[0] Diving Back Into the Foundations: How to Build the Future of Agentic Digital Twins
A position paper for practitioners in Information Management, BIM, and Digital Twins.
Building truly intelligent digital twins requires moving beyond static 3D models and workflow automation to create semantically structured, machine-interpretable building logic—combining EXPRESS/STEP standards with modern AI agents. Rather than choosing between cutting-edge LLM capabilities and foundational data standards, the future depends on mastering both to enable buildings that actively reason, optimize, and coordinate operations throughout their lifecycle.
Read as:
Over the past few weeks, I have been working on a fundamental question: how do we move beyond workflow automation toward buildings that can truly think, adapt, and optimise themselves? Why do "digital twins" still behave like static models?
The answer isn't in the latest plugin or another automation script. It's in going back to the foundations. You sometimes need to go back in order to move forward.
Over the past few weeks, I've been focused on a fundamental challenge: how do we move beyond simple task automation toward buildings that can actively monitor themselves, respond to changing conditions, and continuously improve their own operations? Why do so-called "digital twins"—digital representations of physical buildings—still function like static snapshots rather than living systems that learn and adapt?
The answer isn't found in the latest software tool or automation script. It's in returning to core principles. Sometimes you need to understand the fundamentals before you can innovate effectively.
For professionals in construction and real estate: Think of this as the difference between having a floor plan and having a building that knows when something needs maintenance before it breaks.
## Executive Brief: From Static Models to Intelligent Buildings
The industry has invested heavily in digital twins, yet most remain passive 3D representations rather than operational intelligence systems. The opportunity lies not in incremental automation improvements, but in fundamentally rethinking how buildings process information and respond to changing conditions.
**Strategic Implication:** Organizations that transition from static asset documentation to dynamic, decision-making systems will capture significant competitive advantages in operational efficiency, risk reduction, and capital optimization.
[1] Why EXPRESS Matters for AI-Driven Buildings
To build genuine digital twins, not just 3D models, but operational intelligence layers, I've found myself studying ISO 10303-11 and the EXPRESS language that underpins STEP and much of the CAD world.
EXPRESS represents one of the most rigorous attempts at describing geometry, topology, and product data in a machine-interpretable format. For those working in BIM and ISO 19650 Information Management, this might seem like academic archaeology. But here's the practical implication: if we want LLMs and agentic systems to reason about buildings, spaces, and operations, they need structured, semantic data they can actually parse and act upon.
To create genuine digital twins—real operational intelligence systems that can guide decision-making in buildings—I've been studying ISO 10303-11 and EXPRESS, a formal data language that powers STEP (Standard for the Exchange of Product Data) and much of the CAD industry.
EXPRESS is one of the most comprehensive attempts at creating a machine-readable language for describing physical properties: geometry (3D shapes), topology (how parts connect), and product data in a way computers can understand and act on. If you work with BIM (Building Information Modeling) or follow ISO 19650 standards for information management, this might feel like going back to basics. But here's why it matters practically: if we want AI systems and autonomous robots to make decisions about buildings, spaces, and operations, they need structured data that clearly defines relationships and constraints—not just descriptions in natural language.
For construction teams: This is the difference between telling an AI "repair the HVAC system" and giving it a precise, machine-readable model that shows exactly where every component is, how it connects, what its specifications are, and how it relates to the rest of the building.
## Semantic Foundation: The Missing Element in Digital Transformation
Current BIM implementations excel at geometric representation but lack the semantic structure needed for AI-driven decision-making. Rigorous data standards—developed decades ago for precision manufacturing—provide the machine-readable frameworks that enable AI systems to understand not just what a building looks like, but how it functions and why design decisions matter.
**Business Impact:** Without semantic data structure, AI and robotic systems remain limited to surface-level automation. Proper semantic architecture unlocks genuine operational intelligence and autonomous optimization across assets.
[2] From CES 2026 to the Construction Site
At CES 2026, the industry celebrated robots arriving on factory floors. I'm focused on the orchestration layer that makes them effective.
