed to AI."
