We built our own AI tools, and expertise runs every one.

We keep a deterministic database of over 1,000 Design 1st projects spanning dozens of industries: budgets, timelines, complexity, requirements, results. Our AI budget estimator pairs that record with machine learning and AI inference to produce instant estimates, so you see realistic ranges and the assumptions behind them. It is not a chatbot wrapper. The value is the knowledge underneath it, and how our AI tool presents the information. Be aware the budget is as good as the prompt you give it. Use the tool, then call us and we can make sure a budget includes the key features and functions of your new product idea.

Every product development is unique, so the database points an ROM arrow in a direction. A thorough budget still comes from a conversation with our technical team to ensure all your requirements are included. For example: did the type of radios the electronics have to run get specified correctly? If not, hundreds of additional hours may be required.

Ian Kayser, our VP of Sales, has budgeted more than 7,500 projects in 25 years at Design 1st. His intuitions, plus our technical team reviewing your product requirements, get amalgamated into a more accurate budget for the work required to go from your idea to a product in the market.

Treat our AI budget tool as a financial planning instrument for ballpark dollars and timelines. It is not a Gantt chart activity roadmap. The many activities specific to a new product idea have to be assessed case by case by our development experts, and their analysis with you may move the estimate up or down.

Our public AI budget tool includes six example real projects we worked on, with their budgets, as part of the training data, from a simple hand tool to a complex wearable. Our private AI budget draft tool is what produces the actual budget released for sign-off paperwork, after we have discussed the details of your project with you and our technical team.

Try the budget estimator →

Dreaming up the next out-of-the-park product concept that addresses a real market need and improves a user's workflow: this is step one. Drafting the targets and requirements to guide the development team is step two. Cost, price, feasibility and supply planning follow from there.

Drafting the requirements (PRD) is labour intensive, boring, has to be meticulous, and is the last thing anyone wants to spend time on or document. There are hundreds of details to work through, and many need research before they can be pinned down, which makes it an excellent place for AI to help. Our AI PRD tool drafts requirements across the five critical areas (business, market, user, product and supply) and outputs them as a tabbed Excel document for humans to filter, adjust and refine the last 10%.

The draft PRD is never correct. You will research further, tune, remove, add and re-specify. What it does is take roughly 90% of the research and documenting work, typically months of effort, out of the process, and hand the development team a running start. Many items cannot be fully defined early on and will be updated as decisions get made while the development proceeds, and the PRD is an excellent place to track all the decisions that get made on the road to finishing the product development.

Try the PRD drafting tool →

A second early stage AI Target Market Research tool we created looks through internet information on markets, distribution channels and likely competitive products, and identifies the opportunities, constraints and risks you may face selling into your chosen market, along with geographic and ethnographic detail estimating the size of the opportunity.

Try the target market research tool →

A third early stage AI Personas tool we created is complementary to the Target Market Research tool: it drafts both primary and secondary user personas for the likely consumers, buyers and users of your new product. This feeds into the market research.

Try the user personas tool →

The PRD requirements tool available on our website includes four real projects with their requirements drafted, from a water bottle to pet tech. You can also find a funding help search tool there. There are example projects with all of the tools to try out if you do not want to share any information about your specific product. Get a feel for the tools, then sign an NDA with us and we can run the tools with you after that.

Try the funding search tool →

Our AI Visual Concept Exploration tool is used by senior product leads to explore form, fit and function, and to move quickly through a dozen concept directions early, while the core of the product is still being decided. We pull from the many products and images in the world via public LLMs as well as 6,500+ of our own concept sketches. The tool outputs are used for conversations, debates, option exploration, client feedback and synthesis of various visual elements as the team begins the creation of the geometries that will define the direction of the product to be developed in CAD tools on the path to prototyping, tooling and manufacture.

Hand sketching still does a higher fidelity of nuance that the AI tools do not yet provide. We draw dozens of rough sketches per client, not photorealistic renders, because rough is where the emotional wow, usability and the human factors get explored, along with quick build prototypes to hold, assess and refine.

