Leveraging the vast interconnection of language and ideas through Large Language Models, a designer’s understanding of the needs, wants and desires of intended stakeholders defines the value proposition and product design requirements of a product or service through implementation of the Value Opportunity Analysis (VOA). The resulting VOA LLM Bot explores emotion, aesthetic and other human-valued attributes, and significantly increases perception of the VOA as a useful method for identifying product requirements, and analyzing opportunity solutions.

We introduce a method that turns online customer reviews into design insights. By analysing smartphone reviews, we extract the product features customers talk about and identify the sentiment linked to them. The approach combines Large Language Models (LLMs) with the Kano model, showing how specific features influence satisfaction or dissatisfaction. The results are coherent with the dimensions of the Kano model. The work demonstrates that LLMs can be informed and constrained by established design frameworks, bridging LMMs and design reasoning to provide theory-grounded insights.

Data-Driven Design (DDD) is emerging as a transformative approach in engineering design, leveraging AI tools to extract knowledge from design data that drive product development and innovation. While large language models have advanced DDD through the analysis of textual data, technical drawings remain largely unexplored. To address the limitations of current vision-language models, this study presents a novel object detection pipeline that automatically identifies components in patent images, enabling data-driven analysis of component geometries, interfaces, and spatial configurations.

Dominant designs establish de facto standards for all products within an industry, shaping both competition and innovation dynamics. Studying dominant designs enables firms to make informed decisions for new product development and to anticipate technological shifts. This paper presents a computer-based method that automatically extracts the spatial configuration of components from patent drawings to support the analysis of dominant designs and anomaly detection. A case study on eyeglasses validates the approach, demonstrating its potential for data-driven design innovation.

Chatbot-based surveys offer low-burden, in-situ data collection, yet unconstrained LLMs often drift from research aims. We conducted 359 ultra-short, post-experience voice interviews in a public venue to compare a framework-guided LLM, an unconstrained LLM, and fixed questions. The guided approach produced significantly longer responses than fixed questions and yielded the richest diversity of process-specific accounts. These findings show that probe control is essential for eliciting actionable, experience-grounded feedback in real-world, time-limited settings.

Artificial intelligence influences requirements engineering, but it remains unclear which activities benefit and how. This paper reviews 15 studies from the last five years, classifying AI approaches with an established RE framework. Current work focuses on operational tasks: requirements determination, analysis, consolidation, and traceability. About two thirds address single activities rather than integrated solutions. Early-phase tasks like knowledge elicitation receive little support despite being central to practice. The mapping clarifies existing AI support and gaps for future work.

We present FORGE (Framework for Optimization and Reinforcement-driven Generative Engineering), a probabilistic programming framework for generative design that unifies declarative, symbolic modeling and reinforcement learning (RL). FORGE can learn and refine a design generator through RL based on simulator-derived rewards. We demonstrate FORGE across several vehicle domains. FORGE creates an extensible, interpretable foundation for generative engineering. It can act as both a data generator for machine learning and a design optimizer, offering a practical alternative to purely neural methods.

This work explores Reinforcement Learning (RL) for the circular design of planar truss linkages using available bars and pins. A bipartite graph representation and elementary action formulation enable agents to assemble mechanisms in a physics-based environment. Results for a force-inverter design problem show 98.5% success for fixed-stock training and 66.0% for shuffled stocks. The method demonstrates RL’s potential for inventory-constrained mechanism synthesis, with future work targeting scalable, indexing-invariant architectures and more flexible connection actions.

This study examines how engineers formulate natural language prompts for searching existing assemblies in mechanical design. A survey with 48 engineers produced 169 prompts for different assemblies. Results show that prompts are mostly written as bullet points with an average of three and up to seven requirements. The engineers describe assemblies mainly through implicit functional descriptions and geometric or physical parameters. These findings form an empirical basis for developing generative AI-driven, prompt-based systems to foster design reuse.

Design thinking fosters creativity but it’s susGceptible to cognitive biases. We propose a rule-based framework supported by large language models that uses a Prompt-Reflection-Reframe loop to identify bias mechanisms in designers’ verbal reasoning and generate theory-grounded reflective prompts. Through scenario-based evaluations, we validate the framework’s theoretical foundations and establish a methodological basis for supporting bias-aware design practice.

