The AI flagged it. Did anyone go and look?

Korea's AI flags risky workplaces, inspectors reach one in 20

South Korea's Occupational Safety and Health Agency uses AI to score workplaces on about 200 indicators and predict which are high-risk. Figures released by lawmaker Park Jeong on October 28 show what happened next. Of 24,418 construction sites the model flagged from January to September, 63.8% were inspected. Of 373,567 flagged workplaces in manufacturing and other industries, 5.2% were. Park's question: is the field ignoring the predictions, or does the model need work?

With Qforce Pulse: a flagged site gets a checklist assigned to a named person, and the dashboard shows which flags became a finished inspection with photos and which are still open. A risk list only helps if someone works through it.

Source: asiae.co.kr

Builder's AI turns Korean safety briefings into 16 languages

Daewoo Engineering & Construction began standardizing AI translation across its Korean sites in September. A supervisor speaks instructions in Korean, and foreign workers get them as text and audio in 16 languages. Early tests put Korean speech recognition at 94.6% accuracy, the company says. A tablet handles sign-in, health checks and safety training. Its Q-Box app, where workers record quality inspections on a phone, cut documentation time by more than 90% in field trials.

With Qforce Pulse: send the morning safety briefing as a broadcast and see who has acknowledged it, and let crews log inspections on their phones with a photo, so the record exists before anyone writes it up.

Source: koreaherald.com

AI routes every hotel audit fault until someone fixes it

RedDoorz, which runs budget hotel brands across Southeast Asia, built its own audit app, Inspecto. Hotel teams check rooms, bathrooms, amenities and staff procedures without paper, and AI sorts each issue and routes it to the team that fixes it. An issue closes only once the fix is done and verified. This year its teams have run 4,900 audits across more than 1,500 properties, the company says. Guest reviews feed the same ticket queue.

With Qforce Pulse: the housekeeper photographs the fault, the AI gives a first read, and a ticket goes to maintenance with the photo attached. It stays open until someone closes it with evidence, and you see every step.

Source: manilatimes.net

Orchard managers beat AI fruit counts with local knowledge

University of California, Davis researchers spent three years with 16 fruit growers in California and Washington and ran 165 interviews. Camera systems that count fruit to forecast the harvest tended to overcount. At one Washington orchard the system predicted 420 bins; the ranch manager said 385 and was within 2%. A ripeness sensor agreed with an experienced worker 76% of the time, and when they differed, the manager almost always sided with the worker.

With Qforce Pulse: the crew logs what they see in the block, with a photo and a note next to the machine's number, so the manager's correction and the reason for it are kept for next season.

Source: agdaily.com

Italy adds scheduling software to workplace safety risk assessments

Since October 23, Italian employers must cover software that shapes how work gets done in their mandatory workplace risk assessment, with no transition period. Legislative Decree 179/2026 applies to any system that affects how work is organized, the pace of production, how tasks are done or safety decisions, whether or not the vendor calls it AI. Scheduling platforms and production monitoring count. Another article keeps hiring, discipline and dismissal decisions with a person.

With Qforce Pulse: every AI first read on a photo sits next to the person's confirmation, with who decided and when, so the risk assessment can point to a human check that actually happened.

Source: techtimes.com

One big number

89%

of hotel guests want either a hybrid stay that mixes AI and human service, or AI kept in the background on repetitive tasks. Almost none want a stay run by AI. Hotel software company Mews surveyed 3,250 travelers in eight countries in July. People beat AI at all nine touchpoints tested, even booking a taxi (44% against 28%). Put the software in the back office and the people at the front desk.

More adventures

A robot dog for gas plants. ANYbotics began delivering ANYmal X, a four-legged inspection robot certified for explosive atmospheres; its robots already run more than 120,000 autonomous inspections a month for over 50 customers, the company says.

Monitored without knowing. Only 9% of more than 2,000 Australians polled knew technology was used to monitor staff at their own workplace, as a University of Technology Sydney report calls the country's surveillance laws "dangerously out of date."

Confidence after training. Walmart surveyed its 2.1 million employees and got 1.7 million responses; 84% of its US frontline workers said they were confident using digital tools.