Thought Leaders in Artificial Intelligence: Prashant Fuloria, Chief Product Officer at Fundbox (Part 3)

Sramana Mitra: Let’s double-click down on the AI. What signals are you looking for? What is the algorithm learning on the basis of?

Prashant Fuloria: We capture a number of data elements about a business. First and foremost is the data that we find in the small business graph. It’s important because the health of the business does depend on the health of the businesses around it and their interactions.

If I run a construction business, I may have a number of clients. The Fundbox small business graph captures the clients that I have as well as my invoicing and payment history with each of them. For example, I could be building out a new wing for a gym. This could be a two-year long project. Every month, I’m submitting invoices for work that has been done. There may be certain terms associated with it.

Frankly, the payment coming in from that gym does not have to correspond with those terms. Usually in B2B, payments are often late. Looking at the actual invoicing and payment history is an important part of looking at the small business graph. On the other hand, I might work with an accounting firm that closes my books every quarter. I may work with a marketing firm that helps me run a campaign.

By looking at the set of customers and vendors, we get some very valuable information. That’s one key element that we use for assessing the health of a business.

Sramana Mitra: This is not data that is sitting around in public domain. How are you getting that data?

Prashant Fuloria: When I come to Fundbox, I’m going to connect something to Fundbox. Let’s suppose I use QuickBooks. Fundbox is one of the premiere developers on the QuickBooks platform. When I connect my QuickBooks to my Fundbox, I’m giving Fundbox access to my accounting information that includes information about all my customers and all my vendors. It’s data like that that helps us build out this small business graph.

Sramana Mitra: You’re saying that you are looking in that graph to see who are the other vendors that they’re doing business with and what timeframe they are getting paid in. All of those signals you are picking up from there and you’re modeling it.

Prashant Fuloria: Exactly. We could be getting this information from bank account data. We pull a lot of data as well. We’re looking at a variety of unstructured data sources like your website presence and social media footprint. We pull structured data from payment exchanges like the small business financial exchange. There’s a variety of data sources that we pull. Our secret sauce is the ability to combine these datasets and run models with different levels of granularity based on how much data we have about a customer.

Coming out of this model is an assessment of the business health. One thing very important about Fundbox is that we do not have a traditional underwriting function. We have 140 people. Not one person in that team is a traditional underwriter in that their only job is to look at business applications and manually underwrite them.

We have credit analysts that look at the results of our models and over time, look at the health of our portfolio. A vast majority of our credit decisions are automated. There’s no human intervention between the point a customer comes onboard and connects their accounting software to the time they get a credit decision.

This segment is part 3 in the series : Thought Leaders in Artificial Intelligence: Prashant Fuloria, Chief Product Officer at Fundbox
1 2 3 4 5