Skip to the stories
Mechanical & Manufacturing

AI Product Concepts Can Leave Design Agencies Out of the Work That Shapes a Manufacturable Product

  • DEVELOP3D
  • 602 reads

AI-generated concepts can help startups visualize ideas, but late design involvement may raise costs and leave weak products unchallenged.

AI Product Concepts Can Leave Design Agencies Out of the Work That Shapes a Manufacturable Product

AI-generated product concepts can make an idea look ready for development before anyone has checked whether it can be built, priced or sold. For product design consultancies, that changes when clients seek help: some arrive with polished visuals and expect engineering to turn them into physical products, after the decisions that most benefit from early design work have already been made.

Nathan Partington, founder of Lancashire-based product design and engineering consultancy npd, says that pattern has become common among enquiries from product start-ups. He has worked in product design for two decades, and describes a shift in how prospective clients approach agencies: they may use AI to develop a concept, then treat the consultancy’s role as making that concept real.

That can save a start-up money at the outset and help it explore an idea on a limited budget. It can also leave less room for designers to question the product’s purpose, simplify its design or resolve manufacturing constraints before development gathers pace. Design for manufacture depends on those fundamentals being considered early, when changes are generally easier to make.

Visual concepts can obscure the cost of development

Partington recalls discussing a prospective client’s concept work and estimating that comparable conceptual development by the consultancy would have cost about £20,000. He believed that work could have produced a better end product and made development simpler. But if a client completes that early work with AI, it may also remove a significant part of the agency’s work from the project.

The business effect isn’t limited to lost concept fees. A project that reaches specialists only after its core decisions have been made can be harder to steer. The client may already be attached to a particular appearance or feature set, even if those choices complicate tooling, manufacture or the route to market. A convincing image can communicate intent, but it doesn’t establish that a product will work for users or justify its development cost.

Partington also sees AI-generated estimates shaping expectations about project budgets. If those estimates understate what development requires, a start-up may commit to a plan that doesn’t account for engineering, tooling or redesign. His argument is that specialist input early in a project can cost less than correcting avoidable problems later. The source doesn’t give a measured comparison of those costs, but identifies tooling changes, redesign and time to market as areas where early decisions matter.

What npd’s enquiry data shows

To understand the change, npd began tracking how incoming clients found the consultancy, why it declined enquiries and the estimated value of projects described in AI-generated briefs. After several months, the average estimated project value was around £15,000, compared with the firm’s target of £50,000.

The consultancy rejected nearly all of these enquiries or tried to discourage clients from pursuing their ideas. Partington says the concepts often hadn’t been validated or qualified, and some didn’t fit the company’s ethos. The rejection rate rose to 89%. In his view, the volume of concept images can create a stream of speculative enquiries without a matching increase in viable projects.

AI can produce a requested concept without assessing whether it has a credible path to becoming a product people want to buy.

That distinction matters for engineering teams. Generating a visual response to a prompt is different from evaluating user needs, ergonomics, materials, manufacturing processes and commercial fit. An image can help communicate an idea, but it doesn’t supply the evidence needed to decide whether the idea deserves development. Partington’s concern is that users may interpret a responsive system as validation, even when it hasn’t tested the assumptions behind the concept.

Where AI fits in the design workflow

Partington reports that established clients have not shifted their requests in the same way, and that senior design managers he has spoken with describe limited adoption inside their companies. He attributes that caution, in many cases, to concerns about intellectual property ownership and cybersecurity requirements. Those comments describe the experience of the people he spoke with, rather than a measure of adoption across the industry.

At npd, the core design process continues to use Onshape for CAD and KeyShot for visualisation. The team still develops concepts through discussion, observation and work with people who have practical product-development experience. It uses automation for some administrative tasks, including project planning, note-taking, data analysis and reporting, while keeping client-facing concept work in the hands of its designers.

That division reflects a practical way to assess AI in product development: ask which tasks benefit from faster processing, and which require judgment grounded in users and manufacturing. Automating routine documentation can free up time. Generating a concept can help a client express an idea. Neither step, by itself, resolves whether the design is comfortable to use, feasible to produce or suited to a real need.

Partington argues that human experience remains central to those questions, including the first-use response to a product and the feel of its interaction. AI tools may change quickly, and their capabilities will keep developing. For design teams, the immediate task is to use them where they support the work while keeping experienced engineering judgment involved early enough to shape the product.

About the author: AI

Nathan Partington is founder and CEO of npd Studio, a product design and engineering consultancy based in Lancashire. He started the company in 2014 after two decades in product design, with a stated principle of working only on products that deserve to exist.

Source: DEVELOP3D · develop3d.com

From the same set