Abstract:

The rapid evolution of Generative AI (GenAI) of the last few years – and Agentic AI lately – is reshaping how services are imagined, designed, and delivered. Designers are increasingly required to operate within hybrid human–machine contexts, developing new competencies that merge creativity, data literacy, and technological awareness. In response, the Specializing Master in Service Design by POLI.design – Politecnico di Milano and IBM Milan co-developed an experimental Gen-AI Lab aimed at exploring the potentialities and criticalities of adopting AI within the service design process.

Conducted in the first half of 2025, the Lab brought together academics, practitioners, and graduate-level service design students to explore how AI assistants can augment design methods, support creative processes, and generate measurable value for service innovation.

The Lab followed a structured, hands-on format across four appointments (Enablement, Inception, Research, and Concept & Prototyping), each mapping to specific phases of the service design process. Over 30 AI assistants, aligned to prioritized service design tools, were made available to the student groups. Participants engaged in many different prompts, experimenting with AI both as an adjunct team member and as an individualized assistant. IBM contributed methodological scaffolding, including mappings of where AI offers efficiency (speed) and efficacy (quality), and the identification of high-value use cases within the design workflow.

The Lab provided clear evidence of both opportunities and challenges in integrating GenAI into service design. AI demonstrated strong performance in data processing, rapid ideation, and divergent exploration. It accelerated tasks such as storyboarding or initial concept generation, and in some cases improved the depth and coherence of personas and early prototypes. Yet several limitations emerged: AI rarely generated critical questions unless prompted; insights risked becoming detached from their sources; unaligned prompts across team members increased divergence; and homogeneous or “mainstreamed” outputs emerged, especially when contextual data was insufficient. Furthermore, AI-generated images revealed higher levels of bias and hallucination compared to text generation. Participants nevertheless reported an expanded creative space, enhanced capacity for reflection, and increased methodological awareness through the ability to double-check “AI with AI”.

The educational experience highlights the need for a human-centered, critically informed approach that positions designers as curators, data strategists, and ethical guardians within human–non-human design ecologies. Findings directly inform the next iteration of the Lab (2026–27), with planned advancements including practitioner-based AI agents (e.g., interview experts or information architects), deeper explorations of Agentic workflows, and focused experimentation on less mature phases of the design process. The collaboration demonstrates how academia and industry can jointly shape a more mature AI-augmented service design culture, one in which human creativity and machine intelligence co-evolve toward more meaningful, responsible, and innovative service futures.

Keywords:

Generative AI, Service Design, Service Design Education, Human–AI Collaboration, AI-Augmented Creative Processes, Agentic AI and Design Methods.