What AI PCB design can do today, and where engineers stay in charge

“AI PCB design” covers everything from a chatbot that answers questions about a datasheet to a tool that produces a routed board. The useful question is narrower: which steps of a real design can AI take on today, and how do you check its work? This is how we think about it while building Resin.

Where the time goes in a board design

A typical board passes through the same steps: gather requirements, choose an architecture, select components, capture the schematic, lay out and route the board, check it, and produce manufacturing files. Much of the effort is reading and transcription. Engineers read hundreds of datasheet pages, copy pin tables into symbols, match land patterns to footprints and wire pins one at a time. That work is careful but repetitive, and it is where AI helps most.

What AI does well today

• Turning requirements into a parts list. Given a goal and its constraints, AI can propose architectures and search for real, purchasable parts that fit them.
• Reading datasheets. It can find the pin table, the absolute maximum ratings or the reference circuit in a long PDF and cite the pages it used.
• Drawing symbols and footprints from the datasheet, followed by deterministic checks of pin counts and dimensions.
• Drafting the schematic: placing the parts, connecting the nets and laying the sheet out so a person can read it.
• Reviewing a design. A review pass can flag problems such as a missing pull-up or an unconnected pin, and explains each finding so you can accept or dismiss it.
• Placing and routing a board under your design rules, with design rule checks run on the result.

Where the engineer stays in charge

AI drafts; engineers decide. The choices that set a product’s cost, reliability and certification path still belong to the person who will sign the board off:

• Architecture and trade-offs: which interface, which power topology, how much margin.
• Critical circuits such as high-speed links, RF, power stages and anything safety-related.
• Sign-off. A clean ERC and DRC run is necessary, not sufficient; someone still reads the schematic and inspects the board before it goes to fab.

That is why every AI change in Resin is an ordinary edit in the same editor: you can inspect it, change it or undo it, and you can work without AI at all.

How to judge an AI PCB design tool

Whatever tool you try, a few questions separate a demo from something you can ship with:

• Are the parts real? Look for manufacturer part numbers you can order, not generic placeholders.
• Can you see the evidence? Facts taken from a datasheet should point to the page they came from.
• Is the output a design you can edit? A picture of a schematic is not a schematic.
• Does it check its own work? ERC and DRC should run on what the AI produced.
• Can you leave? You should be able to export editable files, such as KiCad or Altium projects, and the full manufacturing package.

Try it on a real board

The quickest way to judge AI PCB design is to give it a board you already know. Describe it in Resin, or import the KiCad or Altium project, and compare the result with your own. The documentation walks through a first project, and every new account starts with a free seven-day Pro trial.