Bringing Generative AI to the Factory Floor: Designing the Siemens Industrial Copilot
3 min read · 738 words
The Challenge
Siemens AG is one of the world's leading industrial automation companies, and its TIA Portal is the backbone of how engineers configure, program, and maintain factory systems. When Siemens set out to integrate generative AI into that environment, the stakes were high. Engineers on the shop floor work with complex, interconnected systems where a confusing interface is not just a frustration — it is a genuine operational risk.
The core problem was one of integration without disruption. Siemens needed a copilot experience that would allow engineers to orchestrate AI agents and plugins, and to work fluidly with their existing TIA Portal project context, without forcing them to change the way they think or work. The challenge was not simply to design a new feature — it was to design an entirely new interaction paradigm, and to do it in a way that felt native to a deeply technical, precision-driven environment.
This required more than visual design. It demanded a clear architectural vision for how data would flow between the AI layer and the existing toolchain, and a design process disciplined enough to move from open-ended concept all the way to launch-ready screens without losing coherence along the way.
The Approach
The work began at the concept stage — the right place to start when the product itself does not yet have a defined shape. Rather than jumping to wireframes, the focus was first on understanding the mental models of the engineers who would use the copilot: what they needed to know at any given moment, what actions they needed to take, and where AI assistance would genuinely reduce cognitive load rather than add to it.
From there, the design progressed through structured low-fidelity prototyping, which allowed key interaction decisions to be tested and challenged before any significant investment was made in visual execution. The questions driving each iteration were practical ones: How does an engineer invoke an AI agent mid-task? How is plugin context surfaced without overwhelming the primary workspace? How does the interface communicate confidence and uncertainty in AI-generated suggestions without creating hesitation?
The data-layer specification was developed in parallel with the UX work — a deliberate choice. Designing the interface and the underlying data architecture together meant that the eventual handoff to engineering was grounded in reality rather than aspiration. This reduced the distance between design intent and technical execution, and it kept the scope honest throughout.
The Solution
The final deliverables comprised three interconnected components: concept documentation capturing the design rationale and interaction principles, a data-layer specification defining how the copilot would access and surface TIA Portal project context, and a full set of launch-ready UI screens covering the plugins and agents interface.
Together, these outputs gave Siemens's engineering and product teams everything they needed to move from design into development with confidence. The concept documentation ensured that design decisions would not be misread or quietly reversed under build pressure. The data-layer specification closed the gap between what the interface promised and what the system could deliver. The launch-ready screens translated the entire vision into production-quality UI.
"Shiraz is committed to the customer's needs and endeavours to implement the best solution for the user. Very good communication and independence — a good networker with high professional competence."
The Results
The Siemens Industrial Copilot plugins and agents interface shipped with a coherent, well-documented design system behind it — one that had been stress-tested from concept through to execution. Engineers at Siemens now have a way to interact with generative AI agents and plugins directly within their existing workflow context, without being pulled into a separate tool or forced to adapt their process to the technology.
The quality of the output was recognised in the working relationship itself. The engagement spanned the full design lifecycle — from the earliest whiteboard thinking to launch support — which reflects the level of trust placed in the design judgment throughout. That kind of sustained involvement is only possible when the work consistently meets the standard the client needs.
"Very good communication and independence — a good networker with high professional competence."
The project stands as a practical example of what it takes to bring AI responsibly into a high-stakes industrial environment: not just strong visual design, but the discipline to align interaction design, data architecture, and engineering handoff into a single coherent delivery.