Preparing Fintechs for Intelligence in the Middle

Several folks on LinkedIn asked me if I could shed more light on the section of my blog on middleware selection that discussed Intelligence in the Middle. In this blog, I’ll expand on that concept in more detail including what fintechs can do today to prepare for what’s coming.

Overview

The financial services industry is slowly but steadily embracing artificial intelligence (AI) technologies. So far, we’ve seen and heard a lot about AI agents, primarily to replace human resources in customer facing functions. Soon AI agents will also play a critical role in financial services enterprise integration. This is especially relevant for fintechs as AI will dramatically increase their ability to grow and scale.?

Fintechs and APIs

Today most integration between fintechs and their customers is handled by APIs (application programming interfaces).

Typically, this involves communication between fintech APIs and those of their customers’ backend systems, especially the banking core.

APIs have been around for decades and are the most common technology used by enterprise applications to communicate with each other. While APIs are a proven and robust technology, their structure and scope of coverage will soon undergo changes.?

?

From API to AII

Fintech connectivity is evolving to enhance APIs with what we might call AII (artificial intelligence interaction). Effectively this is a brand-new communication pattern, made possible by the rapid advancement in LLM (Large Language Model) technology over the last two years. The resulting change places new demands on APIs and poses additional challenges for fintechs.

Conventional API-only Pattern

Integration between systems was originally based on static route API invocation between systems and response from the API call was accepted as is.??

APIs with Rules Engine Added

Over time, many enterprise systems added rules engines that evaluated the API response and retried the API call as needed. While this approach proved more robust, it also added complexity to the underlying system.

?

AII with a Single System

The emerging AII approach, fronts APIs with AI agents that provide intelligence in the middle. An AI agent can iterate through API calls with a downstream system until it receives a response that meets the rule criteria.?

?

AII with Multiple Concurrent Systems

If required, the AI Agent can also make multiple concurrent calls to different systems and select the best response to accept on behalf of the originating system.??

Preparing APIs for an AII Future

Fintechs can perform simple optimizations to enable their APIs to support AII.

Closing the API Coverage Gap

Most fintechs build APIs incrementally to support integration requirements on an as-needed basis. As a result, some fintech capabilities still lack API coverage. With the emergence of AII, the number of possible fintech use cases will dramatically increase and system-to-system interaction will continue to replace direct end user interface. For fintechs to deliver on this new model, their APIs must go from incremental development to complete and comprehensive coverage.?

Optimizing APIs to work with LLM AI agents

Traditionally, API models are designed to be order agnostic. For example, if a schema contains fields A and B, it does not matter which is listed first in the message payload.? LLM AI agents, however, are highly precedence sensitive. Because the generative algorithms are all based on Markov chains, the parsing sequence is very important.? To work with LLM AI agents, fintechs must enhance their APIs to incorporate consistent ordering of objects and fields.

Conclusion

?The groundwork for fintechs to optimize their APIs to work with emerging AI capabilities can and should begin today

?

AI provides new capabilities that dramatically increase the range of use cases fintechs can offer their customers. Fintechs that fail to leverage AI capabilities will offer a narrower range of functionality than their AI-enabled competitors and risk becoming obsolete.?

?

?