The software made the call. Can anyone check it?

AI scheduler puts nurses on shifts they asked to avoid

HCA Healthcare, the largest hospital chain in the United States, has rolled out Timpani, an AI scheduling tool built with Palantir, at about 130 of its 190 hospitals. It builds nurse rosters from forecasts of patient volume. Six nurses told Wired it ignores requested days off and leaves shifts short of senior staff. HCA says nursing leaders make the final decisions. A former HCA data manager is suing, alleging the tool's data was deleted, so nobody could audit it.

For managers: if software drafts the roster, keep a record of what it proposed, what a manager changed and why, because that record is your only answer when staff say a shift was unsafe.

Source: wired.com

AI cameras warn port drivers when workers share their lane

Busan New Container Terminal in South Korea has installed 28 AI cameras on 14 ship-to-shore cranes and satellite positioning on about 70 vehicles. The system calculates the distance between every worker and every straddle carrier, the tall machines that move containers. Drivers get alerts at 15, 10 and 5 meters, but only when the worker is in the same lane, to cut false alarms. The control room can replay any alert to investigate near misses.

For managers: the replay is the part to copy, because an alarm that leaves a record turns a near miss into something you can learn from instead of a story nobody wrote down.

Source: rivieramm.com

Ticket data cuts a ballpark's cleaning water by 198,000 gallons

ABM, the facilities services company, now plans cleaning at Chase Field, home of the Arizona Diamondbacks baseball team, from the ticket manifest: who sat where, and how many. Crews skip washing sections nobody used. In one month that saved 198,000 gallons of water and about $2,500 compared with cleaning everything, the company says. With coreless paper rolls and automatic restock alerts from restroom sensors, paper waste fell to 8%, against a typical 15% to 20%.

For managers: ABM's own vice president says the data still needs "field verification," so if software tells a crew where to skip, someone has to confirm the skipped areas were actually fine.

Source: facilitiesdive.com

AI scores every call, and agents cannot see the rules

University researchers, writing in Social Europe, surveyed more than 700 workers in British telecom and finance call centers in November 2025. They found 81.3% are recorded on every call, usually with AI sentiment analysis. Supervisors used to listen to a sample; software now scores everything. Workers report unexplained score drops, and union representatives say these scores can move up to 40% of variable pay. Where a manager can correct the AI's score, workers feel less negative about it.

For managers: give every person a simple way to challenge a software score, and a named manager who can overrule it, before the score touches pay.

Source: socialeurope.eu

One number

3.4 hours

a week is what employees in the United Arab Emirates spend clarifying, correcting or rewriting poor AI output, against 3.6 hours a week they say AI saves them. Censuswide surveyed 500 full-time AI users at UAE companies with 1,000 or more staff for Workday, from June 23 to July 2. The two are separate averages, so they do not give anyone's net gain. They do show where much of the saved time goes: into checking.

Try this Monday

Take one thing software produced for your sites last week: a roster, a cleaning plan, an alert or a score. Ask two people on the floor whether it matched what happened. Where it did not, find out whether anyone could change it and whether that change was written down. If nobody could, decide this week who can, and where they record it.

Worth a look

Sensors instead of night rounds. South Korea's health ministry plans smart diapers and fall sensors to replace two-hourly night checks in care homes, piloted in 400 households and 12 facilities from 2027.

Ordering from the shop floor. Distribution Casino France says handheld computers with AI substitution suggestions cut order time by 20% to 30% across 4,500 convenience stores, in a vendor announcement by Zebra Technologies.

Into the drawer. Why store staff stop using a tool that confidently sends them to the wrong aisle.