This working paper examines how generative AI tools can be integrated into service design and broader co-creation workshops while keeping participation, psychological safety, and collective ownership central. Drawing on ten qualitative interviews with service and social designers, an exploratory questionnaire, and self-initiated facilitation trials, the study identifies where AI adds practical value and where its use introduces risks. Findings show a consistent pattern: AI is most effective in the preparation and follow-up phases of co-creation sessions—supporting research, agenda design, synthesis, and documentation—while its role during live facilitation should remain limited to avoid disrupting rhythm, attention, and participants’ sense of ownership. To structure these insights, the paper introduces the Workshop Compass framework, which distinguishes three phases (preparation, execution, follow-up) and three layers of facilitation attention (environment, people, process). Across these dimensions, recurring benefits include efficiency gains and scaffolding for novice facilitators, while key risks involve over-generalisation, automation bias, and undue deference to machine-generated outputs. The paper proposes practical heuristics to mitigate these risks, such as separating divergent and convergent phases, disclosing AI contributions, treating machine-generated clusters as suggestions, and pairing decisions with short human rationales. The contribution is primarily methodological and practice-oriented. Rather than evaluating AI systems, the paper offers grounded guidance for designers and facilitators seeking to integrate AI responsibly into co-creation workshops, while preserving the fundamentally human and relational nature of facilitation. A revised version is planned for submission to a peer-reviewed venue.
Máté Attila Barna (Sat,) studied this question.