Inside the launch of ELECTRIX OS: The future of AI-native engineering

What Does Engineering Intelligence Mean for Electrical Engineers?

Zein starts with a definition. AI-native engineering, he explains, is “the software architecture behind a new category of software that we call engineering intelligence.” To see what changes, it helps to look at how an electrical engineer works today: the product is engineered in the head and then documented step by step in the CAD system.
This approach turns the process around. “With AI native engineering, you start with the engineering objective.” Zein illustrates this with a packaging machine that is specified directly as a goal:

• An output of 250 bottles per minute
• Allen-Bradley hardware
• Compliance with UL standards
• Redundant safety circuits
• A maximum cabinet width of 50 inches for the US market

The system then works through the requirements and constraints, asks questions where necessary, applies the engineering knowledge stored in it beforehand and develops a technical solution. Only once the engineer has reviewed, changed and approved that solution does the CAD system come into play: “It’s used to execute and document everything.”

Why Not Simply Add AI to CAD?

Could a similar result not be achieved by adding more AI capabilities to conventional ECAD software? Zein’s answer is a clear no. “If you simply add AI to a CAD software, CAD remains the centre of the engineering process, which is the de facto standard in most companies. We wanted to change that.”
More AI inside CAD, he argues, only improves the way an engineer interacts with the software. “But you’re not fundamentally changing the process how engineering happens itself.”
ELECTRIX OS therefore creates an intelligent engineering layer above the CAD tools, one that understands the objective, the standards, the requirements and the components. Zein stresses that its design is completely agnostic of the CAD tool itself, as it should be. The CAD system becomes the execution layer of a more intelligent piece of software.

What Is the Difference Between a Copilot and Engineering Intelligence?

The difference is architectural, Zein says. A copilot sits inside an application and helps the user operate it, for example by changing a wire from red to green or generating the bill of material. It then does what it is told to do, “hopefully”, as Zein adds.
ELECTRIX OS works at a different level. “But ELECTRIX OS sits above the CAD system, and it starts with engineering intent.” It asks what is being built, which requirements and constraints apply and which standards and rules must be followed. From this, it develops the engineering solution that today exists only in the engineer’s head. CAD systems, in Zein’s view, are “intelligent, sophisticated documentation systems”, and that is fundamentally what they do.
He therefore separates the two approaches clearly: “A copilot helps engineers use the software, while an engineering intelligence software like ELECTRIX OS participates in the engineering itself.”

Why ELECTRIX OS Uses a Multi-Model Architecture

ELECTRIX OS is built on a multi-model architecture. “Why? Because we don’t believe that one AI model can or should do everything.” Zein points to the clear differences between prominent models such as ChatGPT and Claude. Tying the software to a single model would, in WSCAD’s view, be a mistake. Instead, different technologies can be used for different tasks.
The same flexibility allows ELECTRIX OS to work with a company’s own internal language models, which matters for organisations with strict data requirements.

A Semantic Model, Not a Digital Twin

Does understanding engineering context mean that ELECTRIX OS builds a digital model of the machine? Not in the sense of a virtual prototype or a digital twin, Zein clarifies. What the system builds is much closer to a semantic model of the machine.
The system needs to understand what the machine is supposed to do, its functions and the requirements associated with them. When it receives requirements, it checks whether anything is missing. It also understands relationships: “If you change one requirement, it can track what else is affected.” The same applies to standards, whose meaning the system can interpret as well.
This changes the role of the drawing. “So drawing is no longer the model of the engineering problem. It becomes one of the outputs of the engineering process.” Electrical engineers, who today spend a great deal of time drawing, will leave that task to the system and focus on architectures, engineering concepts and technical problems. “Honestly, this is why you became an engineer in the first place.”

Can AI Propose Alternative Engineering Solutions?

For Zein, the ability to develop alternatives is one of the most valuable aspects of engineering intelligence. “Because if you look at product engineering, there is rarely only one technical valid answer. In fact, most of the time, there are several possible components, several cabinet configurations, or even different architectures that could be used.”
The system can work out these alternatives, and the engineer evaluates them. “You can nudge the system and say, tell me what are the pros and cons for each architecture. But you are ultimately the one deciding.” According to Zein, that is a far more rewarding job than searching the menus and submenus of a CAD program for a particular function.

