Generation Is Accelerating. Decision-Making Is Becoming the Bottleneck.
Generative AI is changing the economics of product exploration. Product teams can move from requirements and rough sketches to multiple visual concepts, design variations, and increasingly sophisticated 3D representations far faster than conventional development cycles once allowed. That expansion of the design space is valuable, but it creates a new constraint: when generating another option becomes inexpensive, the scarce capability is no longer ideation alone. It is the ability to determine which option deserves engineering resources, validation, and ultimately manufacturing investment. The issue is particularly important in physical products, where a visually convincing concept still has to satisfy requirements for usability, geometry, materials, human interaction, production constraints, and commercial viability. This shift from generation toward evaluation is already visible in From AI-Generated Design to Human-Fit Validation: A New Product Development Workflow, where faster concept creation moves the development bottleneck downstream toward judgment and validation.
For product organizations, this changes what “speed” should mean. Producing 50 concepts instead of five creates little economic value if teams still spend weeks debating which concept to pursue, discover fundamental problems after detailed engineering, or restart development after physical prototypes expose issues that could have been identified earlier. Decision confidence is therefore not simply confidence in an AI recommendation. It is the degree to which a product decision is supported by relevant engineering constraints, human data, simulation evidence, and clearly understood uncertainty. AI-based product development becomes strategically useful when it helps teams reduce the distance between exploration and commitment: identifying weaker concepts earlier, directing expensive engineering work toward stronger alternatives, and preserving the ability to change a design before the cost of change rises.
Human Evidence as a Requirement for Better Design Decisions
The difficulty is that physical-product decisions cannot be made from product geometry alone. A headset may satisfy dimensional requirements while concentrating load at the temple. A wearable may appear correctly positioned on a static model but lose sensor contact as the wrist rotates. A vehicle interface may technically fall inside a reach envelope while requiring uncomfortable shoulder or trunk motion for particular users. Furniture, healthcare equipment, protective products, robotics, and consumer hardware present similar problems because product performance depends partly on how geometry interacts with human variability. Traditional anthropometric tables, percentile values, static measurements, and small test samples remain useful, but they simplify the population. People with similar stature can differ substantially in limb proportions, local curvature, body depth, joint mobility, posture, and movement strategy.
Structured 3D human body data, anthropometric measurements, anatomical landmarks, posture data, movement information, and representative digital human models give product teams a richer basis for evaluating those differences. Instead of asking whether a concept fits an average user, teams can examine accommodation across meaningful body-shape variation, potential interference, reach, clearance, contact regions, posture, and movement before a design is locked. That principle is developed further in Product Development AI Needs to Understand Humans, Not Just Generate Designs. The objective is not to claim that digital models eliminate engineering judgment or physical user testing. Their value is to create better evidence earlier, so that real-world testing can focus on the users, conditions, and design risks that matter most.
Virtual Validation, Optimization, and the Economics of Earlier Decisions
Once product geometry and human evidence can be evaluated together, virtual validation becomes more than a visualization step. Candidate designs can be screened against representative users, fit constraints, contact conditions, movement scenarios, engineering requirements, and other relevant parameters while modification is still relatively inexpensive. AI can then support comparison rather than generation alone: ranking alternatives, identifying constraint violations, revealing where evidence is insufficient, and directing teams toward the next simulation or physical test. Generate → evaluate → validate → optimize → commit becomes a more valuable model for AI product development than generation followed by late-stage correction. How Digital Human Simulation Validates Products Before Physical Prototyping provides a complementary view of why moving evaluation upstream can narrow the designs that require costly physical testing.
The economic implications can be substantial, although industry benchmarks should not be treated as guaranteed outcomes for every project. McKinsey has estimated that generative AI could create productivity value equivalent to 10–15% of overall R&D costs, with product R&D opportunities including faster virtual design, simulation, and physical-test planning. In a separate analysis of generative design applications across industries, McKinsey reported cases involving 30–50% shorter development time and 6–20% lower part costs. More recently, BCG estimated that AI applications in automotive CAD could reduce design cycles and rework by up to 60%, with time-to-market reductions of up to 40% becoming possible in broader AI-enabled engineering environments. These figures describe different technologies, industries, and maturity levels, but they point toward the same business mechanism: earlier, better-informed decisions can reduce downstream iteration and improve the productivity of expensive engineering resources.
From Product Intelligence to Confident Manufacturing Decisions
This is the broader direction behind Comfo Labs. Rather than treating AI design, human data, simulation, and validation as isolated tools, the company is developing them as connected layers of product-development intelligence. SIZE LAB provides anthropometric and 3D human-body resources that can help development teams characterize target populations and work with representative 3D Personas instead of relying only on abstract demographic profiles. Doodll extends the development environment toward AI-assisted ideation, sketch exploration, design variation, 3D asset development, and production-oriented product development. Together, these capabilities support a shift from designing first and validating later toward evaluating product decisions earlier, when teams still have room to change direction.
The strategic opportunity is larger than automating individual design tasks. A more mature AI product-development system should help an organization understand what can be generated, what can be rejected early, what requires further evidence, what should be optimized, and what is ready to move toward manufacturing. Human data and product-human fit analysis become important because they add evidence about the people for whom the product is being built; simulation adds evidence about likely behavior before physical testing; AI helps connect and interpret those inputs across a growing design space. The result is not certainty, nor the removal of designers and engineers from the decision process. It is a transition from assumption-heavy development toward data-driven product decisions with greater decision confidence—giving teams a stronger basis for committing engineering time, prototype budgets, tooling, and manufacturing resources to the designs most likely to deserve them.
