Commercial Real Estate (CRE) investment analysis today is widely conducted using a bottom-up approach on an individual property basis to assess the value and/or risk of subject properties—an approach that involves gathering street level information on the property, tenants, management, leverage, finances, submarket, and other pertinent data to form an opinion of prospective health and future performance.
More often than not, CRE debt market participants additionally employ a “model” which converts certain important property, submarket, and performance metrics into a quantitative answer that informs investment actions and/or accounting—a model which can then be adjusted to bring in factors not captured explicitly within it; qualitative adjustments informed by the “bottom-up” analysis of the analyst.
“Ai SPARK is pioneering the next step forward in credit analysis. Its AI based platform is built FOR and BY CMBS professionals and helps our team assess risk and adapt to changes in the macro environment faster than ever before. No other platform comes close."
Why AI?
The application of AI in CRE investment analysis cannot be described as “top-down”, at least not in the same way as a man-made credit risk model, because of the sheer amount of granular information that it considers. Neither can it be described as “bottom-up” because, after all, it is built using data.
It is in between the “bottom-up” and “top-down” approaches. It is “bottom-up” when set next to a man-made credit risk model, and “top-down” when set next to a CRE investment analyst’s qualitative approach.
The granular nature of AI makes it more dynamic and accurate relative to man-made credit risk models. Its accuracy is driven by exponentially more data than a regular model, and as such has the capacity to speed up the analyst, focus analytical efforts, and fill in with greater reliability than a regular model. This allows the analyst to get on with more in-depth analysis with confidence.
In addition, AI has a superior ability to draw inferences in data. It can mimic any model, decision tree, or mathematical function known to man, when/if it deems it necessary to do so. This ability gives it an inherent advantage over typical man-made models.
