Operating Systems
3M is shrinking its factory network while speeding up invention. It has closed 16 manufacturing and distribution sites this year; product-development cycles are roughly 20% shorter. Third-quarter launches rose 24% to 87. These improvement are both in-house claims, so we don’t know if one change causes the other. Still, an encouraging combination for a company whose future depends on making new things. Source: earnings call
GE Aerospace says redesigning an engine fan-assembly operation more than doubled its output. Across its maintenance network, LEAP (the engine that powers A320 NEO) turnaround times are two weeks shorter than at the end of 2025. The evidence points to more self-managing operations at the local level: teams appear to be improving workflows and removing bottlenecks without waiting for every decision to move up the hierarchy. The company still relies on centralized standards, engineering controls and management oversight, but its operating system gives frontline teams meaningful room to improve how work gets done. Source: earnings call
Why train employees who can take their skills elsewhere? A new paper finds better-managed firms train more, even within the same industry and labor market. Its proposed explanation is that the organization makes portable skills especially valuable inside that company. A practical implication for training budgets: teaching someone more is a better idea if their colleagues, tools and authority let them use it. Source: Deming, Sadun and Stanton
P&G’s Gillette factory in Berlin can run for four hours overnight without anyone operating the production line. This month, the company reiterated plans to expand unattended operations. Automation changes when people need to work, alongside how many are needed. Keeping an expensive factory busy at 2 a.m. needn’t require keeping its employees there too. Source: P&G
Artificial Intelligence
AI-agent teams tend to develop internal hierarchies. In a new preprint, researchers let agents improve themselves under different team arrangements. Successful groups often concentrated authority in their strongest member, and that agent improved more than its counterpart working alone. The result could reflect learned social patterns, optimization dynamics, or the way the experiment rewards coordination. Even artificial intelligences appear sensitive to team dynamics. Source: AGORA paper
Cooley, a law firm, has created a separate “Cooley AI” company, giving technical teams more freedom to hire and build than the law firm’s bureaucracy permits. It also offers startups software to handle routine legal tasks themselves. The bet: sacrifice some billable work now to build a better reputation with founders building the next generation. (Invoking Christiansen: protecting the client relationship may require undermining the service that started it.) Source: Business Insider
