AI as a Copilot Keeps Engineers in Control of Electronics Design

When incorporating AI in electronics design, the most important question is not how much of the design process AI can automate, but how to introduce automation while retaining the engineer as the principal decision-maker.

What you'll learn:

  • The migration from automation to assistance.
  • The pathway to implementation.
  • Why agentic AI changes matter.

Electronics design isn’t simply a matter of producing an answer from a set of inputs. Engineers must interpret incomplete requirements, evaluate competing architectures, understand physical and electrical constraints, make tradeoffs, and ultimately take design responsibility. Artificial intelligence (AI) can accelerate many of these activities, but that doesn’t necessarily mean it should take ownership of them.

The emerging AI model is therefore less about autopilot and more about a highly capable copilot: A system that can investigate options, perform repetitive work, highlight problems, and suggest solutions. In the meantime, the engineer stays responsible for the ultimate design direction.

From Automation to Assistance

Electronics engineers already work with enormous quantities of information. Component datasheets can run to hundreds of pages, while selecting a suitable device may involve comparing specifications, availability, cost, power consumption, and even compatibility with other components.

The challenge isn’t necessarily a lack of engineering tools. It’s the amount of information and number of possible pathways that engineers must navigate within finite project schedules.

This factor is particularly evident at the beginning of a project, when requirements may still be incomplete and architecture not yet established. At this stage, engineers are often working in what’s sometimes described as “fuzzy” space. They know what the product needs to achieve but probably remain unaware of the optimal implementation strategy.

Here, AI can be particularly useful, turning loosely defined intent into something more structured. An engineer might describe the intended function in natural language or begin with a rough functional sketch. Using such inputs, AI can help identify the required building blocks. It’s also able to ask questions about missing information and present possible approaches.

Crucially, this doesn’t require AI to make the architectural decision. Instead, it gives the engineer a faster way to explore the design, keeping the “what” with the engineer.

The Pathway to Implementation

One useful principle for AI-assisted electronics design is to separate the definition of what the system needs to do from the detailed question of how to execute implementation.

High-level functional architecture could provide that level of separation. Rather than immediately committing to individual components and connections, an engineer can establish the major functional blocks and their relationships, creating a representation of design intent that allows for subsequent refinement.

AI can prove an enormous help at this stage. It’s able to search component information, identify candidate devices, draw up functional circuits, and generate an initial schematic or bill of materials. It also makes it practical to explore several possible implementations rather than spending significant time developing just one.

The engineer, meanwhile, remains able to ask pertinent questions, such as why one particular approach has been proposed. Changing the constraints, rejecting a component, or choosing a different architecture are further options that remain open to the engineer.

Ultimately, some decisions are routine and can reasonably benefit from automating. Others require the engineering knowledge and experience — as well as the product-specific judgment — that determine whether a design is genuinely suitable.

Why Agentic AI Changes Matter

The development of agentic AI introduces another dimension. Early AI-assisted tools often behaved like a conversational interface attached to a large language model (LLM). The engineer issued a prompt and received a response.

Agentic systems can take a more structured approach. Rather than simply generating an answer, an agent can break a goal into tasks and use specialist capabilities to execute those tasks in sequence, then check the result against original demands.

This capability is key for electronics design applications. A robust AI-assisted workflow can begin by capturing the requirements and separating what’s known from what has been assumed or left unspecified. It’s then able to create a plan and execute individual stages. AI can also verify the resulting design and, if a requirement changes, revisit the process rather than simply modifying one isolated output.

Such behavior is easier for an engineer to supervise. A visible plan, progress through defined stages, and explicit verification provide considerably more insight than receiving a finished design from an opaque system and being expected to trust it.

Verification Can’t be Optional

The physical nature of electronics places a particularly high burden on verification. A software developer can usually run a test suite against a change and quickly identify whether it affects existing functionality. In contrast, electronics design requires its own layers of checking. These extend from electrical connectivity and design rules through simulation, signal integrity, power integrity, manufacturability, and other physical constraints. As AI agents gain greater ability to modify design data, these checks become even more pertinent.

One promising approach is to make verification continuous rather than treating it as a final-stage activity. An AI-generated component choice, for example, could be checked against the requirements that led to its selection. A proposed change might trigger appropriate design-rule or simulation checks. The results can then be returned to the engineer or, where appropriate, used by another agent to refine the work.

This workflow creates a feedback loop rather than a one-way process. The principle is straightforward: AI should generate as well as demonstrate that what it generates remains within engineer-established boundaries.

The Engineer as Orchestrator

There’s another benefit to this model. Engineers should not have to become supervisors of disconnected AI systems. If an engineer has to transfer information manually between component-selection software, schematic capture, simulation, mechanical design, and other tools, adding more AI could simply create another layer of complexity. The more useful vision has agents and engineering tools exchanging relevant information while the engineer retains control of the overall design intent.

This approach could eventually make cross-domain engineering far more practical. A change to an electrical design might have implications for thermal performance or mechanical packaging, or even power requirements. Rather than requiring an engineer to investigate each domain separately, connected agents could identify the potential consequences and inform the engineer for consideration.

The engineer becomes less of someone manually operating individual tools and more of a design process orchestrator. This isn’t a reduction in engineering expertise. It potentially gives that expertise greater leverage.

Designing Faster Without Thinking Less

The strongest argument for AI in electronics design is not that it can replace engineering judgment, but that it can make engineering judgment more valuable.

If AI can reduce the time spent trawling through datasheets, creating repetitive schematic connections, transferring information between tools, and checking routine conditions, engineers can devote more attention to key factors like architecture and tradeoffs, as well as innovation and product requirements.

It can also make experimentation cheaper. Instead of committing substantial time to a single architecture before discovering a problem, engineers could explore several alternatives earlier in the process and compare their implications. This possibility is particularly important as electronics applications become more complex. More capable tools allow engineers to tackle problems that were previously impractical with available resources.

The same pattern has occurred repeatedly in engineering and software development. Better tools have automated lower-level tasks without eliminating the need for people who understand the underlying discipline. They have merely changed where those people spend their time. AI can do something similar for electronics.

A Copilot that Knows Its Limits

For AI-assisted electronics design to deliver on that potential, trust will depend on more than the sophistication of the underlying AI model. Engineers need visibility into what the system is doing and control over the constraints it’s working within, as well as mechanisms for checking its output.

This is the thinking behind the current development of platforms such as CELUS. Its Design Assistant is evolving toward an agentic architecture in which requirements are captured before design development; work is planned and executed in stages; and the resulting design is checked against requirements (see figure).

The platform's approach also emphasizes keeping design decisions open to the engineer rather than presenting AI-generated results as immutable answers.

No Need to Choose

AI-assisted design will be most useful when it respects the distinction between automating work and automating responsibility. Engineers should not have to choose between embracing AI and retaining control of their designs. With the right workflow, they can have both.

AI operating as a copilot is able to process information, explore different routes, perform routine tasks, and flag potential issues that lie ahead. However, the engineer still decides where the aircraft is going.

About the Author

Tobias Pohl

Tobias Pohl

Co-Founder and CEO, CELUS

Tobias Pohl is the CEO and Co-Founder of Celus. He earned his BS in Mechanical Engineering and his MS in Automotive Engineering from Technische Universität München in 2016 and 2017, respectively.

Sign up for our eNewsletters
Get the latest news and updates

Comment About the Article

To join the conversation, and become an exclusive member of Electronic Design, create an account today!