Opinion: The Cognitive Void: How AI Psychosis is Erasing Corporate Judgment

The current corporate rush toward artificial intelligence is not a strategic evolution; it is, as Nikhil Suresh describes in a piece for Hermit Tech, a state of "mass psychosis." From the halls of Fortune 500 companies to the corridors of government institutions, banks, and hospitals, a dangerous pattern has emerged: the outsourcing of critical cognitive infrastructure to tools that frequently fail, overseen by leadership that is either incapable of formulating a plan or too terrified to admit the failure.

As a skeptic, I find the most alarming aspect of this trend not the technical limitations of Large Language Models (LLMs), but the erosion of the human judgment required to manage them. We are witnessing a systemic risk where the market is ignoring a fundamental truth: the people in charge are no longer making rational decisions. Instead, they are operating in a feedback loop of "frothing excitement" and obfuscation.

**The Illusion of Productivity**

There is a widening chasm between the narrative sold to shareholders and the reality on the ground. According to reporting from Hermit Tech, many publicly traded companies are issuing press releases claiming massive productivity gains from AI. In reality, some of these organizations have done nothing more than purchase Microsoft Copilot licenses and declared victory.

This is not merely a case of optimistic marketing; it is a coordinated effort to hide failure. Suresh notes that boards, executives, consultants, vendors, and employees all have a vested interest in misrepresenting the success rate of AI projects. The cost of honesty is high: executives who admit the insanity of these pivots are removed, and honest employees are frequently fired or targeted during layoffs.

**A 0% Success Rate**

When we strip away the corporate jargon, the actual output of these investments is staggering in its inadequacy. Suresh reports that his team has observed a 0% success rate across all AI projects encountered over the last year and a half. This failure persists not only in projects they were directly involved in but also in those observed in passing.

Even when AI tools accelerate specific, narrow workloads, the scale of the investment remains senseless. The failure is twofold. First, many companies are fundamentally incapable of running software projects effectively, and AI projects inherit all the standard failure modes of software development while adding the volatility of a novel technology. Second, there is a hard ceiling on what LLMs can actually accomplish.

Consider the ubiquitous internal chatbot. These tools are designed to streamline information retrieval, yet they fail because they rely on documentation that is often low-quality. As Suresh points out, an LLM is not psychic; it cannot provide answers that have not been written down and made accessible. Consequently, internal uptake is virtually non-existent.

**The Metrics of Failure**

Perhaps the most systemic risk is the deliberate avoidance of accountability. Project leaders are carefully avoiding the tracking of basic metrics—such as whether a tool is actually being used—or are relying on metrics that are easily gamed.

Suresh provides a poignant example involving Mitsubishi. After an automotive failure, he interacted with a customer-facing AI bot that was polished, natural-sounding, and promised a swift callback. That callback never happened. In the eyes of the company's metrics, the interaction likely appeared successful because it did not trigger an error. However, the real-world result was the loss of a customer who decided not to purchase another Mitsubishi vehicle. The bot functioned perfectly as a piece of software, but it failed as a business tool because it severed the human connection necessary for resolution.

**The Systemic Risk**

We are currently pivoting entire organizations around "agentic workflows"—a space Suresh describes as having only about four viable applications—despite a lack of evidence that these tools provide any meaningful value. The result is a corporate landscape where the "saner people" are living in a state of constant fear and frustration, unable to have rational conversations with leadership who have succumbed to AI psychosis.

When we outsource the core of our decision-making processes to black-box algorithms, we aren't just risking a few failed software projects. We are eroding the institutional capacity for critical thought. If the primary goal of a project is to look successful on a dashboard rather than to solve a human problem, the company has already failed. The market may be pricing in the potential of AI, but it is completely ignoring the cost of the cognitive decay happening in the C-suite.