3.1. Research design

We adopted a qualitative case study approach to investigate the research questions (Reference YinYin, 2009). A single-case design was chosen because our “how” research questions required exploring a contemporary phenomenon (automotive CMF design for circularity with AI support) in its real-world context. The case is bounded by the CMF Circularity AI project environment – focusing on automotive Colour, Material, and Finish design practices as they relate to sustainability and AI. This design allowed us to gather rich, contextual insights from practitioners, aligning with Yin’s guidance that case studies are appropriate when the boundaries between the phenomenon and context are complex and when multiple data sources are needed to understand why or how something occurs (Reference YinYin, 2009). By treating the automotive CMF design domain (within a sustainability-driven innovation project) as a single holistic case, we could deeply examine designers’ experiences and needs without the confounding variability of multiple case settings. The case study was exploratory, aimed at uncovering user needs and tool requirements rather than testing a specific hypothesis.

3.2. Participants

We employed purposive sampling to recruit professional automotive design professionals with direct experience in CMF design and sustainable materials. A total of eight participants took part in our study. These individuals were drawn from different automotive OEMs – including major vehicle manufacturers (OEMs) and a Tier-1 supplier – to capture diverse perspectives across the design value chain (see Table 1).

Table 1.

Study participants demographics overview

Prior to the interviews, all participants provided informed consent in accordance with the Swedish Psychological Association’s standards. They were assured of anonymity (we assigned codes and generic role labels instead of real names) and informed that the data would be used for research purposes only, in line with ethical guidelines and GDPR requirements. This careful selection and consent process helped ensure participants were comfortable sharing candid insights into their workflows, challenges, and expectations.

3.3. Data collection

Our primary data source consisted of in-depth, semi-structured interviews with the eight participants. An interview guide was developed (based on the study’s aims and the literature) to ensure consistency while allowing flexibility for probing insights. Each interview lasted approximately 60–90 minutes. Most sessions were conducted via video conferencing (given participants’ locations and schedules), though a few were in person where feasible. We began each interview with a brief introduction of the research and obtained permission to record the conversation. The interviews opened with questions about the participant’s background and current role in automotive design, to establish context. We then explored their design process and challenges, focusing especially on how (and if) circular economy principles figure into their CMF decision-making (addressing RQ1). Participants described how they select materials/finishes, what sustainability criteria they consider, and any difficulties in integrating circularity (e.g., lack of data, time pressures, or perceived quality trade-offs). We used follow-up prompts to delve into specific pain points – for instance, asking for examples of conflicts between sustainable material choices and design requirements. In the second half of each interview, we shifted focus to the topic of AI design support (addressing RQ2). Participants were asked about their familiarity with AI tools and what they would expect or desire from an “agentic AI” assistant in their workflow. We discussed potential features such as real-time sustainability metrics (e.g., carbon footprint calculators, recyclability scores), alternative material suggestions, trend analyses from customer insight data, and so on. Participants were encouraged to imagine how an AI tool might help them overcome the previously mentioned challenges – and to voice any concerns (such as trust, reliability, or creativity issues). The semi-structured format enabled us to cover key themes while adapting to each expert’s domain of expertise (some conversations went deeper into design quality evaluation, others into material engineering details, depending on the person’s background). All interviews were audio-recorded to ensure accuracy of data capture. The recordings were then transcribed verbatim using a secure, AI-assisted transcription tool for efficiency (locally deployed Whisper - an automatic speech recognition (ASR) system (OpenAI, 2022). After transcription, we reviewed each transcript against the original audio recordings to correct errors and removed any identifying information (names of people, specific projects or companies) to protect confidentiality. These clean transcripts formed the basis of our qualitative dataset. We also collected secondary materials to inform the case context – for example, internal project documents and literature on automotive sustainable design – but the interview data was the central source of evidence for our case study. Throughout the data collection, we followed Yin’s guidelines of creating a case study database: organizing interview notes, audio files, transcripts, and any related documents in a structured archive. This ensured that we had a clear audit trail from raw data to the conclusions, reinforcing the study’s reliability.

