The Shift from Generating Designs to Making Product Decisions
Generative AI has changed the speed and economics of early product development. A written requirement, rough sketch, or reference image can now become multiple visual concepts and increasingly sophisticated 3D candidates far earlier in the process. Designers can explore more alternatives before engineering resources are committed, while AI-assisted modeling, simulation, and digital workflows are moving more development decisions into virtual environments. But faster generation creates a new constraint: the ability to produce more designs does not automatically improve the ability to choose the right one. As the cost of creating another option falls, evaluation becomes more important than generation.
This distinction is especially important for physical products. An AI system may generate a convincing headset, wearable, seat, hand tool, medical device, or automotive control without knowing whether the geometry will accommodate the people expected to use it. A concept can look plausible in a rendering and still create interference around the ear, unstable sensor contact, excessive reach, poor pressure distribution, or restricted movement. Product development therefore requires a different level of intelligence from image or geometry generation alone: AI must reason about the relationship between the product, the human body, and the conditions of use. This transition from rapid generation toward evidence-based evaluation extends the workflow explored in From AI-Generated Design to Human-Fit Validation: A New Product Development Workflow.
Human Variability as an Engineering Input
Human-centered product design has traditionally relied on anthropometric tables, selected percentiles, reference mannequins, small user studies, and a limited number of physical prototypes. These methods remain valuable, but each simplifies the population being designed for. Two people with similar height and weight can differ substantially in shoulder slope, torso depth, facial contour, hand geometry, limb proportions, posture, and local body shape. A single average body therefore cannot describe the combinations of characteristics that determine whether a product fits, remains stable, can be reached comfortably, or distributes load appropriately across a target population.
The limitation becomes more visible when products interact closely with the body. Wearables must maintain contact while the user moves. Head-mounted products have to accommodate differences across the nose, face, head, and ears simultaneously. Seats must respond to changes in posture and supported tissue, while workspaces and vehicle interiors have to account for reach, clearance, visibility, and occupied space across different body proportions. Static dimensions alone cannot describe all of these relationships. 3D human body data, body-shape diversity, anatomical landmarks, joint locations, posture, and movement data make it possible to treat human variation as a design variable rather than as uncertainty left for late-stage testing. The same shift from isolated movement metrics toward richer human-product interaction data is examined in Beyond Joint Angles: How 3D Human Body Data Is Redefining Human-Centered Product Design.
From Human Data to Computational Human Understanding
The next step is not simply collecting more measurements. Product development AI needs structured human representations that can participate directly in analysis. A 3D body scan preserves spatial relationships that conventional measurement tables cannot fully capture, including local curvature, cross-sectional shape, asymmetry, surface geometry, and relationships among multiple body regions. When those scans are combined with consistent anthropometric definitions, landmarks, skeletal or joint structures, population attributes, and relevant postures, they can support digital human models and representative 3D personas that allow a product to be evaluated against realistic differences among users.
This creates a different role for AI. Instead of generating product geometry and handing the result to humans for all subsequent judgment, AI can increasingly help identify relevant human data, select representative user models, compare product alternatives, detect potential interference or accommodation problems, and determine which conditions deserve deeper simulation or physical testing. Virtual evaluation can then examine questions such as fit, clearance, contact regions, reach, posture, and eventually more complex interaction behavior while design decisions remain reversible. The objective is not to replace physical prototypes or user studies, but to use them more selectively and with better evidence. Teaching AI to Understand the Human Body: 3D Human Data, Digital Humans, and Product Validation provides the technical context for this shift from demographic descriptions of users toward computational representations of human geometry and interaction.
A Human-Aware Workflow for Product Development AI
This is the direction in which Comfo Labs is developing its human-data and AI product-development capabilities. SIZE LAB turns anthropometric measurements, 3D body-shape data, and representative 3D Personas into resources that designers and engineers can use when defining target users and evaluating human diversity. The platform is intended to make human evidence usable during design rather than leaving it in separate databases or reference documents. Doodll extends that foundation into an AI-assisted product-development environment connecting concept development, 3D exploration, review, human-product evaluation, and design decisions before repeated physical prototypes are built.
The larger objective is not simply to add another AI tool to the design process. It is to change what product-development systems can reason about. A useful AI workflow should eventually connect product geometry, human variability, physical interaction, and design constraints so that teams can ask not only what can be generated, but who a design works for, under which conditions, and where it is likely to fail. How Digital Human Simulation Validates Products Before Physical Prototyping shows how moving this evaluation upstream can complement rather than replace physical testing. The defining capability of product development AI may therefore be less about the number of designs it can create and more about whether it can understand the humans who will ultimately have to wear, hold, reach, sit in, operate, and live with those designs.