How Engineers Can See Why a Solution Was Proposed

How does an engineer know why the system proposed a particular solution and which assumptions it made? Zein calls this one of the cornerstones of the development. Some creativity has its place in engineering, but there are parts of it where the hallucination that comes intrinsically with AI is not acceptable.
ELECTRIX OS therefore documents every important decision, as well as every requirement that triggers a chain of decisions. “And second, you have the traceability for everything the system does.” Five years after a packaging machine has gone into operation, the reasoning behind a component can still be traced back to the requirement, norm, standard or engineering rule it is based on.
Very few engineers can do that today, Zein says, because the CAD system does not hold this information. “So you have the documentation, you might have the schematics and you see the component, but you have no idea why that component was placed there.”

Who Is Responsible When AI Proposes a Design?

If a design created with engineering intelligence later proves to be wrong, where does responsibility lie? For Zein, the answer is built into the system: “it is crystal clear that the responsibility, the absolute responsibility for everything remains with the engineer.”
Every proposal from the system can be approved, declined or redirected by the engineer. “But you are ultimately in charge of that decision. There’s nothing the system does without your approval. And I think this is the way it should be.”

What an Engineering Day Looks Like with Engineering Intelligence

Asked about productivity, Zein describes the gain WSCAD has seen so far as “staggering, but it’s a different kind of work.” He illustrates the difference with a typical engineering day.
Today, an engineer who starts a new project at 8 a.m. spends an hour or an hour and a half reviewing and refining the requirements. They then switch to the CAD system and spend the rest of the day documenting the product they have developed in their head.
With engineering intelligence, the day runs differently. The engineer tells the system the goal and provides the requirements, whether as Word documents, PDFs or in another form. The system works through everything, raises questions and points out what has not been defined. The engineer decides, makes changes and revalidates. The system then generates the CAD files in the CAD system of choice. “So probably around 10 or 10.30, you’re ready and you’re off to your next project. And guess what? You haven’t even started your CAD system.”

Can AI Help with the Shortage of Experienced Engineers?

Too little engineering capacity is one of the challenges that also emerged from WSCAD’s industry report The State of Electrical Engineering 2026. Zein sees a direct lever here. “Because one of the strong points of this architecture is that you can embed engineering knowledge.” He compares it to taking your best engineer and copying that knowledge into the system, where everyone in the company can use it. “So even a junior engineer is able to grab the know-how of an experienced engineer once you have that knowledge inside the system.”
This differs from earlier knowledge management approaches, in which knowledge sat in a database that engineers had to search actively and then apply manually. “Well, here the advantage is you’re not even looking for it. It’s embedded in the system and the system automatically applies that knowledge.” The engineer still decides whether to approve it, and the knowledge can scale across the entire enterprise.

How Engineering Knowledge Stays Confidential

Engineering data is a company’s competitive advantage, and most companies are correspondingly careful with it. Zein gives the example of in-house engineering guidelines, which can run to 30 pages in some companies and three in others. “This engineering knowledge does not leave your company.”
Much of the processing takes place on the user’s own machine or server. For companies in industries such as aerospace or defence, which do not use public LLMs because they do not want even a single request to leave the company, ELECTRIX OS can switch to an internal LLM. “So nothing, really nothing, goes outside your company walls.”

Which Part of Today’s Engineering Workflow Will Seem Outdated?

Looking five years ahead, Zein expects one habit in particular to look strange: “I think we will find it strange that highly qualified engineers spend so much time telling CAD software exactly what to draw, what to connect.” Today, an engineering decision made in the engineer’s head is translated into hundreds or even thousands of interactions with CAD software. In five years, he believes, this process will be entirely automated, and documentation will become a downstream process that happens automatically.
Engineers will instead develop solutions, make technical decisions and execute them through appropriate engineering tools. CAD drawings and CAD models will not disappear, Zein stresses, because manufacturing, installation, commissioning, maintenance and compliance still depend on them. “But drawing? I don’t think that drawing will be at the centre of engineering anymore.”
He sums up the transition in a single sentence: “CAD digitised the drawing and engineering intelligence software begins to digitise the engineering process itself.”