How will Embedded Finance harness Generative AI?

What is Embedded Finance?

To build a successful business in finance, you need three things (read: there are more, but it’s called the rule of three).

1. You need to find customers.
2. You need to educate those customers.
3. You need to underwrite those customers.

Embedded finance, the latest update to fintech pitch decks and go-to-market powwows, has been a key driver in addressing these core pillars. This is especially the case in tackling items one and three – source and underwrite.

When you place financial tools where the customer lives, say within their ERP, CRM or POS (we call them “Digital HQs”), it makes finding those customers a whole lot easier – if someone is about to buy inventory, let’s offer them an inventory finance loan on the spot. Similarly, when that financial tool lives within the Digital HQ, it becomes a lot easier to underwrite them – if we can seamlessly access and analyze every customer invoice and inventory level over the last twelve months, we will be more informed about their credit risk.

🚨 The underwriting data discussed above, we’ll call Digital HQ data, will show up later.

Enter current day embedded finance: the challenge remains in education. Existing solutions, though dripping with benefits (read: hype and attention of whatever VC capital is out there), can struggle as a medium for product discovery and education. If I don’t understand the benefits of borrowing for my inventory, I’ll just pay with cash thank you very much. The current market alternatives aren’t great, as embedded fintechs must sort between:

1. limiting to one product, usually an expensive MCA product,
2. charging the Digital HQ a monthly fee for a sales team to support customers, or
3. relying on good design and video features to guide the customer across a few products.

The biggest variable that impacts the success of an embedded finance program is the customers ability to understand the benefits of the products. They need to confidently step beyond “finance is now available” to “finance is now available and this is how it will help me.” Retailers that use our PayLater product, for example, end up seeing 4-6% gross margin expansion. Distributors that use our FundNow product, will markedly improve cash conversion cycle which aids in growth. These tools run the risk of remaining untapped without the right coaching.

Enter Generative AI

The power of GenAI, of course, sprinkled with plenty of caveats and what-ifs, feels transformational. ChatGPT, a popular GenAI tool, has opened 100mm+ people to a new outlet for the “why don’t you Google thats,” in addition to successfully scaring a few journalists. When used as intended, the seemingly human responses can offer a short explanation to even your wildest questions (read: turns out ChatGPT thinks “we should see other people” @MargotRobbie).

One of the key benefits to GenAI, as flaunted on ChatGPTs homepage, is the ability to “[remember] what the user said earlier in the conversation.” We will refer to this as “context priming” – broadly defined as any information that informs the GenAI’s next response. Context priming is important as, without it, there can be some major misses in responses. “Mac McClung is on fire!” holds two completely different meanings, for example, without context about the date or venue.

When proper context is provided, and Mac has collected his Slam Dunk trophy, the GenAI tools become even more interesting. Imagine the prompt has enough relevant background to offer personalized advice – once primed, ChatGPT is quick to remind Mac that, though his victory is impressive, he should be leveraging this exposure into a longer term career play.

When the GenAi is primed with valuable context about a business, we (read: embedded finance folks) can finally unlock door #2 – educating the customer. Fortunately, as discussed earlier, some embedded finance solutions have unique access to Digital HQ data. As we’ll see below, this Digital HQ data is perfect for context priming.

Let’s go to ChatGPT to demonstrate:

For starters, we ask ChatGPT for advice on working capital. We do not provide any context and, as a result, it sort-of misses – kindly suggesting we sell more or spend less (thanks so much!) before offering a few helpful hints.

Next, we prime with a couple basic details. Data points such as revenue and accounts receivable are easily pulled from our Digital HQ data. As we see, the response starts to take form.

Now we’re really cooking! The GenAI understands we have an asset in A/R and makes a suggestion to borrow against that. Once again, we get a nudge to sell more – the training data must have overweighted the Jack Welch catalog.

Finally, we give ChatGPT a layup. We prime with specific context – again, all easily derived from our Digital HQ data – and the GenAI provides a detailed, specific suggestion.

The demonstration, a small drop in the well of potential, can help understand how a large dataset of context can help prime a GenAI to offer personalized advice. The benefit of embedding this tool, versus, say, a standalone “help app,” is the trove of Digital HQ data available to easily prime the results. A small business that needs help understanding their financial options can get much faster, relevant responses when sharing the right snapshot of their business profile.

As the world of embedded finance grows from thought-bubble to norm, it is likely that customer education becomes more of a sticking point. By leveraging powerful GenAI tools readily available, and combining the treasure chest of Digital HQ data, embedded finance solutions can permanently change the landscape of the financial services industry.