Tim Cook visits Shanghai's Blue Box as brand runs local AI on Mac mini

Blue Box, founded in 2019, sells roll-packed mattresses online and offers a 100-day trial. The company uses iPad and Mac to connect customer service, product design, content production and offline service. For custom mattress orders, customer service passes a request to designers, who use a Mac to design the mattress appearance and pattern, confirm fabrics and generate production instructions for the factory. Its content team uses Macs for multi-camera monitoring and editing of unpacking and rebound demonstrations, and iCloud sync lets customer service, brand and e-commerce teams share visual materials.

The company has grown from a two-person founding team to more than 300 employees, changing its device delivery and permission management needs. With most staff using Macs and other Apple devices, Blue Box moved its internal device management to Apple Business, which launched in April and combines built-in mobile device management, managed Apple accounts, app distribution and directory services. IT uses preset groups to preconfigure apps, system settings and security policies for different roles. When an unopened Mac reaches a new employee, connecting to the internet triggers zero-touch deployment. Founder Lu Zhaoli told Cook it takes about 12 minutes from unboxing the Mac to starting work, and employee satisfaction among those who switched to Mac reached 100%.

Blue Box's sleep convenience vehicle service lets users book online and have a mattress-equipped experience vehicle come to their residential compound or office building. Inside, iPad serves as a mobile workstation for staff, showing product information, syncing operating data, connecting to backend systems and linking offline trials with online orders.

As online and offline services and data from different departments converge, Blue Box says it deployed a 35B-parameter Qwen model locally on a Mac mini with 64GB of memory and developed an internal agent interface. Employees can ask questions in natural language; the system converts them into SQL to query internal data including sales, inventory and store operations. Lu told Cook the system had been in use for about a month, with customer service, product, e-commerce and supply-chain departments already using it. Blue Box said that in the past, compiling sales across channels, comparing store performance or forecasting next-quarter sales required multiple departments and could take one to two days; some queries now return results in minutes.

An Apple-commissioned survey conducted by Omdia found that 76% of respondents were concerned about data leakage through cloud services, and more than one-third of enterprises planned to move more AI workloads to local devices in the next 12 months. The survey was conducted in the fourth quarter of 2025 and covered 1,584 enterprise technology decision-makers and practitioners in five countries; all respondents' companies had 500 or more employees and were concentrated in healthcare, finance and professional services. At the same time, 73% of enterprises still used cloud or hybrid infrastructure to deploy AI, indicating local computing was not replacing the cloud entirely but being used alongside it based on data sensitivity, cost and workloads. Blue Box's approach is close to the survey's findings: it assigns queries involving sales, inventory and store operating data to local AI on the Mac mini to reduce sensitive data transmission to the cloud and make ongoing costs easier to estimate, while other materials requiring cross-department collaboration and on-demand access continue to flow through services such as iCloud.