3.4. Data analysis

We analyzed the interview data using qualitative thematic analysis, combined with techniques recommended for case studies (e.g. pattern-matching and explanation building). The analysis procedure was systematic and iterative. First, two researchers independently read through each transcript to become familiar with the content. We then open-coded the data, assigning descriptive labels to meaningful segments of text. Many codes were derived inductively from the participants’ own words (e.g. “limited recyclability data” or “trust in AI’s suggestions”), while some codes were informed by our research focus (e.g. categories for sustainability challenges vs. AI tool expectations corresponding to RQ1 and RQ2). Using a collaborative coding approach, the team compared and refined these codes, merging similar ones and resolving any discrepancies through discussion. Through this process, a set of salient themes emerged that captured the patterns in the data. For example, under RQ1 (current integration of circularity) we identified themes such as “Lack of real-time sustainability feedback in design”, “Perceived quality trade-offs with eco-materials”, and “Organizational pressure versus circular ideals”. For RQ2 (AI tool expectations), common themes included “Desire for lifecycle impact metrics”, “AI as a creativity support (not a replacer)”, and “Trust and transparency requirements for AI”. We noted where participants converged or diverged in their views – a form of pattern analysis aligning with Yin’s suggestion to look for rival explanations and inconsistencies. Notably, all interviewees agreed on the need for an AI tool to provide quantitative environmental impact data, yet they had differing opinions on how the AI should present recommendations (some wanted strict guidance, others just inspiration). To ensure robustness, we also utilized computer-assisted analysis: we leveraged an AI-based text analysis assistant (Manus AI, 2025) to help aggregate and synthesize the large volume of qualitative data, essentially cross-checking the themes our manual analysis found. This tool helped quickly highlight common phrases and potential insights across transcripts, which we then verified against the raw data. Throughout analysis, we maintained a chain of evidence by linking each theme back to multiple quotes and instances in the transcripts (preserving representative examples for reporting). We also performed peer debriefing within the research team – one researcher would challenge the interpretations and another would double-check that conclusions were supported by the data – to avoid subjective bias. By triangulating findings across different participants and cross-validation within the team, we enhanced the credibility of our results. In summary, the data analysis yielded a comprehensive understanding of (a) how CMF designers currently deal with circularity and where the gaps are, and (b) what features and characteristics they would value in an AI assistant to support sustainable CMF design. These findings directly inform the user-centered design requirements we discuss later.

4.1. Designer needs and current gaps

Interviews revealed that CMF designers strongly desire better data integration in their workflow, especially for sustainability metrics. Nearly all participants emphasized the need for accessible environmental impact data (e.g., CO₂ footprint, material weight, recyclability) at the design stage. Currently, obtaining such information is tedious and unreliable – one designer (P3) described it as“very difficult to get some clear data from the suppliers” when validating material choices. In practice, designers often lack precise lifecycle data and must make educated guesses about a material’s sustainability performance. We found that a core user need is a tool that can quickly provide trusted sustainability insights (e.g., carbon footprint comparisons) to support material decisions, filling a gap left by today’s ad hoc, time-consuming data collection processes.

Another significant gap in current tools is usability and integration. Participants highlighted frustration with the clunky, siloed software they currently rely on. Many still use basic spreadsheets or separate databases to track materials and impacts, which they find cumbersome. “We need something much more friendly than Excel,” one interviewee (P5) stated, echoing a common sentiment that the user interface (UI) should be intuitive. We found that designers want a seamless, low-friction interface for querying data – for example, a chat-style or voice-driven tool – rather than navigating complex enterprise systems or manual Excel sheets. Integration with existing product data systems was also frequently mentioned. Interviewees noted that current PLM/BOM systems are not designed for easy queries, forcing manual workarounds. They expressed a need for the AI tool to pull information directly from internal databases (materials libraries, supplier information, BOM data) so they can automatically retrieve details such as where a material is used or its supplier’s specifications. In short, a well-integrated, user-friendly platform was seen as essential to addressing productivity gaps in today’s workflows.

