Utilities today are asking ever more of their engineers: design the next substation faster, with less engineering talent available and with zero tolerance for error. Agentic AI in substation design is emerging as a new way to automate engineering workflows, improve accuracy, and help utilities deliver projects faster. These improvements are now a must. Electrification and the energy transition are lifting demand, but the need for new data centers is significantly compounding it. At the same time, the grid is aging, project complexity is rising, and backlogs are growing.
There is an irony to this situation. AI, the very technology fueling much of today’s demand growth, may also be the key to meeting it. But only if we approach AI integration with the rigor and accountability our industry demands.
The reality of AI in substation design and infrastructure engineering
There is no shortage of excitement about AI’s potential to accelerate infrastructure delivery. Across the utility sector, engineering leaders are asking the same questions: How do we move faster? How do we capture and harness the institutional knowledge of our aging workforce? How do we maintain safety and quality while scaling output?
Yet alongside that excitement is well-founded uncertainty. A highly intelligent but unpredictable system can be alarming in an area where safety is critical. Hallucination, a known drawback to large language models, is simply not acceptable in substation design, where personnel safety, public safety, and grid reliability are non-negotiable. If an engineer must take the time to verify one hundred percent of AI-generated output, the technology hasn’t reduced their workload, it’s added to it.
What the industry is seeing today is largely early experimentation. General-purpose AI tools and copilots are being used in substation design primarily for document retrieval, onboarding assistance, and ad hoc queries. Some organizations are building custom agents with specific instructions and limited context. But deep, workflow-integrated AI, the kind that can fundamentally change how substations are designed, remains rare.
The gap is clear. In software development, workers can share their entire codebase with AI agents and supervise automated workflows. In infrastructure engineering, that level of integration barely exists. The reason comes down to four prerequisites that must work together:
- Intelligent agents that can adapt their approach to solve complex problems
- The right context from standards, past projects, and organizational data
- Specific skills with detailed instructions for domain tasks
- Access to trusted engineering tools that can manipulate complex enterprise data
Without all four working in concert, AI in substation design will remain a novelty rather than a transformation.
Why general-purpose AI falls short in substation design
It’s tempting to view the latest large language models as a universal solution for all kinds of tasks. But infrastructure engineering is not a general-purpose problem. Substation design operates under rigorous safety standards, regulatory frameworks, and physical constraints that demand deterministic, verifiable outputs. A bespoke AI tool generated through casual experimentation isn’t trustworthy enough and can’t easily interact with the large, complex enterprise systems and data formats that utilities have built over decades.
What’s needed instead is AI that is purpose-built for the domain, embedded directly into the engineering tools professionals already trust, operating on structured and governed data, and producing outputs that can be traced, checked, and stamped by a licensed professional.
This is the approach Bentley Systems is taking with OpenUtilities Substation+, a new model-based substation design platform with AI capabilities built in from the ground up, not bolted on as an afterthought.
A purpose-built foundation for AI in substation design
Substation+ represents a fundamentally different approach to integrating AI into engineering workflows. Rather than creating a standalone AI tool that engineers must learn and trust independently, Bentley has embedded agentic AI directly into a trusted, model-based design environment.
The foundation matters. True model-based design, combined with a robust information schema, is built on over a decade of work with Bentley’s BIS framework and extending concepts from CIM and IFC for substations and power systems. The combination creates the structured, high-quality data environment that reliable AI requires. Without this foundation, AI is guessing. With it, AI understands the engineering context of every element in the model.
The AI in Bentley’s Substation+ isn’t just answering questions about documents. It fully understands the model—every breaker, every disconnect switch, and every transformer and its properties. It has access to the functions of the application itself. And it can call on specialized agents with domain-specific skills to perform real engineering tasks within the model.
Crucially, Bentley’s commitment to data stewardship ensures that users retain full ownership and control of their design data. Organizations decide whether and how their proprietary data is used for AI model training, with complete transparency into how models are created and what data was used to develop them. For engineering leaders evaluating AI platforms, knowing that intellectual property is protected and that AI models are auditable through tools such as Bentley’s Data Agreement Registry is foundational to building the trust required for adoption.
How agentic AI works in substation design workflows
The capabilities of Substation+ illustrate what becomes possible when agents, context, skills, and tools come together in a single integrated environment.
At the foundational level, engineers can upload design criteria documents and query them in natural language—they can ask, for example, what the breaker requirements are for a specific site. They can search the model for a particular piece of equipment by device tag, list all elements of a given type with custom-formatted exports, highlight selections, and review properties including identifying which fields still need to be populated.
But it’s the deeper AI integration that truly brings value to design. Engineers can ask the AI to cross-reference element properties against uploaded design criteria to flag conflicts. They can also calculate derived values, such as high and low side full load ampacity for a transformer, and can automatically compare them against other equipment and standards. Compliance workflows, such as FAC-008 documentation, can be guided by AI using application guidance documents to identify what’s missing and how to find or calculate the required information. Ad hoc bills of materials can be generated from model selections, grouped and formatted to required specifications.
Perhaps most significantly, AI can directly assist with modeling. It can search equipment catalogs, help select and place components, update properties individually or in bulk across selections, and even generate placeholder geometry for equipment not yet in the catalog. All of these actions are performed within the governed, model-based environment, where every action is traceable.
The AI in Substation+ isn’t a chatbot sitting beside the engineering tools. It’s an integrated copilot that speaks the language of substation design and operates within the constraints that make engineering output trustworthy.
From AI assistance to optimization in substation design
The current AI capabilities are powerful, but they point toward something even more transformative. Consider the possibilities: AI agents invoking proven simulation engines to iterate on thousands of design permutations, varying geometry, materials, cost, and equipment availability, then presenting qualified, superior alternatives for the engineer to review and approve. Imagine extending this intelligence into operations and maintenance, where continuously updated digital twins enriched with maintenance and operational data create environments so data-rich that the AI can surface insights no human team could find manually.
This shift is profound. Rather than measuring engineering productivity in work hours and inputs, the industry can begin measuring it in outcome quality. The engineer’s role evolves from manual execution to expert supervision, scoping problems, delegating to intelligent agents, and applying professional judgment to review and approve results.
This is not a distant vision. The building blocks, model-based design, structured data schemas, trusted applications, and domain-specific AI skills, are coming together now. But realizing the potential requires the industry to engage with these tools, learn how to work effectively with probabilistic technology, and help shape the workflows that will define the next era of substation engineering.
How AI improves substation design outcomes
Every decision in substation design ultimately serves the same goals: safe, economic, and reliable power delivery, with increasing emphasis on security, sustainability, and flexibility. AI supports these goals while changing what’s possible while pursuing them.
The organizations that move early to integrate purpose-built, trustworthy AI into their engineering workflows won’t just work faster, they’ll produce better outcomes. The result will be designs that are more thoroughly optimized, more rigorously checked, and more effectively carried forward into construction and operations.
Substation+ is next-generation substation design software, purpose-built for intelligent 3D design, real-time collaboration, AI-assisted workflows, and connected cloud delivery. Apply for early access to explore how concurrent design can accelerate your projects and improve coordination across teams.