A.I. – Mind the [expectation] Gap

Science fiction has always been a precursor of what’s to come, acting as a beacon for the wide-eyed wanderers and innovators. At the same time, Hollywood has spoiled our expectation of time it takes to get there – I blame montages for that.

Nowhere is that expectation more misaligned than in the Fintech (including InsurTech & RegTech) space. For a phrase, let alone the industry, that did not exist a few years ago the expectation of what the practitioners of these “arts” can deliver is phenomenal.  This leads to a hype-cycle, further widening the expectation gap of what the technology can deliver. But the misalignment is not about what but “when the technology can deliver”.

Here is my attempt at aligning or at least managing the expectations gap for A.I. within FinTech.

A.I. is about data

I am sure it comes as no surprise that A.I. is all about data but I am always surprised at how many people don’t truly appreciate that point. So A.I. is a data play. Almost all A.I. companies today, with a few exceptions, are focused on teaching their machines – translation, busy with large-scale data cleansing exercise. The few companies that are offering elements of programmatic intelligence (likes of Google and Facebook) have been data companies for a while now.

Data Quality

Financial institutions are notoriously bad with data management, largely because the industry has grown through acquisitions resulting in a data soup. This lack of data consistency across multiple repositories is often mislabeled as the “Legacy systems” issue. It’s not a giant leap to conclude that what ties banks and insurers to their legacy systems is not the dated IT infrastructure but rather the data-soup within it.

The reluctance to open the pandora’s box has what, in my opinion, lead to inflated expectations from A.I. further fueled by a barrage of FinTech, InsurTech and RegTech conferences. I often walk away from these conferences with a sense that the expectation gap that a bit of code will make the legacy data issues go away is snowballing. Technology will definitely ease that pain, however, the task for data cleansing still fall on data owners.

Data Volume

Like with our learning needs, the volume of information needed for machines to learn is substantial. You would not expect someone to push the boundaries of theoretical mathematics just by teaching them basic sums. A large volume of knowledge-base is a pre-requisite to smarter machines. However, the volume of data within financial services operations is not large by any stretch of the imagination.

One of the biggest challenges to the application of A.I. within RegTech is the low volume of data. 200,000 pages of regulatory text is a lifetime of reading but for machines, it’s not enough to sink their teeth in. What makes great lawyers is not the law they have digested, rather their experiences that provide the context for their work. The challenge for RegTech’s working with A.I. is around how best to provide that context.

A.I. Rollout plan

Given the pivotal role played by data quality and volume in building a truly intelligent and autonomous machine, I expect the customer touchpoints to be the first to rollout A.I. in the next 2 – 3 years. This is because these areas relay on newly generated data (decoupled from the “legacy data-soup”) and learnings from other verticals such as e-commerce can be applied here.

Application of A.I. to back-office operations will come but it will start life as an advanced rules engine in the near future before truly becoming autonomous in the next 5 – 7 years.

 

FinTech’s “Henry” moment 

cards_dribble_1.jpg
Image by Joshua Jouppi

Application of robotics to the automotive industry would not have been possible without Henry Ford’s re-designed manufacturing process. I feel that Fintech is currently going through their “Henry” moment, exploring new approaches to delivering financial services. However, unlike Mr. Ford, we are more aware of the incoming technologies in the near to mid-term future and can design for them.

At CoVi Analytics we are re-designing the current approach to risk and compliance with a mission to simplifying and automating business activities. Although we are clear on what technologies will help us deliver our vision, we build a technology-agnostic approach – e.g. our process mapper that’s fundamentally changing how business documentation is produced and shared. This allows us to remain agile and respond to new technological innovation – Augmented reality perhaps.

Is your expectations gap aligned? Over what timeline do you see A.I. truly rollout within the financial services? Let’s start a conversation.