Box CEO Aaron Levie Warns CEOs Are Prone to 'AI Psychosis'
Box founder and CEO Aaron Levie stated on X that CEOs are particularly susceptible to AI psychosis because they are distanced from practical implementation work and cannot see the final steps needed to generate sustainable AI value.
He pointed out that CEOs often only see the perfect prototypes generated by AI (such as product demos or contracts) but overlook the substantial real work that follows, including code review, problem fixing, validating terms, and system integration.
Levie suggested that CEOs should extensively use AI to truly understand the challenges of deploying agents in enterprises, recognizing both the potential and the actual efforts required.
Source: Public Information
ABAB AI Insight
Aaron Levie, as a veteran in SaaS, has long emphasized the practicality of AI tools in serving enterprise-level documentation and collaboration. His criticism of 'AI psychosis' continues his long-standing observation of the gap between technological hype and corporate reality, warning against the general optimism bias among top decision-makers.
On the capital front, corporate CEOs are driving rapid expansion of AI budgets, but actual engineering teams face code review, integration, and maintenance costs that far exceed expectations, leading to delays in ROI validation. Resources are shifting from purely prototyping experiments to tools and processes that address the 'last mile' issues.
Similar to early cloud computing adoption where CEOs focused on demos while overlooking integration complexities, and the long-term maintenance cases of CRM systems, AI agents are currently transitioning from laboratory/prototype enthusiasm to sustainable enterprise-level deployment.
Essentially, this is a technological substitution: AI agents superficially replace human tasks efficiently, but actual deployment requires significant human oversight and system restructuring. The mechanism is that the complexity of existing corporate processes, compliance, and legacy systems far exceeds the capabilities of the models, forcing capital to shift from blind optimism to a rational reassessment of the true costs of 'human-machine collaboration'.
ABAB News · Cognitive Law
CEOs see prototypes, engineers fix production; the gap determines AI success or failure.
After the perfect demo, the real 100 troubles begin.
The further away from the last step, the easier it is to overestimate the actual value of AI.