Designers currently struggle with inconsistent or opaque data, which undermines trust in digital outputs. They stressed that an AI system must use credible, up-to-date sources and clearly document its references. For instance, a studio engineering manager (P4) explained she would require “clear statements on where this AI tool gets its information, what kind of databases are feeding its answers” before trusting the results. Similarly, others said they would only treat AI suggestions as preliminary indicators unless the tool can show confidence levels or source details for its recommendations.

We found that building trust through transparency is crucial: users want to double-check the AI’s advice against known benchmarks, and they expect the tool to facilitate this (e.g., by citing an industry-standard LCA database or flagging data uncertainties). This highlights a gap in current processes, where designers often have to compile and verify data themselves; an AI assistant could fill that gap by serving as a validated information hub.

Participants also pointed to collaborative and decision-support gaps in their current workflow. Because sustainable design decisions often involve justification to other stakeholders (management, engineering, or suppliers), designers said the tool should help communicate and document the rationale behind choices. For example, one interviewee (P1) suggested the AI could generate a brief report or summary of the decision process to “show design history and justifications”, lending weight to CMF proposals. At present, assembling such evidence is manual and time-consuming, so this feature was seen as a valuable support for gaining buy-in for novel or eco-friendly materials. Likewise, several interviewees wanted easier ways to share outputs or collaborate on analyses across teams. Today, results are often shared through email or presentations. In contrast, respondents envisioned features like a shared workspace or one-click export of the AI’s analysis for colleagues.

Finally, the interviews underscored some technical limitations in current design practice that designers hope AI tools could overcome. One noted issue was the disconnect between digital visualization and physical reality: CMF professionals often find that what they see on-screen (digital renderings of colours/materials) does not match the real-life outcome, due to lighting conditions or material behaviors that are hard to simulate. This gap in material fidelity means designers still rely heavily on physical prototypes, causing delays. While the proposed AI assistant was mainly imagined as a decision-making tool (not a rendering engine), some participants suggested that future versions might integrate with visualization software or AI image generation to quickly preview material choices more realistically. In addition, knowledge dissemination is a challenge – one interviewee (P7) noted having to “educate and build up the use of CMF” in the company because many colleagues lacked familiarity with design considerations. Such comments show that, beyond specific features, designers need support in bridging knowledge gaps and ensuring consistency. In summary, current tools and processes leave CMF designers without readily available data, with clunky interfaces, and with weak support for collaboration and visualization. The interviews highlighted these as priority gaps that an AI-driven CMF assistant should address.

4.2. Attitudes towards circularity

The participants generally expressed a strong personal commitment to sustainability and circular design, but practical tensions and contradictions within the industry tempered their attitudes. A clear theme was the conflict between sustainability goals and cost or organizational priorities. Many interviewees lamented that, in practice, eco-friendly materials or circular solutions are often overridden by budget concerns. As one CMF designer said (P1), “the one that sits on the money… they care if it costs too much,” implying that management will cut sustainable ideas if they are not cost-competitive. Indeed, some OEMs still lack a formal requirement to use sustainable materials, making such efforts rely on individual leadership support rather than mandates. Conversely, a few participants highlighted positive examples where their companies are increasingly prioritizing sustainability— for instance, one interviewee noted that their firm had established a dedicated sustainability team and was aligning material choices with clear environmental targets. These mixed experiences reveal an attitudinal divide: in some organizations, circularity is becoming a strategic goal, whereas in others it remains a secondary consideration. Designers operating in the latter context often feel frustrated that “they don’t care” enough about greener materials, highlighting a need to demonstrate that circular solutions can meet business constraints.

