In the modern digital age, companies need to be able to quickly explore the process innovation affordances of digital technologies. This includes exploration of Machine Learning (ML), which when embedded in processes can augment or automate decisions. BPM research suggests using lightweight IT (Bygstad, Journal of Information Technology, 32(2), 180–193 2017) for digital process innovation, but existing research provides conflicting views on whether ML is lightweight or heavyweight. We therefore address the research question “How can Lightweight IT contribute to explorative BPM for embedded ML?” by analyzing four action cases from a large Danish manufacturer. We contribute to explorative BPM by showing that lightweight ML considerably speeds up opportunity assessment and technical implementation in the exploration process thus reducing process innovation latency. We furthermore show that succesful lightweight ML requires the presence of two enabling factors: 1) loose coupling of the IT infrastructure, and 2) extensive use of building blocks to reduce custom development.

在现代数字时代，企业需要能够快速探索数字技术的流程创新可供性。这包括对机器学习 (ML) 的探索，将其嵌入流程中可以增强或自动化决策。 BPM 研究建议使用轻量级 IT（Bygstad，Journal of Information Technology，32 (2), 180–193 2017）进行数字流程创新，但现有研究对于 ML 是轻量级还是重量级提出了相互矛盾的观点。因此，我们解决了研究问题“轻量级 IT 如何为嵌入式 ML 的探索性 BPM 做出贡献？”通过分析丹麦一家大型制造商的四个诉讼案例。我们通过证明轻量级 ML 大大加快了探索过程中的机会评估和技术实施，从而减少了流程创新延迟，为探索性 BPM 做出了贡献。我们进一步表明，成功的轻量级 ML 需要存在两个促成因素：1）IT 基础设施的松散耦合，2）广泛使用构建块来减少定制开发。