Why hasn't AI taken over the Finance Industry?

Brian Collins
Written by:
Brian Collins
Why hasn't AI taken over the Finance Industry?

AI has already proven to be a revolutionary force around the world. From deciphering dead languages to finding previously unknown patterns in historical data, it is redefining our understanding of our past. With content creation and recommendation systems on platforms we use daily it changes the way we view our present. And with assisting on health diagnosis and the creation of new pharmaceutical drugs it is already changing our shared future.

Silicon Valley is also betting big on AI as the most important tech breakthrough since the smartphone.

For all of this, AI has yet to be a revolutionary force in the financial services industry (FSI). Despite the huge amount of blogs, articles, and papers dedicated to the subject, as someone who’s job it is to watch for fintech trends around the world I have been extremely underwhelmed by the so called “impact” of AI in financial services. Unlike the humanities, entertainment, and healthcare industries listed above, when you read any article on the impact of AI in finance it usually comes down to a few key areas: credit decisioning, fraud and AML, and (sigh) chatbots.

The lack of imagination is not the author’s faults - there are simply very few real world examples thus far of AI working in the financial industry.

The better question then “what is possible” for AI in finance is “why isn’t finance adopting AI more quickly?”.

What’s Blocking AI Rollout

There’s two options here: First, that there is a limited opportunity for AI in finance. We know this not to be the case.

FSIs are some of the most data centric organisations in the world. There remains a huge opportunity around filling gaps in existing workflows, speeding up timelines of decision making, and mitigating risks taken by these organisations through core their products. Even if that wasn’t the case, many of the existing technologies these companies use (Salesforce, Slack, Atlassian etc) have announced new AI tools themselves. So AI is clearly coming for the industry whether they want it or not.

Which leaves option two: something is blocking FSI’s from engaging more deeply.

And that brings us to the very large whale in the data lake: the risk of access to permitted data that a corporation undertakes and that regulators expect be upheld.

AI is run off a Catch-22:

  1. AI is only as good as the data it can consume. To be a good decision maker, it requires huge amounts of the highest quality data imaginable. For the finance industry, this is mostly permitted data.

  2. Financial service businesses are the most regulated industry in the world, and access to permitted data is the holiest of holies. Because of the heavy regulations around data protection (GDPR in Europe, CDR in Australia etc) access to permitted data is strictly controlled even internally.

And this is why I believe we haven’t seen more actual impact of AI into finance: because without access to the best data, AI’s simply cannot provide value.

The big question: is there a solution to this? Potentially…

Large corporates always have the option of a buy/build approach. The problem is AI is expensive. Monumentally expensive, even for the largest financial companies in the world. There is another way however.

When blockchain was all the rage, FSIs wanted access to the power of blockchain technology without the burden of cryptocurrency. Middleware companies like Blockdaemon popped up to allow corporates a safe middle ground of using DLT without unnecessary regulatory burden.

This is exactly what Redactive does so well act as a middleware solution that acts on behalf of the corporate to be a buffer between a third party service provider (OpenAI, Slack, Salesforce, etc) and the companies internal permitted data. These tools would package up and protect permitted data, acting as both the organising librarian who tracks and checks out the data, and the security guard ensuring nothing leaves that shouldn’t.

With access to permitted data, AI’s would be able to move quickly in training and deployment of actual services that both FSIs and consumers need.

And who knows, maybe this will work and we’ll all finally get the chatbots we’ve been waiting for. Or hopefully, we get something much, much more powerful.

Learn more about Redactive

Forget about chunking, embedding and Vector Databases. Redactive's managed RAG pipelines make it easy to access the context you need via our SDK.