"If we misunderstand the client, we misspend 100 hours": Exploring conversational AI and response types for information elicitation
Authors
Paludan, Daniel Hove ; Bährentz, Kasper Vedsted ; Fredsgård, Julie
Term
4. Term
Publication year
2025
Abstract
A shared understanding between clients and designers is vital for successful projects, yet we know little about how digital tools shape this alignment. This thesis explores how digital systems can support requirements elicitation—the early process of identifying a client’s needs and preferences. We designed and tested a digital tool that combines a conversational agent (chatbot) and choice-based response formats (e.g., multiple choice) to see how these features affect early collaboration. The study had three phases: (1) Semi-structured interviews with 10 design companies to understand current practices and inform the tool’s design. (2) A controlled evaluation with 50 mock clients using a 2×2 factorial design to isolate the effects of the conversational agent and response format on user experience and perceived preparedness for collaboration. (3) A presentation of the system to 7 of the original 10 companies to gather feedback on its value, limitations, and how it could fit into practice. Findings show that both the conversational AI and choice-based responses were associated with lower ratings on the dependability scale of the User Experience Questionnaire, yet they produced client input with greater clarity. The thesis offers design implications for integrating conversational AI and structured responses into elicitation tools to support mutual understanding in early-stage client–designer collaboration.
En fælles forståelse mellem kunde og designer er afgørende for et godt designforløb, men vi ved stadig lidt om, hvordan digitale værktøjer påvirker denne forståelse. Dette speciale undersøger, hvordan digitale systemer kan støtte kravindsamling – altså det tidlige arbejde med at indkredse kundens behov og ønsker. Vi udviklede og testede et digitalt værktøj, der kombinerer en samtaleagent (chatbot) og valgbaserede svarformater (f.eks. multiple choice), for at se, hvordan disse funktioner påvirker samarbejdet tidligt i processen. Studiet havde tre faser: (1) Semistrukturerede interviews med 10 designvirksomheder om deres nuværende praksis, som informerede værktøjets design. (2) En kontrolleret evaluering med 50 simulerede kunder i et 2×2 faktorialt design, der isolerede effekterne af henholdsvis samtaleagent og svarformat på brugeroplevelse og oplevet forberedelse til samarbejde. (3) En præsentation af systemet for 7 af de oprindelige 10 virksomheder for at indsamle deres vurderinger af værdi, begrænsninger og muligheder for at integrere værktøjet i praksis. Resultaterne viser, at både samtaleagenten og valgbaserede svar var forbundet med lavere scorer på "dependability"-dimensionen i User Experience Questionnaire, men samtidig gav de kundebidrag med større klarhed. Specialet tilbyder designanbefalinger for, hvordan samtaleagenter og strukturerede svar kan indarbejdes i kravindsamlingsværktøjer for at styrke den gensidige forståelse mellem kunder og designere i projektets tidlige faser.
[This apstract has been rewritten with the help of AI based on the project's original abstract]