Keywords: BIM, digital twin, knowledge graph, ISO 19650, openBIM, IFC, material passports, circularity, building lifecycle information management
1. Introduction: A Building That Knows Its Own Story
Walk into most buildings today and ask a simple question — why is this beam here, and not fifty centimetres to the left? — and the honest answer is usually: nobody knows anymore. The architect who made the call retired. The structural calculation lives in a folder on a server that was migrated twice. The site engineer who found a soil condition that forced a redesign never wrote it down anywhere the next owner could find. The building stands, but its memory doesn’t.
Building Information Modelling solved a real problem: it gave the industry a shared 3D model instead of a stack of disconnected drawings. But a model is not a memory. It tells you what a building is — geometry, quantities, systems — far better than it tells you why it is that way, who decided, what they assumed, and what happened after occupancy that proved them right or wrong. That gap is becoming impossible to ignore as buildings are asked to last longer, adapt more, disclose more about their materials, and justify every design decision to regulators, insurers, and future generations of owners.
This piece looks at what comes after BIM and the conventional digital twin: not a better 3D model, but a structured, living record of a building’s decisions, materials, performance, and lessons — queryable across its entire life, and increasingly legible to AI agents working on the project’s behalf. It’s a concept this piece calls building memory, and it’s already implicit in where the industry’s own standards are heading in 2026.
2. From Drawings to Models: A Short History of Losing Context
2.1 The paper era and its blind spots
For most of the twentieth century, a building’s official record was a set of drawings plus a paper trail of specifications, change orders, and correspondence — usually scattered across the offices of the architect, the engineer, the contractor, and the owner, with no single party holding the whole picture. Once construction finished, that trail rarely followed the building into operation.
2.2 BIM: geometry gets structured, but decisions don’t
BIM, maturing through the 2000s and 2010s, changed the units of exchange from drawings to structured 3D objects carrying properties, quantities, and relationships. It was a genuine leap — clash detection, quantity take-off, and coordinated multidisciplinary models became normal practice. But BIM’s default unit of memory is the object, not the decision. A wall knows its material and thickness; it does not natively know that its position was fixed by a neighbour’s right-to-light claim, or that its insulation spec was tightened after an early energy model flagged a compliance risk.
2.3 The digital twin: real-time, but often shallow
Digital twins added a further layer — live sensor feeds mirrored against the model, useful for facilities management and operational optimisation. Yet in most implementations today, the twin is a live dashboard bolted onto a static geometric model. It can tell you that a zone is running hot; it typically cannot tell you that the same zone was value-engineered out of its original shading system three years earlier, and that this is why.
3. The Industry Is Already Moving Toward “Information,” Not Just “Models”
This shift isn’t speculative — it is happening in the standards right now. Draft revisions to ISO 19650 Parts 1 and 2 were opened for public consultation in March 2026, with Part 3 following in June and formal publication expected in 2027. The direction of travel is unmistakable: the standard is moving away from “BIM” as a modelling exercise and toward information management as a continuous lifecycle discipline, replacing the old split between the delivery phase (Part 2) and the operational phase (Part 3) with a single, unified process that runs from concept through to decommissioning. The project’s Building Execution Plan is being reframed as an “Information Production Plan” — a small naming change that signals a large one: the deliverable was never really the model, it was always meant to be the information.
At the same time, national and regional regulation is pushing hard on material-level traceability. The EU’s Construction Products Regulation entered into force in January 2025 and began phasing in through January 2026, introducing a Digital Product Passport requirement for construction products alongside textiles, batteries, and electronics — mandating that composition, origin, environmental impact, and compliance data travel with a product rather than disappearing into a supplier’s private records. Full rollout across all EU member states is expected to run from 2026 to 2030.
Put these two threads together — lifecycle information management at the building level, and digital product passports at the material level — and the outline of “building memory” becomes visible: not a single new tool, but the convergence of governance frameworks that were, until recently, developed in separate silos.
4. What “Building Memory” Actually Means
A living knowledge system for a building would retain, in structured and queryable form, at minimum:
- Design intent and assumptions — not just the final wall position, but the constraint that produced it, and the load case, occupancy assumption, or code clause behind every calculation.
- Decision traceability — a chain from requirement → option considered → decision → approval → the person or role who signed off, so a future renovation team can see not just what was built but what was rejected, and why.
- Material provenance and passports — origin, composition, embodied carbon, and disassembly/reuse potential for the actual materials installed, not a generic product-family average.
- Sensor and performance history — how the building has actually behaved against the model’s predictions, feeding discrepancies back rather than letting them evaporate into a facilities log nobody reads.
- Maintenance and intervention history — every repair, replacement, and retrofit, linked back to the original component and its passport, so the record compounds instead of resetting with every ownership change.
- Lessons and rationale for deviation — the informal knowledge that normally lives only in the heads of the people who were on site: why a detail was changed on day 40 of construction, and what it prevented.
None of this requires abandoning BIM or IFC-based digital twins — it requires treating them as one layer in a larger knowledge graph, where geometric objects, documents, sensor streams, approvals, and material passports are linked entities rather than files sitting in adjacent but disconnected systems.
