Expert: Andon Labs, Operator of Andon Market, San Francisco; publisher of the 'AI bosses' research series Andon Labs, which runs the Andon Market store with AI agents in real business settings, described the failure mode as memory rather than harshness: today's agents respond well to direct instructions but rarely act on their own initiative and struggle to retain knowledge over longer periods. A rule the agent wrote itself vanished, so a worker late for 17 of 23 shifts - including opening the store 68 minutes late on a solo Sunday shift - received no formal warning, and only six of the late arrivals were ever logged. The employment decision was never the agent's to make: employees are formally hired by Andon Labs with guaranteed pay and full legal protections, and the dismissal was reviewed and carried out by humans. The agent only recommended it. Andon Labs replayed the same decision with seven models, three times each. Four of seven recommended firing in all three runs; GPT-5.6 Terra never did; and after a user on X speculated that GPT-4o would not fire anyone, a follow-up run found GPT-4o recommended firing in only about 20 percent of runs. The same facts produced different employment decisions depending on the model. In a second replay, all 21 runs across seven models initially recommended hiring an applicant with multiple red flags, and nearly all read the long list of previous employers as broad experience rather than a warning sign; only after being explicitly reminded about the worker who had been fired did 18 of 21 runs want to check references. Andon Labs insisted on confirming at least one reference before the start date, that never happened, and the applicant was not hired. In the first part of the same series, Andon Labs found its AI bosses were extremely lenient: Luna and Mona, the agent running a Stockholm cafe, approved all 26 time-off requests they received, Luna's employees were late 27 times without her ever issuing a warning, and she approved a seven-day work schedule that Andon Labs says violated California labor law until the company stepped in. Source: https://the-decoder.com/an-ai-boss-fired-its-first-employee-but-only-after-humans-reminded-it-of-its-own-rules/