Speech-capable AI systems introduce new possibilities for communication and collaboration in design, yet methods for analysing human-AI interactions through speech remain limited. This paper proposes and applies a method for analysing conversational interactions in speech-based human-AI design activity. Grounded in conversation analysis, this method reveals how conversational structure and designer roles emerge through spoken interaction, offering an analytical framework for examining communication, cognition, and collaboration in design.

Engineering organisations increasingly aim to reuse historical BOM, CAD, and requirements data to identify recurring components. A key prerequisite is Entity Matching (EM), whose performance on heterogeneous engineering data is unclear. This paper evaluates classical models, zero-shot LLMs, and hybrid EM on Amazon–Google and a multimodal engineering dataset. Random Forest and XGBoost achieve near–state-of-the-art results; LLMs perform well but are costly, hybrids add little. EM transfers under controlled conditions and forms a foundation for reference architecture reconstruction.

The use of artificial intelligence, especially large language models (LLMs), is increasingly explored to support early system development. This paper evaluates low-threshold LLM-based tools for supporting conceptual design. Through an experiment, two LLM-based tools were compared generating alternative solutions using an existing function model of an electro-mechanical system as input. Functions were provided in natural language and using the Functional Basis. Results show limitations and significant potentials for effective and efficient conceptual design support.

OEMs are shifting from product-centric offerings toward smart services, but adoption is still hindered by technical development barriers. We conduct a systematic literature review of peer-reviewed studies with original industrial evidence and identify eight barrier categories across data, semantics, integration, governance and modularity. We map them onto a smart-service architecture and key analytics roles, and relate them to GenAI building blocks such as LLMs and knowledge graphs, outlining a research agenda for overcoming technical barriers towards scalable OEM smart service development.

Patents contain valuable design insights, yet manual analysis remains time-consuming and complex. This study explores Large Language Models’ capacity to automate patent analysis for engineering design. GPT-5 and Gemini 2.5 Pro were evaluated across Motivation, Novelty, and Key Invention Features using three patents and expert evaluators assessed outputs through Accuracy & Fidelity, Comprehensiveness, and Analytical Depth. Results indicate LLMs demonstrate proficiency in feature synthesis but exhibit inferential limitations in motivation analysis, underscoring the necessity for human oversight.

Data-driven design increasingly relies on sensor data, yet these thin measurements often lack the experiential context needed to explain why events occur or what users feel and need. This limits their value for human-centred design. Passive and active contextualisation are introduced to describe how meaning is produced through inference and user participation. A real-world case study using See.Sense cycling data from Melbourne shows how combining thin and thick data produces more situated understanding and actionable design insight.

Small and medium-sized enterprises often lack the time, expertise, and tools for effective scenario management. This paper proposes a modular, AI-enabled scenario architecture integrating a guided wizard and expert environment on a shared knowledge backbone. The design aims to reduce effort and tool fragmentation while preserving human judgment, structural quality, explainability, and traceability. The proposed pattern outlines a provenance-aware foresight pipeline with human-in-the-loop capabilities that aims to transform one-off projects into reusable organizational knowledge.

Engineering software is evolving through the integration of artificial intelligence, creating new opportunities for enhanced assistance within product development. This paper proposes a use context model to systematically align and classify the functionalities of Generative Engineering and Design software with respect to the combination of product development phase, the nature of the task, and the level of support provided. Based on this model, a methodological guideline is proposed, offering a structured framework for the development and application of these tools in product development.

This study evaluates the efficacy of various freely available Large Language Models (LLMs) in conducting semi-automated purpose-oriented technology searches to support design activities as well as Technology Intelligence for innovation management, using a systematic manual search as a baseline for comparison. The case to run the comparison focuses on identifying water purification technologies suitable for mobile systems. The results show that LLMs can target more technologies than human-based searches, reducing time demands and providing wider entry points for additional technology analysis.

Imperfect CAD models with non-smooth features are common outputs of the latest digital tools. These are unsuitable for the feature recognition needed for end applications like computer-aided manufacturing. This paper proposes to recognise features from imperfect models by contributing a comprehensive dataset, a novel data surrogation method, and ML-based automated feature recognition model. Results show that the data surrogation method accurately replicates manual imperfections with voxel accuracy >0.9 and a Dice coefficient >0.6. Ultimately, feature recognition achieves 92.8% test accuracy.