Box CEO: Executives Must Avoid ‘AI Psychosis’

Box CEO Aaron Levie says C-suite executives are facing a form of “AI psychosis” as they look to integrate AI technology into their businesses.

In a post on X, Aaron says CEOs are “uniquely” prone to this phenomenon as they are “sufficiently distant from the last mile of work” required to turn AI outputs into reliable business tools.

“When [CEOs] play with AI, they see the happy path results, often not considering the next 10 or 20 things that have to happen to get sustainable results from agents,” he says.

“Look, I made this awesome product prototype,” Levie adds, before saying that executives often overlook the need to review the code before it goes into production, resulting in the potential need to fix problems later on.

Aaron’s theory is backed up by data found throughout the industry, with 42% of companies abandoning most AI initiatives in 2025, up from 17% the prior year, according to S&P Global Market Intelligence.

This market data shows that the average organisation scrapped nearly half AI proofs-of-concepts before they reached the production stage.

CEO oversight of AI integration

Despite some industry uncertainty over deploying these systems, MarketsandMarkets states that agentic AI companies have raised more than US$2bn across more than 40 rounds in the first quarter of 2026, compared to US$1.09bn in the same period a year earlier.

The global AI agents market was valued at US$7.84bn in 2025 and is projected to reach US$52.62bn by 2030.

Aaron suggests companies that can prove their AI technology is capable of consistent task completion at scale, audit trails and governance infrastructure, are more likely to receive larger funding rounds as they’ve shown they can solve “the last mile”, as he puts it.

Using the analogy of an AI-generated contract, Aaron says just because you can produce one, it doesn't mean you understand the process that made it, nor are you safe from potential errors.

“Look I generated a contract,” Aaron says. “Yes but you didn’t verify all the terms before it goes out to the counterparty and didn't have to wire up all the past contracts to work with.”

CEOs are rarely privy to the “last mile” of technology and process deployment and often only experience AI in controlled demonstrations, not during the production phase.

Research from the RAND Corporation highlights this, finding that more than 80% of AI projects fail, twice the failure rate of non-AI technology projects.

A 2025 survey from McKinsey also found that organisations reporting meaningful financial returns were twice as likely to have redesigned end-to-end workflows before selecting any modelling technique.

Implications of agentic AI in business

At a TechCrunch Disrupt in 2025, Aaron argued that mission-critical processes need a “church and state” separation between deterministic systems and probabilistic AI, citing real-world examples of agent failures, such as data leaks and unexpected database modifications.

Major errors like these could lead to dissolutions of entire AI projects within the industry. According to data from Gartner, more than 40% of agentic AI initiatives are predicted to be cancelled by 2027.

Aaron suggests that despite these concerns, CEOs should use AI more, but the method in which it’s used is important.

He discusses how generating outputs is not equal to understanding systems, adding: “The best thing you can do as a CEO is to use AI a ton to figure out the real implications of agents in the enterprise, and come out the other side with an appreciation for both the upside and the real work that goes into them.”