What machine learning means for real estate

Machine learning (ML) – a particular form of artificial intelligence (AI) – is already affecting all areas of our lives, often without us even realising.

Construction and real estate are also increasingly using digital tools to reduce costs, optimise workflows, achieve business objectives and become more competitive.

The ultimate goal is establishing a common data environment (CDE) for every asset that is accessible from different systems – and even between organisations.

The BSI defines a CDE as 'a single source of information for any given project or asset, used to collect, manage and disseminate all relevant approved files, documents and data for multidisciplinary teams in a managed process'.

Although real estate faces particular challenges due to the wide variety of documents, data and actors, digitalisation has already found its way into many organisations.

Documents are increasingly being digitised, digital twins can be created even for existing buildings, and operating data is being recorded by sensors at short intervals or even in real time. This makes much more data accessible to machines and digital processing.

The property technology or proptech industry is enabling and encouraging this transformation by optimising workflows and allowing the development of new business models using digital approaches and evolving business intelligence (BI) platforms.

In addition, international initiatives such as the Global Alliance for Buildings and Construction have made it their primary goal to standardise data and to close information gaps to improve energy efficiency in the building sector. 

But the surveying profession is still particularly reliant on building documentation and manual efforts for collecting information. Paper remains widely used in real estate, not only in archives but also in day-to-day business. 

It is important, though, not only to have access to digital or digitised documents from any workstation, but also to make them useable in future digital processes.

ML requires that documents be prepared in such a way that they can be read by machines. In order to do so with a scanned document, we must first extract its words and characters using optical character recognition (OCR).

ML algorithms are already in use, and others being developed, to classify documents, automatically extract key information, and transfer it to document management systems (DMS) and BI platforms.

A recent report sponsored by the Property Research Trust presents the key requirements for digitising analogue building documentation to support digital workflows and applications. It outlines requirements for digital document archives, scanning rules and digital archiving workflows. 

In particular, the report maintains, documentation governance must minimise the need for human intervention. Data managers should be installed at scale and take responsibility for data quality and content, supported by AI.

Ideally, digital twins should also be shared in transactions and between owners and service providers, including surveyors, throughout the building lifecycle to maintain a high level of information and avoid recurrent data collections.

A key requirement is multi-tenancy, meaning that data is only accessible to stakeholders for whom it is relevant and who are authorised. Eventually, this information could be shared using blockchain technology as a further means to ensure data integrity.

To some extent, digital workflows will also support automated valuation models (AVMs), although these are still considered controversial in the surveying profession.

Valuations, especially for larger and more complex properties, require a high degree of expertise, responsibility and accuracy.

Nevertheless, ML can and will certainly help in this task. Simple valuation cases or even mass appraisals are likely to be handled by AVMs, at least initially. 

The final decision and liability, however, must remain with the human professional and the surveying company, unless the client agrees to waive its right to make claims in exchange for a low-cost or expedited appraisal.

A major drawback of AI is that the way it produces results remains obscure – 'black box' – pretty much like a human brain without the possibility of questioning the actual cognitive process.

Ultimately, the users of AVMs, not the software developers, will remain responsible for the results they sell to their clients. Similar to the case of autonomous vehicles (among others), there will be an ongoing debate about this.

Offshore firms could offer poor internet AVM valuations to building owners looking to minimise costs. The surveying profession will have to demarcate these to avoid damage.

Those who adapt best will succeed. Surveying firms could even become involved in processing data and documents as data managers for specific subsets of data such as building and maintenance documentation, or even for digital twins and a CDE.

In this respect, they should partner with proptech or engineering firms that provide digital facilities, but which are not usually capable of or even suited to becoming data managers themselves.

Cross-border document or data management remains a major challenge, given the diversity of documentation and the continuing lack of common international data processing standards. Those who can manage this, however, will inevitably grow their businesses.

As data and information become more accessible, its quality and services such as valuations that use it will become key to competitive advantage in real estate. 

Skilled people with a sense of responsibility, an entrepreneurial spirit, leadership skills and the ability to make informed decisions should remain the foundation for this, now and in the future.