The math: concept exploration used to require a table of industrial designers at $1,000+ an hour. Now one senior lead with the tool brings the same creativity and output to the larger product level planning and decision making team. Development cost drops because this exploration stage moves faster. Our hours required stays similar and the team explores many more options in detail before committing to a direction.

Finding physical parts. Our AI Finder internal tool scans open-source multi-country customs records to find the factories that could actually make a part at the yearly volume you anticipate, instead of trusting a web search and a catalogue listing. From those early supplier options it collects specification sheets, contact information and comparative reliability for the intended application, for our design team to review as we select parts against the targets the business and marketing teams set.

FEA thermal analysis. We use simulation for feasibility from deterministic physics: SolidWorks FEA for heat, stress and deflection, and CFD for fluid flow. We design and check heat dissipation with traditional CAD and FEA to get system-level feasibility for our housing concept when the chance of thermal overload is identified as risky. Traditionally we then collect real data with bench tests, thermocouples and a thermal chamber. AI comes in at the beginning and the end of the process: to assess system level feasibility of heat generated inside a volume prior to FEA, so no time is wasted on obvious no-go designs, and afterwards to help with data analysis. This speeds up the analysis and guides the decision on which option for product direction to carry forward.

Bill of materials. A BOM starts with critical parts selection at the concept stage, and then gets further filled in right through detailed engineering, prototype and testing, where every component is selected and prototype and production suppliers are finalised. The evolving BOM becomes the order list for the first pilot manufacturing builds. Our AI BOM creation tool makes the creation and evolution of the BOM throughout development stages faster, and eliminates the manual errors traditional to assemblies with hundreds of parts that all have to come together with no confusion.

None of these three AI tools replaces an experienced mechanical engineer's judgment, as they must review all outputs for accuracy. The AI tools let an expert move faster, cover more options, and collate a large amount of information into organised documentation.

PCB Layout AI Estimator 3D. A spreadsheet of components and a board shape goes in; an optimised board component layout comes out, in three dimensions. Arrange parts by group, flip and rotate them, check top and bottom views, all before a schematic exists. Working closely with the industrial designers and mechanical engineers, board size, shape and depths can be optimised early, which is essential for small products where there is almost no room to give. Same rule as the AI budget estimator: it gives the team a direction to chase, not a finished circuit board. The tool is evolving toward trace route planning, layer count, and risk assessment on critical components.

Wireless power design AI tool. Our electronics team has spent a year deep in wireless power transfer theory, and AI has been implicit in the review of more than 100 research papers on this new technology. The current status is an optimisation and modelling engine that captures candidate solutions against product requirements in days. In the senior engineer's own words: “It's more complicated than it looks on the surface.” A year of accumulated expertise gives our clients access to leading-edge wireless power application effectively instantly, without the parameter, application, materials and testing research that takes months.

Our firmware team built and runs our in-house AI server, with open models. This internal AI server is also the same one that the rest of the company's other AI tools work on. We scale the selected open-source models as our needs grow. New models are tested against real work from our wide array of past projects, and our VP of Software, Yih-Shyang, signs off before any new implementation is used on client work.

AI firmware creation toolset. Our AI coding tools are built for designing and testing physical device code: peripherals, microprocessors, PLCs and display drivers all need specialised integration code. AI drafts libraries and the layers around the core: test harnesses, build scripts, protocol glue, documentation. Register-level and timing-critical code stays human-written, and every AI-drafted line gets the same review as a human-written one.

A recent example: a major telecom client's outdoor control device had a 1,000-hour software budget, far smaller than a traditional non-AI-supported build. AI assistants carried the scaffolding; our engineers created and tested the code that matters. Modules reused from previous designs already carry their testing and validation hours and needed only integration testing. Your budget goes to human architecture and judgment, not typing.

