How AI can improve automated processes in valuation

The valuation profession has been slow to adopt new technologies; perhaps with good reason.

This could include a lack of client confidence in outputs without a human sense check as well as the valuation industry trying to understand the risks and limitations of technology in what has been and remains a litigious sector.

But the pace of change is accelerating, with firms looking for efficient ways to produce valuation reports that are accurate, provided to clients quickly, and compliant with RICS professional standards.

Valuers benefit from much better availability and transparency of comparable transaction evidence than ever before, with the growth of data companies such as CoStar, EIG, Rightmove Plus and LonRes.

However, the data now available needs to be balanced with information sourced from local or specialist professional contacts, who retain valuable, confidential market intelligence.

Valuation firms have therefore been challenged to establish where automation using artificial intelligence (AI) can help their work without affecting the quality and rigour of the conclusions given to the client.

How can the automation of some parts of the process release the valuer to focus on the core aspects of their work?

Do we need valuers to analyse comparable evidence, assess the property's positives and negatives, understand the market and arrive at a rationale for the valuation with well thought-through conclusions if machine learning (ML) can improve at a fast enough pace?

This has driven the valuation profession towards greater use of service providers such as Valos and Edozo, which automate the sourcing of property-related data to include in report templates.

This allows valuers to devote more time to clients' increasing expectations for a comprehensive picture of the asset, including potential risks or opportunities such as energy performance certificates (EPCs) and lease events, as well as fulfilling the core role of advising on value after analysis of relevant comparable evidence and market trends.

Some may argue that this analysis to arrive at logical conclusions is what AI and in particular ML is made for. 

For the simplest of asset types, I would agree that AVMs can provide what a client may need; but their suitability and robustness depends on the number of variables that affect a particular asset, the client's attitude towards risk, and the tolerance they can accept when acting on a valuation conclusion.

The suitability and robustness of AVMs is also affected by the quality of the data being analysed, including its integrity, accuracy and standardisation, e.g. a building might be known as 1 High Street but also High Street House, so when you look at the business rates list, planning and EPCs, the same property can be referred to in multiple ways.

Data quality varies greatly across the multitude of data providers, including central and local government and quasi-governmental bodies, as well as publicly available and subscription data services.

Layering comparable data with information from interactions with other property professionals, who may have dealt directly with a significant transaction, isn't possible with AI data collation alone.

Similarly, a valuer's input in checking the veracity of the information AI has sourced is essential to the integrity of the outcome. 

There are frequently scenarios where legal titles aren't clear, property addresses aren't consistent, publicly recorded planning information is not correct or complete, and recorded EPCs aren't current, for example. 

These simple examples of possible data inaccuracy or inconsistency can materially affect the valuation advice provided and, with the quality of available data, a valuer's due diligence is essential to the process.

Although the availability of data has improved considerably, its integrity and quality remains a weak point in the accelerated adoption of AI in all but the simplest and lowest-risk valuation projects, and I don't see this changing in the short term.

A valuer's network of experienced property professionals is another essential and valuable resource, and something that is developed over time.

So much information can be secured only by speaking to peers in the market, and there is a risk that the relationship aspect of the profession, which is important in sharing information, won't be available in an AI-dominated profession. 

We need to educate the next generation of valuation professionals, so they do not become solely reliant on AI, but see it as a tool to make the process more efficient, supplementing their analytical and personal skills.

Reducing repetitive administration functions, minimising rekeying of information and standardising report production should enable the valuer to provide the robustness needed in the valuation process that ensures public confidence.

As AI in property valuation continues to develop, we need to consider how we refer to data sources, the extent to which the technology is used in the production of a valuation report and how the vendor ensures quality in any system that uses AI.

This should not currently need regulatory function. It is for the valuer and the valuation firm to ensure transparency in their processes for clients, adhering to RICS professional standards. However, as the use of AI grows, it is something that the organisation will have to offer guidance on and keep under review.

The increased use of technology, and AI in particular, means valuation is now less an art and more a science. While this promotes integrity and trust in the valuation profession, we need to ensure that the valuer can still paint a clear picture for the client.

AI is there to help the valuer; but for the vast majority of valuations it is not a viable replacement for a human professional – and is unlikely to offer that prospect for some time yet.