Personagram: Bridging Personas and Product Design for Creative Ideation with Multimodal LLMs

TL;DR

Personagram addresses the challenge of turning static personas into actionable design inputs by using a multimodal LLM pipeline that maps census-based persona attributes to product references and extracted design features. The system couples a Canvas UI for exploration with a Production UI for prototyping, grounded in an AI workflow that moves from text personas to images and back to text prompts, enabling structured, iterative ideation. In a within-subject study with 12 professional designers, Personagram yielded higher engagement with persona-driven outputs, greater perceived transparency and reliability, and faster, more satisfactory ideation cycles than a chat-based baseline. The results highlight the value of multimodal, scaffolded workflows for integrating AI-generated personas into professional design practice, while also acknowledging limitations and offering directions for future collaborative and longitudinal research.

Abstract

Product designers often begin their design process with handcrafted personas. While personas are intended to ground design decisions in consumer preferences, they often fall short in practice by remaining abstract, expensive to produce, and difficult to translate into actionable design features. As a result, personas risk serving as static reference points rather than tools that actively shape design outcomes. To address these challenges, we built Personagram, an interactive system powered by multimodal large language models (MLLMs) that helps designers explore detailed census-based personas, extract product features inferred from persona attributes, and recombine them for specific customer segments. In a study with 12 professional designers, we show that Personagram facilitates more actionable ideation workflows by structuring multimodal thinking from persona attributes to product design features, achieving higher engagement with personas, perceived transparency, and satisfaction compared to a chat-based baseline. We discuss implications of integrating AI-generated personas into product design workflows.

Figures (9)

• Figure 1: Personagram Production UI Components. Features identified from the Canvas UI can be modified in the H) Dimensional Scaffolding Interface, with changes to the feature list being reflected on the I) Text-to-image Prompt Input Box in real time. Pressing the "Explore Design Possibilities" button triggers image generation and results are displayed on the J) Image Gallery.
• Figure 2: Iterative design process using Personagram. Upon filtering persona attributes, users can proceed to the ideation process by 1) selecting a persona, 2) generating product reference images, 3) extracting design features from product image, 4) identifying and moving features to the production UI (i.e., prompt interface), 5) refine feature list, 6) generate product images, iterate by remaining on the production UI or 7) returning to the canvas UI
• Figure 3: Baseline system. Users can interact with an LLM through a chat interface akin to commerical LLM products.
• Figure 4: Study procedure flow.
• Figure 5: Example images of outcomes generated by participants using Personagram (two helmets for older adults, and two chairs for children). The figure also shows the corresponding features participants selected for incorporation into prototyping, extracted from persona-relevant products in Personagram.