Artificial intelligence capabilities are developing rapidly and companies globally are frantically trying to keep up and implement AI tools, but there are consequences to sloppy execution.
In fact, 79% of companies globally expect to incur an “AI debt” as a result of poorly implemented autonomous tools, according to a new report by Asana on the State of AI at Work which surveyed over 9,000 knowledge workers across the U.S., U.K., Australia, Germany, and Japan.
The report highlighted that companies are unprepared and lack the infrastructure and oversight required to foster a smooth collaboration between human employees and autonomous AI agents. Differing from generative AI, agents act independently, can initiate actions, and recall previous work they performed. Some examples include OpenAI’s Operator and Anthropic’s Claude.
AI debt is the cost of not implementing nascent autonomous systems correctly, Mark Hoffman, an expert at Asana’s Work Innovation Lab, told CNBC Make It.
“Those costs could be money costs. They could also be lost time, which relates to money. It could also be a lot of things that you have to undo, which is costly from a financial standpoint. It burns people out to have to do it. It’s all of the costs associated with poor implementation,” Hoffman said.
The report outlined that the debt could manifest as security risks, poor data quality, low impact AI agents which will waste time and resources for human employees, and a management skills gap.
Hoffman said this is not an exhaustive list and the “debt” could look like a bunch of code created by AI that doesn’t work right or AI-generated content that nobody is using.
New research from BetterUp Labs and Stanford Social Media Lab even found that 40% of desk workers in the U.S. have received AI-generated “workslop,” which the researchers defined as content that looks good but lacks any substance.
It’s created almost two hours of extra work for people who encountered it, a $186 invisible tax per month, and a $9 million hit to productivity in a year, per the research.
“There’s large investment going into this space right now, and ultimately it’s a question of whether those investments will pay off,” Hoffman said.
Henry Ajder, founder of AI consulting firm Latent Space Advisory, and an advisor to the U.K. government, Meta, and AI video startup Synthesia, emphasized the need for thoughtful implementation and structures.
“People who are CTOs or innovation officers, the good ones I’ve worked with, the ones who I think I did the best position to succeed with it, they aren’t sugar coating the disruption that this is going to cost … as with any kind of fundamental rework, you are going to have problems, you’re gonna have bumps in the road,” Ajder said in an interview.
‘It’s not a magical silver bullet’
Asana’s report found that despite AI adoption surging to 70% in 2025 from 52% in 2024, workers are also facing higher levels of digital burnout.
Digital exhaustion increased to 84% in 2025 from 75% the prior year, while unmanageable workloads also rose to 77%, per the report.
Mona Mourshed, founding global CEO of Generation, a U.S.-based employment organization, told CNBC that despite companies rolling out AI tools and encouraging the use of it, workers are still struggling.
“The core reason that they’re struggling, and we know this from also talking to our own alumni, is that the use case for how and why are you supposed to use this AI tool in the flow of your work is often missing,” Mourshed said.
“Without a clear understanding of what is the use case that’s going to make this particular task better, faster, cheaper … that’s what leads to the exhaustion, because you don’t know what the intended outcome is,” she added.
Mourshed noted that companies are investing in AI in the hopes that overnight work will be performed better, faster and cheaper, but they aren’t offering the necessary training or guidelines to enable improvements.
“It’s not a magical silver bullet, and all of a sudden it does everything you want once you install it … it’s going to be a much more painful journey to get to those benefits than companies that have thought it through.”
AI expert Ajder said the correct strategy is carefully testing AI use and building infrastructure around it rather than rushing into the race unprepared.
“You don’t start by just embedding, you start by piloting, you start by scoping, by sandboxing, by trialing these systems,” he said.
This includes everything from the correct training for employees, to thinking about the kind of AI models the business might need. It’s much harder to respond to mistakes or malfunctions when there’s no procedure in place.
“So I’m not saying that you can’t take risk thoughtfully when it comes to using AI, but it has to be calculated and it has to be scoped,” Ajder said.
Source: https://www.cnbc.com/