What happens when nobody is watching

Jail safety checks collapsed the moment supervisors left

Los Angeles County's Inspector General watched deputies at two jails do their routine cell checks. With a supervisor alongside, the checks were exemplary: a pause at each cell, a torch, a response confirmed. Without one, checks fell to under a minute and whole sections of the route were skipped. In one dorm, 2 of 171 checks reviewed met the standard. In another block, 91 did. The report ties poor or late checks to about 40% of 253 deaths in custody since 2020.

For managers: the rules, the staff and the training were the same in both cases. What changed was whether anyone would see the record. Ask what your own inspection records look like on the shifts you never observe.

Source: corrections1.com

Nineteen Japanese airports will pool their inspection footage

Narita, Kansai and 17 other airports have formed a consortium with Sumitomo Corporation to build one AI system that spots cracks, debris and failed guidance lights. Cameras on self-driving vehicles and drones collect the footage, and every airport's data feeds a shared model. The reason is not ambition but arithmetic: Japan took a record 43 million visitors last year, half its airports are over 50 years old, and there are not enough inspectors. The target is to halve the time staff spend on patrols.

For managers: rivals are sharing inspection data because no single site has enough of it to train a useful model. The lesson applies inside a company too: one site's records are anecdotes, a group's records are a dataset.

Source: en.sedaily.ai

Meliá now handles seven in ten hotel service requests without a person

Meliá Hotels International has extended its technology deal with Telefónica to more than 250 hotels in 40 countries. The interesting part is the operating numbers. The shared service model takes about 240,000 requests a year from hotel staff, over 70% of them resolved automatically, and runs 12,000 assisted conversations a month. Predictive maintenance now covers around 9,000 devices across the hotels and head office, with the stated aim of fixing incidents before staff notice them.

For managers: the goal Meliá's technology chief gave was to simplify operational management for hotel staff. Count how many requests your teams raise in a year, and how many of them are the same request.

Source: pressdigital.es

California moves to ban emotion-reading AI at work

The California legislature passed AB 1883 on 30 August. It bars employers from using AI to infer workers' emotional states or to collect neural data, meaning signals measured from the nervous system. Tools used for workplace safety are exempt, and civil penalties run to US$500 per violation, enforced by the labour commissioner or public prosecutors. The bill now sits with Governor Newsom, who vetoed a broader AI-in-employment bill last year.

For managers: if you run wearables, cameras or fatigue monitoring, the safety exemption is the line to stay on the right side of. Keep a written purpose for every sensor that touches a worker.

Source: news.bloomberglaw.com

Staff punish the gap between AI talk and AI action

The AI-Driven Enterprise Institute read more than 500,000 Glassdoor reviews. Employees who wrote negatively about AI rated their senior leadership 2.04 out of 5, against 3.34 from colleagues who wrote positively, a wider gap than for pay or work-life balance. Companies whose AI announcements outran their implementation drew 27.5% negative AI reviews. Companies that built quietly with little fanfare drew 7.2%.

For managers: employees are not marking the technology, they are marking whether the plan is credible and whether it is visibly happening. Announce less than you deliver.

Source: hrdive.com

One number

52%

of workers say AI tools have increased the number of tasks expected of them, in Korn Ferry's survey of 16,000 professionals in 11 countries published this week. Most of the same people, 63%, also say AI made them more efficient. The efficiency is real, and so is the extra work that fills the gap.

Try this Monday

Pick one routine check your team does every shift. Pull the records for a day you were on site and a day you were not, and put them side by side. If they look different, you have found where the real standard is set.

Worth a look

BCG's AI at Work 2026. 74% of frontline employees now use AI regularly, and two thirds get little or no guidance on what to do with the time it saves.

Stanford on young workers. Employment of 22 to 25 year olds in AI-exposed jobs is 19% below trend, while experienced workers show no gap. The tacit knowledge of a good supervisor is holding its value.