To AI or not to AI?

By Richard McLean

One of the big topics that was discussed in financial technology (fintech) circles at length in 2016 ― and will continue to be discussed at even greater length as we enter 2017 ― is artificial intelligence (AI).

Machine learning is emerging as the field within AI that is seeing the most amount of real-world applications and use cases among financial institutions, especially in the area of fraud.

Take, for example, a historic event that unfolded in March 2016 that demonstrated the power of machine learning: the victory of program AlphaGo of Google DeepMind over professional gamer Lee Se-dol. This exciting technological breakthrough demonstrates how far AI has come, even in just the past year, and how it is now able to catch humans out.

It’s true, though, that the idea of computers learning autonomously has been around for decades. So what has changed as we enter 2017? Why has machine learning gained so much ground in recent years and why does it continue to surprise us?

As we all know, technology predictions over multiple decades are hard to make. According to 1970s forecasts, mankind should have settled the moon and mars by now. But those who extrapolated from the first moon landings could not foresee inflection points like personal computing, the internet, smartphones and the sharing economy.

Current machine learning technology holds great potential to improve the way humans and machines work together. Machine learning can increasingly free us from many narrowly defined, repetitive, transactional tasks in a steady state.

This enables us to focus on higher-value, complex tasks in dynamic environments. While we used to have to monitor for fraud manually, now machines do almost all of the work.

In recent years, machine learning has gained ground. The technology now exists to deliver enterprise software systems for the financial services sector that can learn how to fully automate business processes at unprecedented levels, react to real-time changes and provide the best possible results for today’s digital business.

The market for AI systems for enterprise applications is expected to increase from just $202.5 million in 2015 to $11.1 billion by 2024. Soon, machine learning will be an integral part of enterprise solutions, making machines our digital co-workers.

However, even today’s most advanced machine learning algorithms learn in a very different way than humans learn. Without additional inflection points, extrapolating current machine learning technology to human-equivalent general intelligence would be a stretch. As we enter 2017, this will continue to be a challenge for the top computer scientists.

Incorporating AI into banks’ enterprise software, for example, opens the opportunity to simplify employee’s everyday lives and allows them to focus on higher value tasks. This is necessary to move industry into the next stages of growth and innovation ― at a global level.

From my own perspective, the highest potential lies in back-office functions and customer service, especially in shared service centers ― use cases beyond just fraud mitigation. As a priority and first step in 2017, businesses and the AI community should focus on automation of repetitive tasks in transactional knowledge work. Subsequently we should tackle use cases that have not been possible before, particularly based on speech, image or video recognition.

The question for banks and other financial services institution shouldn’t be whether to “AI or not,” but rather how to ensure that the world’s leading fintech companies come together as a community to take on the heavy lifting today for the benefit of the generations of tomorrow.

While it is unlikely that machines will exhibit broadly-applicable intelligence comparable to or exceeding that of humans in the next 20 years, rapid progress in the field of specialized AI will continue. It’s a fantastic time to be betting on the field and its fruit.

The writer is regional chief financial officer at SAP Asia Pacific Japan