AI Is Changing Electronics Design

AI Is Changing Electronics Design, and Not Necessarily for the Better

Over the past couple of years I've noticed a big change in how clients approach electronics development. I think AI has a lot to do with it.

Clients used to come to me with a product idea, a problem to solve or a need they had spotted. They'd explain what they wanted to achieve and we'd work together to find the best technical solution. Nowadays I'm getting more and more 20-page AI-generated specifications, complete with BOMs, circuit architectures, PCB layout requirements, firmware plans and even cost and development time estimates. Often the client just wants someone to draw a PCB to their supposedly "vetted" spec.

The problem is these documents are often full of contradictions, unrealistic requirements and questionable engineering decisions. Components that won't fit on the specified PCB. Unrealistic power budgets. Unsuitable communication technologies, overcomplicated architectures and pages of generic manufacturing instructions. And with all that information the document still often doesn't explain what the product is actually supposed to do.

The irony is that I can feed these specs back into AI, ask it to find the technical problems and contradictions and watch the whole thing fall apart. Obviously the findings still need engineering verification. But it shows how much the quality of AI output depends on the knowledge of the person using it.

The illusion of engineering knowledge

Another issue is that AI makes clients look far more technically knowledgeable than they really are. I regularly get long emails discussing technical decisions in great detail, and then on a call I find out the person doesn't understand even the basics of electronics.

There's nothing wrong with that. I've always worked with clients who have no engineering background and that's exactly why they come to someone like me. What's different now is that they're making technical decisions based on what AI tells them, often without understanding the implications.

I've had cases where I explain why an approach isn't suitable and get another AI-generated reply defending it. So I end up arguing with an AI system through someone who doesn't understand either side of the discussion. Most clients do eventually see the value of engineering experience but getting there is getting harder.

AI-generated designs and the loss of engineering judgment

I also used to get design review requests from people learning electronics, where I had a real chance to guide someone and help them improve their designs. Nowadays practically every review request I get involves AI-generated work. Some of it is so poorly structured it's almost impossible to make sense of by hand. And often the person asking for the review can't explain the reasoning behind the circuit because they didn't really design it.

There's also a broader issue that worries me. Give the same engineering problem to five experienced engineers and you'll probably get five different solutions, shaped by their experience, the technologies they like and how they approach problems. Not every solution will be equally good but that variety is healthy. More and more I'm seeing the same architectures, the same component choices and the same generic boilerplate across unrelated AI-generated specs and designs. Very little creativity or variation. I don't think that's healthy for an industry that depends on innovation.

AI is an incredible engineering tool, but that's not the problem

I use AI a lot myself and I really think it's a game changer. An experienced engineer using AI can work much faster, explore more alternatives, catch problems earlier and potentially avoid many of the re-spins that usually come with hardware development.

But an experienced engineer also knows when AI is wrong. They understand the trade-offs, can challenge its assumptions and know what needs to be checked through simulation, measurement and testing.

The problem is that AI gives people without that background the impression they can suddenly make complex engineering decisions. And because the output looks so professional and convincing they often don't realize how much they don't understand.

What does this mean for the electronics design industry?

What worries me most is how this is changing the perceived value of engineering expertise. When someone can generate a schematic, a BOM and a full engineering spec with AI, it's easy to assume most of the engineering is already done and all they need is someone to implement it.

But the value of an experienced electronics engineer has never been just drawing schematics or routing PCBs. It's understanding the problem, choosing the right architecture, making informed trade-offs, seeing problems coming and delivering something that actually works reliably.

I do worry about where the electronics design services industry is heading. AI is making experienced engineers far more productive, but at the same time it's encouraging clients to believe they don't need that expertise anymore. I think it's going to be a hard transition for a lot of engineering businesses.

I'm not against people using AI to develop product ideas. Quite the opposite. I think it's great that people without a technical background can now explore possibilities that used to be out of reach. But there's a huge difference between using AI to understand your options and treating its recommendations as validated engineering decisions.

AI can make an experienced engineer enormously more capable. But it doesn't turn everyone who uses it into an engineer.