Companies expect to see more costs as a result of inadequately implemented autonomous systems.
Shapecharge | E+ | Getty Images
The capabilities of artificial intelligence are rapidly developing, and while companies worldwide are desperately trying to maintain and implement AI tools, sloppy execution has consequences.
In fact, a new AI report from AI in AI, which surveyed over 9,000 knowledge workers in the US, UK, Australia, Germany and Japan shows that 79% of the world expect to incur “AI debts” as a result of insufficient implementation tools.
The report highlighted that businesses are not ready and the lack of the infrastructure and oversight needed to promote smooth collaboration between human employees and autonomous AI agents. When it differs from the Generation AI, agents can act independently, initiate actions, and recall previous work they performed. Some examples include Openai operators and Claude of Mankind.
AI debt is the cost of not implementing the initial autonomous system correctly, Asana’s Work Innovation Lab expert Mark Hoffman told CNBC.
“These costs can be money costs. You can also lose time related to money. It can also be that you have to turn it back. This is expensive from a financial standpoint. It burns people who have to do it.
The report outlined that debt could manifest as a gap in security risks, poor data quality, the impact of AI agents that waste human employees’ time and resources, and management skills.
Hoffman said this is not an exhaustive list and that “debt” can look like a bunch of code created by AI.
A new study from Betterup Labs and Stanford Social Media Lab has even found that 40% of US desk workers receive AI-generated “Workslops.”
Research has created almost two hours of extra work for people who have been hit by $186 per month invisible taxes and $9 million in productivity in a year.
“There are currently a lot of investments in this sector, and it’s a question of whether those investments will ultimately be rewarded,” Hoffman said.
Henry Ajder, founder of AI consulting firm Latent Space Advisory and adviser of UK government Meta and AI video startup Synthesia, highlighted the need for thoughtful implementation and structure.
“People who are CTO or innovation officers, good people I worked with, people who think I’ve been in the best position to succeed, they don’t sugar-coated the confusion that this costs money… Just like any kind of basic redo, you’re going to hit the road,” Ajder said in an interview.
“It’s not a magical silver bullet.”
Asana’s report found that workers also face higher levels of digital burnout, despite a surge in AI adoption from 52% in 2025 to 70%.
According to the report, digital fatigue rose to 84% from 75% in the previous year in 2025, while unmanageable workloads also increased to 77%.
Mona Mourshed, who founded the global CEO of US-based employment organisation Generation, told CNBC that workers are still struggling despite companies deploying AI tools and encouraging IT to use.
“The core reason they struggle and know this from talking to their alumni is that they often lack use cases of how and why you use this AI tool in the flow of your work,” Mourshed said.
“Without a clear understanding of the use cases that make this particular task better, faster and cheaper…that’s why it leads to fatigue.
Mourshed noted that companies are investing in AI in the hopes of overnight work being done better, faster and cheaper, but they don’t provide the training or guidelines needed to enable improvements.
“It’s not a magical silver bullet. Suddenly, if you install it, you’ll do everything you want… it’s going to be a much more painful journey to reach those profits than the companies that have thought of it.”
AI expert Ajder said the right strategy is not to carefully test the use of AI and build infrastructure around it, but to rush to a race that is not ready.
“You don’t just embed it, you start with the pilot, then scoping, sandboxing, try out these systems,” he said.
This includes everything from correct training for employees to thinking about the types of AI models needed for your business. If the procedures are not in place, it is much more difficult to deal with mistakes or malfunctions.
“So I’m not saying that you can’t take thoughtful risks when it comes to using AI, but you need to calculate and scope,” Ajder said.
