All Work and No Flow : Experiences on Automating Product Owners' Workflow using Generative AI

Sammanfattning

Background. Generative Artificial Intelligence (GenAI) is rapidly transforming businesses by automating their operations. The potential of GenAI for code generation tasks has been widely explored, yet its application in non-coding activities within software development remains relatively unexplored. Product owner tasks are not necessarily code-related because they manage both the business and development sides. Automating these repetitive or time-consuming activities presents an opportunity to streamline and improve the workflow. Objectives. The goal of this study is to explore the perceived potential of GenAI to automate and improve non-coding tasks in software development. The non-coding tasks refer to the activities performed by product owners in their workflow from receiving the requirements for a product to delivering them. The study aims to identify non-coding tasks that are suitable candidates for automation based on existing evidence and explore the perceived potential of GenAI implementation along with associated limitations and challenges. Methods. A technical action research is conducted at Ericsson for this study. Semi-structured interviews and observations are conducted to identify time-consuming tasks in the process, expectations of product owners from automation of workflow, and perceptions about GenAI use. A scoping literature is carried out to identify suitable candidates for automation that are then automated in a prototype developed to streamline the workflow of product owners. The prototype is evaluated by product owners and exit-interviews and surveys are conducted to collect product owner feedback on the prototype and automated support for tasks. Results. The findings identify time-consuming tasks conducted by product owners are documentation, artefact management, communication, compiling information, and requirement analysis. The literature suggests that GenAI can be effectively used to automate articulation tasks such as documentation and requirement analysis. Product owners have a positive preference for using the prototype for a streamlined workflow and GenAI automations within it as support for workflow tasks. Conclusions. The study establishes the importance of streamlined, integrated workflows with as few tool chain distractions as possible. The prototype's development and evaluation demonstrate that streamlined processes save time and reduce context switching, while GenAI-based automations improve these streamlined processes.