Picture a building-scale digital twin functioning as a coordination hub, what I call the "queen bee" model:
For non-technical stakeholders: Think of it as mission control for your building. Just as air traffic control coordinates planes without flying them, this system coordinates robots, monitors operations, and simulates changes before they happen physically. The value? Reduced risk, optimized operations, and predictable outcomes.
For technical teams: We're talking about a 3D operational environment where:
- AI agents train and validate in simulation before physical deployment
- Agentic workflows understand geometric constraints, spatial relationships, and design intent — not just natural language prompts
- Robots and human workers operate in a shared environment with real-time digital twin feedback
- Buildings become active participants in their own maintenance, optimization, and adaptation
At CES 2026, the industry focused on robots entering production facilities. I'm more interested in the coordination layer that makes those robots actually useful.
Imagine a building-scale digital twin operating as a central coordination hub—what I call the "mission control" model:
**For facility managers and stakeholders:** Think of it as air traffic control for your building. Just as controllers coordinate hundreds of planes without flying them, this system orchestrates robots and maintenance teams, monitors what's happening in real time, and tests changes digitally before implementing them physically. The payoff? Lower risk, better-running operations, and predictable outcomes.
**For technical teams:** This is a 3D operational environment where:
- AI agents practice and validate their decisions in digital simulation before touching anything physical
- Autonomous workflows understand spatial constraints and design logic—not just responding to text commands
- Robots and human workers share the same space, with continuous real-time feedback from the digital model
- Buildings become active participants in maintaining themselves, optimizing energy use, and adapting to changing needs
## Mission Control for Buildings: The Orchestration Opportunity
The next-generation digital twin operates as a coordinated control center—not replacing humans or robots, but orchestrating them effectively. This "queen bee" model:
- **Validates operations before physical deployment** – reduces risk, prevents costly mistakes
- **Simulates scenarios** – enables predictive decision-making rather than reactive management
- **Coordinates distributed systems** – integrates robots, IoT, staff, and third-party services seamlessly
- **Enables continuous adaptation** – buildings respond to changing conditions and requirements
**ROI Focus:** Predictable operations, reduced downtime, optimized resource allocation, and proof-of-value before full-scale deployment.
[3] The Missing Link: Deep Interoperability
The gap between current BIM workflows and true agentic systems isn't technological capability — it's semantic structure. We have:
- Powerful LLMs that understand natural language
- Robotic systems ready for deployment
- BIM models with geometric data
What we're missing is the connective tissue: machine-navigable building logic that bridges design intent, geometric reality, and operational requirements.
This requires mastering both the cutting edge (LLMs, agentic frameworks, simulation environments) and the fundamentals (EXPRESS schemas, STEP, robust data modeling). One without the other leaves us with either theoretical possibilities or isolated automation islands.
The real barrier between today's BIM systems and tomorrow's intelligent, autonomous buildings isn't raw technology—it's structured data that machines can understand and reason about. We currently have:
- Advanced AI language models that understand human communication
- Robotic systems ready to be deployed
- BIM models containing geometric and spatial data
What's missing is the translation layer: a machine-readable definition of building logic that connects design decisions, physical geometry, and operational needs.
Building this requires mastery of both modern AI (language models, autonomous frameworks, digital simulation) and foundational data standards (EXPRESS schemas, STEP data formats, robust information modeling). Without the modern tools, you have theoretical frameworks that can't operate in the real world. Without the foundations, you build isolated automation projects that don't scale or integrate.
For construction professionals: You need both the latest AI capabilities AND the ability to structure building data so those systems understand what they're working with.
## The Integration Gap: Why Current Approaches Fall Short
The industry possesses individual pieces—advanced AI, robotics, rich BIM models—but lacks the connective intelligence layer. True agentic systems require understanding design intent, geometric constraints, and operational logic simultaneously.
**Strategic Challenge:** Organizations deploying AI or automation without addressing semantic data architecture face islands of disconnected systems and missed optimization opportunities. Mastering both foundational data standards and emerging technologies is essential for competitive advantage.