Another nuanced attitude that emerged is how circularity is defined and balanced against other sustainability criteria. Several participants noted that “sustainability” is multifaceted and can even be internally contradictory. One common tension discussed was longevity vs. circularity: making a product last longer (durability) versus making it easy to recycle or made from recycled content. For example, an interviewee pointed out that a part could be highly recyclable but not very durable, whereas a very long-lasting material might reduce waste even if it’s not easily recyclable. The trade-off between circularity, carbon footprint, and product longevity was explicitly acknowledged – these metrics can sometimes conflict with one another. Therefore, designers approach circularity pragmatically, recognizing that it’s not just about using recycled materials but also about the overall lifecycle impact. We heard that they would like tools to help evaluate trade-offs (e.g., balancing CO₂, recyclability, and expected lifespan) to support informed decision-making. A related challenge is the lack of standardization in measuring circularity: participants gave examples of inconsistent metrics, such as one automaker treating leather as a waste byproduct (thus giving it a low footprint) while another counts the full impact of raising a cow, leading to “very, very different results” for the same material. This inconsistency in circularity definitions complicates designers’ efforts and sometimes breeds skepticism about “sustainable” claims. It underlines why many interviewees want clear, comparable data – so they can cut through contradictory standards and genuinely advance sustainability.

Participants’ attitudes also reflected an understanding of operational constraints on circular design. Many acknowledged that CMF designers alone cannot drive circularity in isolation, because key decisions (materials, part geometry, manufacturing methods) are often made upstream in engineering or product strategy. One respondent observed that a large portion of what affects sustainability “is taken earlier in the chain from the CMF team”, outside their direct control. This reality sometimes led to a sense of powerlessness or dependency: even a highly sustainability-minded CMF designer may be limited if, for example, the vehicle architecture isn’t designed for disassembly, or if procurement will not approve certain recycled polymers. We found that designers adopt a realistic attitude – they focus on what they can change (choosing better finishes or secondary materials), but they also stress that a true circular economy requires cross-functional alignment beyond the CMF domain. Notably, one of the interviewees suggested that an AI tool could aid here by questioning the design choices and prompting more radical ideas like eliminating unnecessary parts entirely. This implies that designers are open to using AI to inject circular thinking at a systemic level (e.g. “reduce the inflow of materials” by design), even though such suggestions might challenge the status quo.

Cultural and aesthetic perspectives emerged as another arena of tension regarding circularity. Automotive design has a legacy of valuing material perfection and consistency, which can clash with the ethos of reuse and natural variation. Several interviewees critiqued the industry’s “obsession with ‘perfection’” – a mindset that every surface must look pristine and identical. This perfectionism can be a barrier to circularity: for instance, using recycled materials might introduce minor colour flecks or patina, and designing for longevity might mean accepting that materials will age and weather over time. One interviewee argued for embracing “beautiful aging” – the idea that materials developing a patina or wear can be seen as adding character, not as defects. Interviewee noted that current quality standards don’t readily accommodate this philosophy, as any deviation is viewed negatively.

We found a cautiously evolving attitude: while older norms prize uniformity, some designers are now advocating for a cultural shift to appreciate natural aging and variation as part of sustainable design. This insight suggests that successful circular design may require not just new tools, but also changes in taste and acceptance within design teams and consumers.

Despite the above challenges, the overall stance towards circularity was one of guarded optimism. Many participants conveyed that both the industry and their own teams are slowly shifting toward sustainability as a core value. One respondent noted a “cultural shift in design thinking” with new leadership that “understands the importance of starting with materials,” signalling organizational readiness for change. Designers take pride in their ability to be creative within constraints, as one interviewee highlighted being “impressed with our work” even when budgets and tools are limited. Such comments reflect a resilience and willingness to adapt: CMF teams are looking for ways to innovate on sustainability, and they appear eager to leverage any new tools (like AI) that can help them do so without sacrificing design integrity. In summary, attitudes towards circularity among CMF professionals are enthusiastic but pragmatic – they believe in its importance, are aware of the internal contradictions and hurdles, and hope that better data and organizational support (potentially facilitated by AI) will let them resolve those tensions in practice.