The "AI Psychosis" of CEOs: When Automation Meets Reality

Aaron Levie, founder of Box, recently introduced the concept of "AI psychosis" to describe a concerning trend among business leaders. According to Levie, those making decisions about AI implementation and its ability to replace job roles are often the same individuals who least understand the true complexity and nuances of those tasks. This disconnect between strategic vision and operational reality can lead to hasty and potentially detrimental decisions.

A striking example of this "psychosis" emerged with the recent news of ClickUp's staff cuts, which reduced its workforce by 22% in favor of AI agents. This incident is part of a broader trend of tech layoffs, with 2026 numbers already approaching the totals recorded for the entirety of 2025. Such dynamics raise fundamental questions about the maturity of AI adoption strategies and the understanding of its actual capabilities and limitations.

The Gap Between Perception and Reality in LLM Deployment

The "AI psychosis" highlighted by Levie underscores a critical gap between executive expectations and the concrete technical challenges associated with deploying Large Language Models (LLM) and other AI systems. Often, the perception that AI can entirely replace complex functions overlooks the need for robust infrastructure, specialized skills for fine-tuning, and a deep understanding of technological trade-offs.

For instance, on-premise LLM deployment requires meticulous planning in terms of hardware. The availability of GPUs with sufficient VRAM, the management of throughput and latency for inference, and the ability to scale infrastructure are crucial factors that cannot be underestimated. A superficial approach can lead to ineffective investments and systems that fail to meet performance expectations or handle real workloads, making automation less efficient than anticipated.

Implications for Deployment Strategy and TCO

This "psychosis" can have significant repercussions on strategic decisions related to AI deployment, particularly for companies evaluating self-hosted versus cloud solutions. A simplistic view of AI can push towards adopting solutions that promise quick returns, without a thorough analysis of the Total Cost of Ownership (TCO). The TCO for an on-premise deployment includes not only the initial costs for purchasing hardware (such as servers with high-performance GPUs) but also operational expenses for power, cooling, maintenance, and the skilled personnel required to manage the infrastructure.

Companies prioritizing data sovereignty, regulatory compliance, or the need for air-gapped environments face additional complexities. These choices imply specific investments in local infrastructure and internal expertise, which may be perceived as obstacles by those with a less realistic view of AI's "plug-and-play" capabilities. For those evaluating on-premise deployments, AI-RADAR offers analytical frameworks on /llm-onpremise to assess these trade-offs, highlighting how a deep understanding is essential for informed decisions.

Future Outlook: The Need for an Informed Approach

To avoid the pitfalls of "AI psychosis," it is crucial for business leaders to adopt a more informed and collaborative approach. Artificial intelligence is a powerful tool, but its effective integration requires a realistic understanding of its capabilities, limitations, and infrastructure requirements. This means actively involving technical experts – CTOs, DevOps leads, infrastructure architects – in the strategic decision-making process.

Only through open dialogue between leadership and technical teams can AI strategies be developed that are not only ambitious but also technically feasible, economically sustainable, and aligned with real operational needs. A fact-based approach, carefully evaluating the trade-offs between different deployment options and considering concrete hardware specifications, is the only way to unlock AI's true potential without falling into the trap of unrealistic expectations.