[4] What This Means for AECO
For the industry, this represents a shift from:
- Document management to Knowledge systems
- Model coordination to Operational intelligence
- Workflow automation to Autonomous optimisation
- Static deliverables to Living digital assets
The buildings we design today will operate for 50+ years. The digital infrastructure we build now will determine whether they remain static structures or evolve into adaptive, intelligent environments.
For architecture, construction, and real estate, this represents a fundamental shift from:
- Managing documents to Building knowledge systems
- Coordinating separate models to Creating operational intelligence
- Automating specific tasks to Enabling autonomous optimization
- Delivering static files to Creating living, evolving digital assets
The buildings designed and constructed today will operate for 50+ years. The digital systems we build now will determine whether those buildings remain static, fixed structures or evolve into intelligent environments that continuously adapt and improve.
## Industry Transformation: From Management to Intelligence
This shift redefines how we approach the built environment across its lifecycle:
| Traditional Approach | Intelligent Infrastructure |
|---|---|
| Document management | Knowledge and decision systems |
| Coordinating static models | Real-time operational intelligence |
| Isolated automation scripts | Continuous autonomous optimization |
| Project deliverables | Living, evolving digital assets |
**50-Year Horizon:** Buildings operational today will operate for decades. The digital infrastructure choices made now determine whether assets remain static or evolve into responsive, self-optimizing environments.
[5] Looking Forward
This is the first entry in a series documenting the technical journey from BIM automation to agentic digital twins. I'll be exploring:
- Practical implementations of EXPRESS and STEP in modern workflows
- Agentic framework architectures for built environment applications
- Simulation-to-reality pipelines for construction and facilities management
The future of AEC won't be built by ignoring our industry's foundations — it will be built by understanding them deeply enough to transcend them.
I don't claim to have all the answers yet — much of this is experimental and involves breaking things in a sandbox before they ever touch a real site. But I'm documenting the process of digging into these schemas and more, to see if they can finally bridge the gap to true agentic behaviour.
This is the first in a series exploring the technical path from today's BIM automation toward buildings with true agentic intelligence. The upcoming pieces will examine:
- Practical ways to use EXPRESS and STEP standards in real construction and facility workflows
- How to design AI agent systems specifically for buildings and infrastructure
- Creating digital-to-physical pipelines: testing changes in simulation before deploying them on actual sites
The future of AEC—architecture, engineering, and construction—won't come from abandoning the field's established standards and practices. It will emerge from understanding those foundations so completely that you can transcend their current limitations.
I don't claim to have this all figured out yet—much of this work is experimental, tested safely in digital environments before touching real projects. But I'm documenting the process of diving deep into these data standards to see whether they can finally enable buildings that truly think and act autonomously.
For construction and real estate professionals: This series is about bridging the gap between the structured data and processes you already use and the autonomous, adaptive buildings we're beginning to make possible.
## Strategic Roadmap: Building the Next Generation
The path forward requires systematic integration of foundational technologies with emerging agentic frameworks:
- Semantic data architecture enabling AI reasoning about space, design, and operations
- Scalable agentic frameworks purpose-built for AEC workflows
- Simulation-based validation reducing deployment risk and capital exposure
**Leadership Focus:** Organizations that systematically master this integration—combining rigorous data foundations with cutting-edge AI—will define the competitive landscape for the next decade of real estate, construction, and facilities management.
Key Takeaways
- Digital twins need structured, semantic data (EXPRESS/STEP schemas) for AI agents to reason about buildings, not just 3D models or automation scripts.
- The "queen bee" model treats building-scale digital twins as coordination hubs where AI agents validate in simulation before physical deployment, reducing operational risk.
- The missing link between BIM workflows and agentic systems is machine-navigable building logic that connects design intent, geometric reality, and operational requirements.
- True agentic digital twins require mastery of both cutting-edge AI (LLMs, agentic frameworks) and foundational standards (EXPRESS, STEP, data modeling).
- The AECO industry must shift from static document/model management to living digital assets that actively optimize and adapt operations over 50+ year building lifecycles.