Buildings as Data: Spatial Memory for Asset Tracking

Jordi Spranger presented at AI Engineer NYC 2026. Spranic, founded in June 2026, collaborates with Neo4j, a graph database company based in Malmö, Sweden, working with Neo4j collaborator Zam.

Spranger focused the presentation on the commercial segment, where the information gap creates the most acute financial risk.

This $400 trillion dwarfs global equities and the combined value of all debt and gold—making real estate Earth's single largest asset class.

Financial instruments—lending, insurance, mortgages—are bookkept and refreshed every quarter. The underlying real estate assets on which they rest update only every few years; a realistic interval between assessments is about five years.

Between the initial assessment and that five-year mark, building information relies entirely on pictures and PDFs—described as a black box with no current data on what has actually changed in the space.

This lag matters because physical conditions in buildings are not static; materials degrade, equipment breaks, and layouts change, yet the financial instruments remain locked to an increasingly stale baseline.

A study by Crawl examined 1,500 properties and discovered that outdated and insufficiently detailed records lead to under-specified insurance coverage—the gap between actual asset value and insured amount.

This systemic mismatch places both property owners and insurers at risk: owners face uncovered losses, and insurers unknowingly carry higher exposure than their premiums reflect.

The underinsurance problem and the lag in asset updates cascade through the entire commercial real estate financing system.

Because lenders and insurers rely on outdated snapshots of properties, the true risk profile of their portfolios remains unknown—a hidden fragility in the financial sector's real estate exposure.

A complete scan of a conference hall took just one minute of walking.

The system does not require specialized equipment or expert intervention; an ordinary person with a sensor can capture a comprehensive record of a building's contents and layout.

As personnel walk the space again—Monday next week, next quarter, or next year—the system aligns both walks on a time axis. The graph database retains all observations from each pass, preserving a complete temporal record.

This alignment enables the system to track not just what is in the building, but how it changes over time. Each subsequent walk captures deltas: items moved, equipment degraded, new assets added, and areas modified.

Graph databases are flexible in what information gets gathered and how it is connected, enabling the system to be hyperpersonalized to each asset and property. A retail property, a warehouse, and an office building each have different priorities; the graph adapts to capture what matters for each.

Evidence of fire extinguishers that have expired or failed safety inspections can immediately inform the insurer, who may then increase premiums based on the degraded condition of the property. Structural damage is similarly captured and flagged.

Rather than waiting for a five-year reassessment, insurers gain immediate visibility into property performance and can adjust their risk posture—or require the owner to remediate before renewing coverage.

The two scans were then aligned on a time axis, creating a timeline between the two periods.

Once stored in the graph, the map becomes an AI-ready asset. Users can query it in natural language: "Where can I heat myself?" returns the radiator location; "Is there a glass or mug to drink from?" points to kitchen glassware.

The spatial map transforms from a collection of pixels into a conversational interface, making building knowledge accessible without technical expertise or manual searching.

Comparing the two kitchen passes, the system detected that three chairs in the middle had been displaced. When queried about what changed, it identified the delta and provided a ground-source reference—the actual photograph from the capture.

This tie to original imagery constitutes proof of change; the system does not merely report a difference, it shows evidence. For compliance, insurance adjustment, or dispute resolution, that evidence is critical.

A single scan of a conference hall can serve multiple purposes: checking for needed repairs, scheduling maintenance, auditing compliance, processing insurance claims, planning interior design changes, and deploying robotic systems.

The spatial record functions as a digital twin of the building—a persistent, queryable model that any downstream process can reference and integrate into its own workflows.

This consolidation eliminates the fragmentation of knowledge scattered across photographs, PDFs, and separate systems, creating a single source of truth for the entire building's current state.

When a policy begins, a first scan establishes the initial state of the property. At renewal, a second scan reveals what has been maintained, repaired, or degraded. The insurer can now measure the property owner's upkeep performance against the policy terms.

Based on observed condition deltas, the insurance company generates a hyperpersonalized premium—one tied to actual property performance rather than broad risk categories or stale data. This precision reduces adverse selection and aligns premiums with true risk.

Properties that are well-maintained receive lower premiums; those showing neglect face higher rates or remediation demands before renewal—creating immediate financial incentive for owners to keep assets in good condition.

Currently, one in three commercial buildings on insurance books carry the wrong construction class, a classification error that distorts risk assessment and premium pricing.

Buildings certified as meeting specific standards can receive hurricane premium deductions of 35 to 60 percent. Seventy-three percent of insurers report significantly fewer hurricane claims when buildings have been checked and certified.

Certification can be automated with AI-powered spatial mapping, making it feasible at scale rather than requiring manual inspection by specialists. This automation is where Spranic's system creates value: it generates the evidence needed for certification without proportional cost.

That core requirement—present-day knowledge rather than five-year-old snapshots—is the fundamental problem spatial mapping addresses.

Real estate is the largest asset class on Earth, yet it operates with enormous inefficiencies and opacity. AI capabilities—particularly the ability to extract semantic meaning from spatial scans and track changes over time—can transform that opacity into transparency.

The result is not just better risk pricing and compliance, but a fundamentally hyper-customized market in which each asset's value, insurance, and management are tailored to its actual condition and performance rather than broad categories and outdated records.