Virtual prototyping before bench testing. To devise high-level fluid flow models that represent the product configuration and its constraints takes deep experience if the results are going to be trusted, because the simplifications and assumptions are what make the output usable or worthless. From there we move to physical bench-model testing, and that is not changing. In all but the simplest designs, the cost and time of simulation, and the trust you can place in it, require time, iteration and a large budget. Blending AI-driven information collection with finite element tools and real-world testing gets to a feasible solution for much less than a simulation-centric approach alone.

We are early to AI-driven simulation of real-world operating environments. Our large database of real testing results is what lets us confirm whether any of the emerging third-party AI tools are actually useful for physical product development.

Prototype BOM tool. A P-BOM is a variation of the production BOM, and it carries parts the final manufacturer has not made yet, which introduces uncertainty into what the testing of a prototype with non-production parts can honestly tell you. Housing parts are not molded from tools yet and come from prototype materials without the strength, ductility, surface finish or chemical compatibility of the final product. The same applies to electronics and firmware at prototype level. Our tool flags every interim part and feeds that into the test plan, so we do not test a requirement on a prototype that cannot properly be assessed during the prototype stage. Some features, functions and product materials cannot be properly evaluated until the manufacturer builds the production product later on. Knowing this reduces test iteration, development time and cost.

Test plan builder and optimiser. The Product Requirements Document (PRD) holds the specifications a product is tested against: business, user, market and product performance, objective and subjective, some solvable and some requiring modification, as decisions are made and changed during the development process. Our test plan tool uses the PRD to separate out the critical and risky elements that deserve careful testing and qualification, and sifts available results from public sources and our own private library so humans spend less time reviewing, debating and planning. Testing is one of the most expensive and time-consuming parts of development, so testing for the risky items that matter is where the money and time is saved.

Our design team supports clients through third-party regulatory testing of the product. Early in concept planning we also help organise the critical testing a client's market and country will require, using our AI research summary tool. As development progresses we run physical tests in our own lab first (EMI, ESD, RFI) to raise the odds the product passes the costly approval testing, hopefully the first time.

AI digs up the relevant information and flags where the risks are, so time and resources go to the right places. Our regulatory expert reviews the output and the client's business team adds their own expertise and learning to the plan. It maps the playing field and the industry contacts far faster than a person with a browser and a phone. It is a starting point, not an output.

Our AI regulatory tool also finds and fills regulatory form drafts, saving the days and weeks of back and forth that are required when humans create these documents. Humans always review and complete the documents ready for submission.

The AI Regulatory Assessment tool serves the development team and the client, to identify the industry experts who guide approvals. On recent medical device projects the time savings, test cost savings and breadth of exploration have paid off clearly.

This last stage of development is where public AI tools are strongest today, tapping into the online information highway full of true and false. With carefully crafted prompts, current public AI tools will produce a stunning array of information for anyone setting out to have a physical device made by a manufacturer somewhere in the world.

You should get AI to answer certain questions:

• Do I start with a manufacturer, or is what I am building going to take several expertise groups to pull it all together?
• Is a similar design already being made by an existing manufacturer, and if so will they build my design as well and protect it from competitors?
• Are my design files and documentation manufacturing-ready?
• What will engaging with the manufacturer cost, to design and to build a quantity of units?

First, getting your product design ready to manufacture, what are your choices?

There is more than one way to design, develop and manufacture a new product idea.

1. Go straight to a manufacturer and have them design and build your product.
2. Build your own development team, design it, then find a manufacturer.
3. Hire a professional design development company, then find a manufacturer.

For an innovative new product idea, setup 3 is often the only way forward. It uses professional teams trained specifically for their role in development, and is by far the most predictable route to arriving at a solution that sells well in your market, with the lowest execution risk, the least time and the lowest development cost.

As companies grow in revenue and in the number of custom SKUs, they often migrate from setup 3 to setup 2 with an internal development team. Setup 1 is most often used when copying an existing design with minor variations, where a manufacturer is already set up to make something similar and your changes are incremental to what they already produce.

Next, which tier of manufacturer can you attract?

