More and more people who contact us start the conversation with: "I talked to ChatGPT about the product, and it wrote down what I need." That's great. Someone who arrives with a plan like that has already thought about the product, knows how to ask good questions and saves time in the first meeting.
But there is a difference between a plan and a product that works. Here is what you should know.
What AI does well
- Organizes your thinking. Turns a general idea into a list of components and stages.
- Suggests reasonable technologies. ESP32 for a Wi-Fi product, LoRa for long range, a lithium battery for a portable device.
- Explains concepts. What a PCB is, what firmware is, why certification is needed.
- Helps you phrase questions for the development company.
What an AI plan usually misses
- Real numbers. "A battery that lasts a month" is a requirement. To know whether it is possible, you need to calculate the current draw in every operating mode, and sometimes measure it.
- The big risk. Every product has one thing that is hardest to do: sensor accuracy, communication range, heat. AI tends to present all parts as equally easy.
- Component availability. A part that looks perfect on paper may be unavailable, or go end-of-life next year.
- Manufacturing. How to build the product in 500 units, what it costs per unit, and how to test every unit on the production line.
- Certification. CE, EMC, and sometimes medical regulation. These requirements affect the design from the start.
- The hardware/software boundary. An AI plan usually splits "hardware" and "software" into two chapters. In reality, the hard problems sit exactly between them. Why this matters.
What AI can't do for you
AI can write a question and suggest an answer. It can't measure a board, see on an oscilloscope why communication drops, feel that a part is getting hot, or talk to the factory. For that you need an engineer.
AI is part of our work too: it speeds up code writing and adds a layer of review. But an engineer designs, checks and is responsible for every decision. Where AI helps in development, and where it doesn't yet.
How to get the most out of your plan
- Bring it to the first meeting. As is. It saves half an hour of explanations.
- Mark what is certain and what is a guess. "The product must work outdoors" is certain. "ESP32" is probably the AI's guess.
- Ask the development company what the biggest risk is. A good answer is the sign of a good engineer.
- Start with a short specification stage. Where the plan becomes measurable requirements, an architecture and a real cost estimate, before committing to full development.
Why continue with a small company
If you planned with AI, you are used to getting answers fast. In a small company you talk directly to the engineer who designs the board and writes the code, and get answers the same day, not after a round between managers. Small vs large company.
The next step
At ProtoMakers the first meeting is 30 minutes and free. Bring the plan the AI wrote, and we will tell you what in it is realistic, what is risky and what is missing. Book a meeting · The complete product development guide.
Written by Roman Alperovich, founder and lead engineer at ProtoMakers.