A new fashionable word is suspended from companies as they rush into AI

Artificial intelligence capacities develop quickly and companies around the world are frantically trying to follow and implement AI tools, but there are consequences for a sloppy execution.

In fact, 79% of companies around the world expect an “AI debt” due to poorly implemented autonomous tools, according to a new ASANA report on the state of AI at work which interviewed more than 9,000 knowledge workers in the United States, the United Kingdom, Australia, Germany and Japan.

The report stressed that companies are not prepared and do not have the infrastructure and surveillance necessary to promote fluid collaboration between human employees and autonomous AI agents. Different from the generative AI, agents act independently, can initiate actions and recall the previous work they have carried out. Some examples include the Openai operator and the Anthropic Claude.

AI debt is the cost of not properly implementing emerging autonomous systems, Mark Hoffman, an expert from the Asana work innovation laboratory, told CNBC.

“These costs could be money costs. They could also be lost time, which concerns money. This could also be a lot of things you need to cancel, which is expensive from a financial point of view. This burns people to do so. These are all the costs associated with poor implementation,” said Hoffman.

The report indicates that the debt could manifest itself as security risks, poor data quality, low impact AI agents that will waste time and resources for human employees and a management gap in management.

Hoffman said that this is not an exhaustive list and that “debt” could look like a lot of code created by AI that does not work properly or content generated by AI that no one uses.

New research by Betterup Labs and Stanford Social Media Lab even revealed that 40% of offices in the United States received a “workslop” generated by AI, which researchers have defined as content that looks good but lacking in substance.

He created almost two additional hours of work for the people who met him, an invisible tax of $ 186 per month, and one hit $ 9 million in productivity in one year, according to research.

“There are significant investments in this space at the moment, and finally it is a question of whether these investments will be chargeable,” said Hoffman.

Henry Ajder, founder of the AI ​​consulting company, the consulting company, and adviser to the British government, Meta and the synthesis of AI video startup, underlined the need for thoughtful implementation and structures.

“People who are CTOs or innovation officers, the good ones I worked with, those who, I think, have made the best position to succeed, they do not cover sugar that it will cost … As for any type of fundamental retouching, you will have problems, you will have bumps on the road,” said Ajder in an interview.

Asana’s report revealed that despite the adoption of the AI ​​70% in 2025, compared to 52% in 2024, workers are also faced with higher levels of digital exhaustion.

Digital exhaustion increased to 84% in 2025, against 75% the previous year, while unmanageable workloads also increased to 77%, according to the report.

MONA MOURSHED, CEO World Founder of Generation, a job organization based in the United States, told CNBC that despite companies deploying AI tools and encouraging the use of IT, workers still have difficulties.

“The main reason why they are struggling, and we know it by also speaking to our own former students, is that the case of use for how and why are you supposed to use this AI tool in the flow of your work is often missing,” said Mourshed.

“Without a clear understanding of what is the case of use that will improve this particular, faster, cheaper task … This is what leads to exhaustion, because you do not know what is the expected result,” she added.

Mourshed noted that companies invest in AI in the hope that work overnight will be better done, faster and cheaper, but they do not offer the training or directives necessary to allow improvements.

“It is not a magic silver ball, and all of a sudden, it does whatever you want once you install it … It will be a much more painful trip to reach these advantages than the companies that thought it.”

The IA expert, AJDER, said that the right strategy carefully tests the use of AI and the construction of infrastructure around him rather than rushing into the unprepared race.

“You do not start by simply integrating, you start with piloting, you start with the scope, with sand, by testing these systems,” he said.

This includes everything, from correct training for employees, to reflect on the type of AI models that the company may need. It is much more difficult to respond to errors or dysfunctions when there is no procedure in place.

“So I’m not saying that you cannot take any risks in terms of using AI, but it must be calculated and it must be extended,” said Ajder.