5. Traceability of Engineering Decisions: The Missing Layer
Engineering already produces enormous amounts of traceable reasoning — calculation packages, design checks, peer reviews, code compliance justifications. The problem is almost never that the reasoning doesn’t exist; it’s that it’s captured in formats (PDFs, email threads, meeting minutes) that are disconnected from the model objects they justify, and that get orphaned the moment a project’s file structure is archived.
A knowledge-graph approach to decision traceability links each significant model object or system to the requirement it satisfies, the calculation that sized it, the standard clause it complies with, and the approval that signed it off — queryable later by a human or an AI agent asking “show me every structural element whose design load assumption depended on a since-superseded wind code.” That query is trivial for a well-structured graph and essentially impossible for a folder of PDFs.
6. Material Passports and Circularity: Memory as a Precondition for Reuse
Circular construction depends on knowing, at the point of deconstruction, exactly what a material is, where it came from, and what condition it is in — information that is nearly always lost by the time a building is thirty or forty years old. Material and building passports are increasingly framed by researchers and EU policy alike as the connective tissue that makes reuse and recycling economically viable rather than aspirational: without reliable provenance data, materials recovered from demolition default to low-value waste, because nobody can verify their composition or structural history well enough to trust them in a new application.
This is where building memory and circularity converge directly. A material passport is only as useful as the record that keeps it attached to the physical material and updates it across decades of ownership changes, retrofits, and repairs — which is precisely the continuity problem that document-based BIM has never solved and that a persistent, structured knowledge layer is designed to.
7. Operational Feedback Into Design: Closing the Loop
Buildings currently function as one-way information sinks: design assumptions go in, and almost nothing about how those assumptions performed comes back to inform the next project. A building memory system inverts this. Actual energy performance, occupant comfort complaints, maintenance frequency, and component failure patterns become structured data linked back to the original design decisions and material specifications that produced them — creating a feedback loop that lets an architecture or engineering practice learn, across its whole portfolio, which assumptions consistently hold up and which consistently don’t.
8. AI Agents Working on Structured Project Knowledge
This is the layer that makes the previous six sections practically achievable rather than an administrative burden. Unstructured documents are hard for both humans and AI systems to reason over reliably; a structured, linked knowledge graph is exactly the substrate AI agents are best at working with. An agent with access to a well-governed building knowledge graph could, in principle:
- Flag that a proposed retrofit would conflict with an assumption baked into the original structural design, before the conflict becomes a site problem.
- Assemble a compliant material passport data set automatically at handover, pulling verified data from supplier passports rather than requiring manual re-entry.
- Answer a facilities manager’s plain-language question about a system’s maintenance history and the reasoning behind its original selection, in seconds rather than through an email chain to someone who may no longer work there.
- Surface portfolio-wide patterns — which assumptions or details recur as later problems — that no individual project team would ever see from inside a single job.
None of this works on top of loose PDFs and disconnected spreadsheets. It requires the underlying data to be structured, linked, and governed — which is exactly what the current wave of standards is quietly building toward.
9. The Governance Layer: ISO 19650, openBIM, IFC
None of this is credible without open, vendor-neutral governance, for the same reason a building’s structural record is worthless if it’s locked in a proprietary format that outlives the software that wrote it. openBIM, built around the IFC (Industry Foundation Classes) schema, exists precisely to keep building data legible independent of any single vendor’s tools — a precondition for information that has to remain readable in fifty years, long after today’s software has been replaced several times over. The 2026 ISO 19650 revision reinforces this by explicitly extending its scope across the full asset lifecycle rather than stopping at handover, and by tying information requirements to defined roles and named accountability at every stage — the same chain of custody that decision traceability and material passports both depend on.
The practical implication for firms is that adopting “building memory” isn’t a separate technology initiative bolted on top of existing BIM practice. It’s largely a governance and data-modelling discipline: deciding, from the earliest project stages, what information is captured as structured, linked data rather than as a document to be filed and eventually forgotten — and building the habit of asking facilities managers and future operators what they’ll need to know, before design decisions are locked in.
10. Conclusion: The Building as a Living Knowledge System
The buildings of the next several decades will be asked to do more than stand: disclose their carbon, prove their compliance, justify their materials, adapt to uses nobody anticipated at design time, and eventually give up their components for reuse rather than landfill. Every one of those demands is really a demand for memory — for a building to carry, in a form that outlives any single project team or software platform, a structured account of what it is, why it is that way, and what has happened to it since.
The direction of the standards, the regulation, and the underlying technology all point the same way at once: not toward a better model, but toward buildings that remember. That is the frame worth building toward now — not as a hypothetical future state, but as the natural extension of information management work that is already, quietly, underway in 2026.
Sources consulted
- ISO 19650 2026 revision coverage (Graitec, REBIM, BARVEA, apiBIM, Coulter BIM Information Management, excelize.com)
- EU Construction Products Regulation and Digital Product Passport analysis (npj Materials Sustainability / Nature; BUILD UP, European Commission; CircularPass)
- Academic literature on material passports, digital twins, and circular construction (ScienceDirect, MDPI Sustainability, ASCE Journal of Construction Engineering and Management, Springer Nature)