Spranic turns repeated walks through a building into a time-linked record of its objects, relationships, and changes.
2
A graph representation lets people query a captured space for objects such as radiators, glasses, and mugs, with references back to the underlying observation.
3
Spranger proposes using the same record for maintenance, compliance, interior design, robotics, and more specific insurance policies.
Summary
Jordi Spranger argues that financial records for property are updated much more often than the buildings those records describe. A building may be represented by old photographs, PDFs, and occasional site visits, even while mortgages and insurance records are reviewed regularly. Spranic's proposed answer is to turn each walk through a property into a spatial memory. An off-the-shelf sensor maps the space, identifies objects, and stores them in a graph database. Later walks remain linked to earlier ones on a timeline, allowing object-by-object comparison and references to the original capture. Spranger demonstrates queries for a radiator and drinking vessels, then shows that three chairs moved between two kitchen scans. He connects this record to maintenance, compliance, interior design, robotics, and insurance workflows. The talk demonstrates scene capture, querying, and change detection. The insurance examples are proposed workflows rather than evidence of insurer deployments or measured premium savings.
Property finance is updated more often than the buildings behind it
Spranger says lending, insurance, and mortgage records are updated every quarter, while the underlying real estate may be surveyed only every few years. A building can go from the creation of a financial instrument to a later modification without a current record of its state. During that period, the available evidence may be a collection of photographs, PDFs, and notes. He connects this gap to Kroll research involving 1,500 properties, which he says found that 70% of buildings were underinsured. His explanation is that old and insufficiently detailed records can lead to insurance that does not match the property.
A walk through a property can become a structured building record
Spranger describes using an off-the-shelf sensor while walking through a space. The scan maps the property, recognizes assets, and stores them in a graph database. He shows a conference hall scan and says the walk itself took place by moving through the hall, with the scan available a minute later. A later walk is retained alongside the earlier one. The walks are aligned on a time axis, so the record contains the space, the objects in it, and how the space changes across observations.
The record must explain changes, timing, and responsibility
Spranger says a useful building record should answer three operational questions. It should show what changed by comparing objects and their surroundings across two walks. It should show when a change occurred, with every observation carrying information about its source. It should also identify who needs to know about a detected difference. He gives structural damage and expired fire extinguishers as examples of findings that could trigger action from property teams or insurers.
A semantic map makes ordinary spatial questions queryable
In the kitchen example, two captures of the same space are recognized as the same space and aligned across time. Spranger then queries the map in ordinary language. The question "Where can I heat myself?" points to a radiator. A question about a glass or mug in the kitchen points to the relevant objects. The point is to turn a scan from a collection of pixels into a representation that people and software can query.
Change detection can preserve evidence for a specific observation
Spranger compares two passes through the kitchen and asks for the difference between them. The system identifies three chairs in the middle that were displaced. The result is linked back to the original capture, giving the user a source reference for the detected change. This connects a statement about the scene to the observation from which it was produced.
One spatial record can feed several property workflows
Spranger describes the building record as an AI-ready asset that can connect to downstream work. The conference scan could support maintenance checks, compliance checks, insurance, interior design changes, or planning for robots operating in the space. The same captured objects and relationships can therefore be reused instead of being left across separate documents and images.
Insurance could compare an initial inspection with the renewal state
For insurance, Spranger proposes capturing the building when a policy starts and capturing it again at renewal. The comparison could show changes and the upkeep of the property, giving an insurer more specific evidence for the policy. He says the insurer could pay for what it can verify and adjust coverage to the property's observed condition. He also cites figures about incorrect construction classes, hurricane-related deductions, and insurer views on certification, while presenting these as the business case around the idea rather than results from a Spranic deployment.
