Generative AI, Stakeholder Interaction, and Iterative Product Innovation: Evidence from Smart Product Development in Manufacturing Firms

Description of Database Materials and Coding Procedures The database contains six types of data sources, which correspond to the data types listed in Table 2, “Data Sources.” Semi-structured interviews. To protect interviewees’ privacy, the database contains only the documents generated from transcribed interview recordings and interview notes, rather than the original audio recordings. Each group interview is represented by one transcript, whereas each individual interview is represented by a separate transcript. The database includes transcripts and notes from three rounds of interviews. These interview records correspond one-to-one with the interviewees listed in Table 3, “Sample Profile of All Interviewees.” All interview records are translated versions of the original Chinese materials. Haier corporate information. This category includes books, news reports, and presentation slides. The book COSMOPlat: Haier Industrial Internet Platform was translated using an AI translation tool. News reports were collected from Haier’s official website, webpages, WeChat public accounts, and other publicly available channels. All news reports are translated versions of the original Chinese materials. The presentation slides were also translated using an AI translation tool on a page-by-page basis. Network information: Publicly available interview videos. This category includes interview audio recordings and transcripts involving Haier executives that were downloaded from publicly available online platforms. All transcripts are translated versions of the original Chinese materials. Journal information. This category includes academic articles downloaded from online sources. Some articles were originally published in English, whereas others are translated versions of Chinese-language articles. Financial reports. This category includes Haier’s annual financial reports from 2012 to 2025. The reports from 2012 to 2014 are translated versions of the original Chinese reports, whereas the reports from 2015 to 2025 are the original English-language reports. Observations. This category includes photographs taken during field visits to Haier experience centers, together with descriptions prepared for each photograph. The research team conducted open coding of the above materials based on the original Chinese-language materials, generating 3,307 coded text extracts and 309 initial concepts, as documented in Coding Results_Coded_Extracts. The coded text was subsequently translated into English for presentation in the database. Therefore, the English coded text may contain minor differences in wording or nuance from the original source materials. These translations were used to facilitate presentation and transparency of the coding process, whereas the coding itself was conducted based on the original-language materials. Through axial coding, the initial concepts were further consolidated into 11 core concepts, with the results presented in Coding Results_Concept_Clustering. The detailed coding principles are described in the Data Analysis section of the manuscript.