You need to know what type of manufacturer your new product can attract. Global Tier 1 manufacturers build high-volume continuous production and deliver quality product at a premium per-unit cost. In your first year or two of sales you are far more likely to need a manufacturer willing to do smaller batch builds while you test the market, tune user expectations and grow sales.

It is simply about yearly volume economics: Tier 1 is 100,000, batch is 1,000, 5,000 or 10,000 per year.

You can save significantly in the first two years with batch builders onshore, and more saved if you go offshore, enough to get you through to a stable continuously producing product. At that point a Tier 1 becomes very interested as your volume climbs from lower-volume batches toward 100,000+ units a year.

Whichever option is open to you, make sure your file package stays with you

No matter which manufacturer you choose in year one, the product file carrying every specification they need to build your product must be professional, detailed and release-controlled. That is the development team's responsibility, in creation and in delivery.

Our output files are Tier 1 level, and manufacturers love Design 1st product files for what is in them, so they can focus on their specialties of review, minor modifications to suit their processes, build instructions, test instructions, quality control, assembly efficiency and the inline test equipment for each batch you order of continuous production.

In year one most products start with batch builds, and the design will keep changing as you tune features and adjust to the feedback from first customer use. The design company can handle these file changes, because manufacturers do not like doing that part; ideally they want build instructions that are flawlessly clear. We use release control documentation to keep your design files straight until you want to take it over yourself. Most of our clients would rather not own the update-and-release-control headache, so they let us keep the factory current with the latest information for each build.

And what will the manufacturer cost you?

There are four major cost areas to expect from a contract manufacturer.

Clients either rely on the manufacturer or on the design team to create the engineering files for any test equipment, and to decide who builds the jigs and fixtures the CM specifies and requires. We have this discussion with manufacturers as early as possible.

Then there is the timeline, which is a cost of its own. Tier 1 manufacturers can be much slower than batch builders, typically four to nine months, because bigger organisations fit smaller work in around their larger production customers. For the first builds of a brand-new product, where you want 10 units, then 100, then 500, then thousands so you can debug with real customers in the field, Tier 1 manufacturers are impossible to attract. They rarely accept new customers who need to go through that learning and tuning process at all, unless you already have a strong relationship with them on other high-volume products, or the category you are entering is strategic to their future plans.

The lower hours are already in your budget. Every budget we write is built with our in-house AI tools and the record of 1,000+ past projects, so the savings are priced in before you see a number rather than promised as a discount later. We break projects into development stages so decisions happen in an efficient order, and the saving passes through.

You can get a draft that looks feasible. That is exactly the problem. A new product crosses nine or more disciplines, and the expensive mistakes live at the boundaries between them. Knowing which shapes and interlocks will actually be moldable, what board size and layer count is producible for the functions it has to perform, what the expensive test labs will do to you, and which assembly process leads to a low manufactured cost is the synthesis work. If you are a designer, engineer or marketer, using AI to sift details yourself genuinely helps the team. It does not replace the integration.

Our AI tools run on our own secure server, in our building, on hardware we own, using open-source models. Where we use third-party cloud AI we use paid services with closed models and strict data-use rules. The in-house rule is blunt: we treat any AI chatbot like a person who has not signed your NDA. No proprietary data gets transferred.

No. AI is used to explore concepts, weigh options, analyse results and set direction. Our designers, engineers and supply teams make every call, and AI-drafted code gets the same review as human-written code. Register-level and timing-critical firmware is written by a person.

Five are public: the budget estimator, the PRD drafting tool, the user personas tool, the Target Market Research tool, and the funding search tool. There are example projects with all of the tools to try out if you do not want to share any information about your specific product. Get a feel for the tools, then sign an NDA with us and we can run the tools with you after that.

In practice the estimator lands within about 20% when your input prompt description carries enough detail. It is a financial planning number, not a quote. A released budget comes after our technical team has reviewed your actual requirements, and that review can move the amount of work required either way.