• Mikhail Dunaev, Chief AI Officer at Comply Control


• 08.05.2024 01:15 pm
#banking #AI

Understanding the landscape of AI adoption in banking is akin to navigating a maze of contradictions. While more than two-thirds of bank executives aim to ramp up digital transformation investments to 10% by the end of 2024, only 6% of retail banks have a roadmap for AI-driven transformation at scale. 

The reason behind these inconsistencies is clear—the banking industry, renowned for its conservatism and heavy regulation, grapples with the disruptive force of AI. Although it is far from being a brand-new technology, it is nonetheless a rapidly evolving one with profound implications for the market, including labour dynamics. 

Despite recognizing the immense potential of AI solutions, retail banks hesitate to embark on large-scale AI deployments. Until unified regulatory stances on AI crystallise—an issue just beginning to surface on the agenda—AI adoption in banking will remain restrained.

Temporary obstacles to AI mass adoption in banking

Banking leadership's uncertainty regarding AI scaling stems from the looming specter of regulatory penalties for potential missteps. In the past few years, we have seen regulatory bodies across the globe paying close attention to Big Tech, including the U.S. and EU’s increased scrutiny of the digital marketing sphere. 

While Europe has embraced the reputation of the world’s fiercest tech regulator with a more centralised approach to data privacy policies, the U.S. tech industry regulation is more state-level and decentralised. 

For instance, when it comes to digital advertising and consumer privacy, California has one of the most impactful of these state-level regulations, the California Privacy Rights Act (CPRA). However, there is a lack of a unified approach to data privacy at the federal level. The AI regulatory landscape is no less confusing, considering privacy concerns and customers’ financial data needed to feed the generative learning tools used to build AI models. 

Furthermore, the regulatory uncertainty prolongs the industry’s wait-and-see approach. It results in the absence of well-tested, reliable, and user-friendly AI solutions capable of streamlining processes, reducing costs, and driving profitability. Presently, viable AI banking products tailored for fintech are scarce. And developing such products requires significant investment in proprietary software solutions. Such efforts are typically feasible only for industry giants equipped with substantial resources and testing capabilities.

Yet, the hurdles impeding AI implementation in banks are temporary. With the advent of proven, off-the-shelf AI products customised for banking, these obstacles will gradually be eliminated. While this transition, both regulatory and technical, will take time, the distribution of adaptable and easily customised solutions will significantly facilitate mass AI adoption in the next five years.

Not a trend, but the inevitable future of personalised banking

Contrary to being a fleeting trend, AI represents the future of banking. AI will revolutionise banking and payment systems by automating myriad routine processes that currently demand substantial human labour.

Crucially, AI’s value lies not in replacing humans but in enhancing their capabilities. For instance, while humans remain pivotal in making final compliance decisions—as required by regulators—an ever-evolving AI and machine learning can complement their efforts by learning, detecting, and predicting compliance risks in a larger volume. 

The development of robust neural networks will catalyse transformations that surpass even the impact of the internet. Thus, in fraud detection, AI will discern suspicious transactions and furnish detailed reports, with humans assuming the role of effective moderators. 

In the next five years, AI will facilitate the creation of personalised financial instruments for every client. Ultimately, conventional notions of internet banking and fintech will wane, replaced by personalised AI banks tailored to each individual user’s preferences.