Towards AI-Enabled Assistant Design Through Grassroots Modeling: Insights from a Practical Use Case in the Industrial Sector

Process modeling is used to understand a business process and to document requirements, but is mostly formalized and limited to modeling experts. This can be a problem when designing interactive systems that incorporate AI elements, since the design can fail to take into account tacit knowledge and context-specific requirements of people that execute the process. While recent discourse has highlighted this gap and called for an exploration into light-weight, grassroots modeling techniques that can be used to model everyday work, it is still unclear how these can be harnessed to design information systems that support work. The aim of this paper is to showcase how a triangulated approach combining three different perspectives - grassroots modeling, theoretical grounding, and first person media, can be used to collaboratively model an informal work activity and design an AI-enabled system to instruct novices to perform that activity. Our experience confirms the assertion that, when provided with the necessary scaffolding, experts without any formal modeling experience can be supported to model their specific, local activities and, in doing so, contribute valuable knowledge to the design of information systems.