Today, we’re excited to announce MerchantVision, SMB street view metadata powered by GPT-4 with Vision. This is the first time LLMs have been applied to automate site inspections.
As the latest addition to our MerchantProfiler product, MerchantVision offers risk teams business intelligence data on an SMB’s physical establishment, which they can match against an SMB’s application information. Automating this manual check saves risk teams critical time in the SMB underwriting process.
Read on to learn more, or check it out yourself.
When an SMB applies for a loan, merchant account, or another financial product, companies need to underwrite the SMB by conducting several credit and fraud checks.
If the SMB is a brick-and-mortar business, many companies will send a site inspector to the business address to confirm whether the physical address’s information matches data in the application. Sometimes, risk teams will skip physical inspections and digitally verify physical address information using a tool such as Google Maps’ Street View. Regardless of the method, this manual analysis is a significant time and resource investment for risk teams.
We heard several fintechs and software platforms complain about this process. Most of the SMB underwriting process has been digitized, so we couldn’t understand why physical address checks were stuck in the past. We decided to dig deeper.
As we’ve mentioned in the past, LLMs offer an opportunity to automate some of the most painful parts of risk management. SMB data is largely unstructured, and LLMs are extremely useful for extracting insights from unstructured data such as images.
When we heard about this site inspection pain point, we began investigating whether we could leverage Open AI’s GPT-4 with Vision to answer questions about physical address images.
How does it work? At a high level, our model pulls images from Google Maps Street View for a given address. Then, we leverage the GPT-4 with Vision model to analyze these images and answer a predetermined set of questions about the images. The responses to these questions are instantly available in the MerchantProfiler dashboard. This process can be done simultaneously for multiple businesses.
MerchantVision automates physical site inspections and provides risk teams with the key insights they need to assess business legitimacy, without visiting the business onsite. Some common analysis it automates includes:
Physical address analysis is a core component of SMB underwriting processes at many fintechs and software companies. It’s especially useful for companies serving businesses in field services, HVAC, plumbing, and related industries. Beyond underwriting, MerchantVision can also be used for periodic site checks during risk monitoring.
Coris is the first fintech risk company to leverage LLMs for enterprises at scale. Customers have already witnessed significant time savings with Merchant Real Industry, our GPT-4 powered industry classification model. MerchantVision will deliver similar efficiencies.
We’ll continue to automate SMB risk management using LLMs when necessary. If you’d like to learn more about MerchantVision – or if you have another risk use case ripe for automation – please reach out.