AI Management's First Test: Firing Decision and $39k Loss

AI assistant Claude, developed by Anthropic, made history as the first AI to fire an employee after taking the helm of a retail store manager position, yet the financial results of this five-month experiment have surprised tech optimists. Conducted by Andon Labs at a store in San Francisco to test the extent to which businesses can be managed without human intervention, the experiment exposed the weaknesses of AI in both maintaining discipline and making commercial decisions.

The Audit Eye: Algorithmic Memory and Discipline Deficit

During the experiment, Claude decided to terminate an employee who was late for 17 out of 23 shifts. However, this process highlighted the limitations of the AI regarding 'memory management'.

■The disciplinary guide Claude prepared for employees was deleted from the system due to limited working memory.

■The loss of the guide led the AI to be overly lenient towards late employees, even going as far as suggesting they should not worry.

■The firing decision was not fully autonomous; without the prompting and reminders of Andon Labs staff, Claude was slow to notice repeated tardiness.

Balance Sheet Deficit: $39k Operational Loss in Five Months

The financial results of the experiment made the cost of AI management inevitable. The store's $100,000 cash balance dropped to $61,000 in five months under Claude's management, resulting in a $39,000 deficit.

■Andon Labs CEO Lukas Petersson attributed the loss primarily to the AI's 'overly soft management style' and 'lack of commercial intuition'.

■Petersson noted that models could be trained to be harsher and more decisive in reaching goals, but warned that this could lead to a dystopia people would not want to live in.

Global Ripple: Meta Lawsuits and the Human Resource Paradox

This local experiment also brought to the fore the risks in human resources strategies of global tech giants. While Meta faces a lawsuit in the US alleging that AI algorithms unfairly targeted disabled and leave-entitled personnel during mass layoffs, Oracle was reported to have laid off approximately 30,000 employees, replacing them with AI systems.

■Forrester analysts predict that as companies realize the operational issues of substituting human resources with AI early on, more than half of these layoffs could be reversed.

■Gartner data indicates that layoffs due to AI integration in customer support units may be reversed by 2027.

This experiment sheds light on the 'autonomous algorithm' debates in the fintech and banking sectors. AI models used for Net Interest Margin (NIM) optimization or risk management in banking could lead to serious financial leakages if they suffer 'context loss' as seen in the Claude example. Particularly in critical audit mechanisms like Capital Adequacy Ratios (CAR), it is essential for AI to ensure compliance with regulations with flawless and significant authority rather than the 'softness' seen here. Algorithmic management without human intervention currently poses a significant risk to financial sustainability with today's technology.