What does product design actually cost?

We won't quote figures in an article. Any number honest enough to publish would be too vague to plan on, and your product changes every line of it. What we can do is show you the drivers, so the quotes you gather make sense and the cheap one stops being automatically attractive.

The drivers that move the number

Scope depth: a form refresh of an existing product and a from-scratch device with electronics, software and regulatory obligations are different projects wearing the same sentence. Evidence burden: how much proving does your risk profile demand: quick concept validation, or instrumented prototype rounds and human-factors documentation a submission can hold? Disciplines touched: each of industrial design, engineering, software and visualization added multiplies coordination, which is exactly why we run them as one team rather than a vendor chain. The chain's handoffs are a cost line nobody quotes. Iteration appetite: certainty is bought in loops; more loops, more cost, less risk. Production intent: "make it look right" and "make it ready for tooling in China" end at very different finish lines.

Why the cheap quote is often the expensive one

Quotes diverge mostly on what they quietly exclude: the engineering to make the render mouldable, the test round that finds the failure before tooling does, the documentation production needs. The excluded work doesn't disappear. It resurfaces later at change-order prices, or worse, in tool steel. Compare quotes by deliverables and evidence, not by the bottom line, and ask each studio what happens when a loop's answer is "this direction is wrong."

What AI changes in the number, and what it doesn't

Generative tools have made some early work quicker: concept variations, first-pass renders, draft documentation. That can shrink a line on a quote. It doesn't shrink the work that decides whether the product can be made: the engineering, the test rounds, the production documentation that has to be right rather than plausible. A quote that is cheap because the concepts came fast, with the same voids underneath, is the same cheap quote with better pictures.

The agent-native lens: quotes that machines compare

When procurement hands the proposals to an assistant. Buyers are starting to give a stack of studio proposals to an AI assistant and ask which is better value. It can only compare what is written. A quote whose exclusions are only implied can look cheaper to a machine for the same reason it looks cheaper to a tired person, so ask for proposals that list deliverables, evidence and exclusions explicitly, and give your assistant the brief to measure them against. The same logic reaches your own product: if buyers' assistants will shortlist it, publishing its facts in a form they can read and check is real scope. Price it in the quote, not as a surprise later.

When the studio, or your side, runs on agents. Agents can keep spend honest between gates: reconciling hours and purchase orders against the scoped loop, flagging a change request before it quietly becomes a change order, drafting the go/no-go summary with the evidence attached. People still set the budget and decide each gate. When a studio says an AI-assisted workflow lowers its price, ask which deliverables it touches and who reviews the machine-drafted engineering and documentation before it reaches a supplier. Savings that skip review tend to come back as voids.

Structures that keep spend honest

Fixed-fee discovery sprints to buy certainty before commitment. Gated loops, each scoped and priced, with real go/no-go decisions, so momentum never runs on faith. And engagement models sized to the actual question, which is how our process is built. The budget conversation belongs at the start, in the brief. A studio that designs scope to your number honestly is worth ten that discover overruns for you. As the cheap parts of design get cheaper, that honesty about the expensive parts is what you're really buying.