How to effectively deliver AI-powered analytics in the age of digital banking

An opinion piece by Sid Bhatia, Regional Vice President, Middle East & Turkey, Dataiku

First, the bad news: COVID-19 hit the region’s banks hard. According to KPMG, H1 2020 net profits for the GCC banking sector were down by a third on the same period in 2019. But the good news is that certain pandemic-independent factors prevail that spell opportunities for FSI companies. By 2019, the World Bank reported almost two thirds of the Middle East and North Africa (MENA) population had moved to cities. For GCC countries this urbanization ranged from 84% in Saudi Arabia to 100% in Kuwait. In such a high-net-worth geography, this is especially good news for the banking industry, given the long-established connection between urbanization and growth in the banking sector.

Meanwhile, across the region, economic-diversification initiatives — from the TASMU Smart Qatar program to the UAE Smart Data Strategy — call for digitization across the FSI sector, and many organizations have leveraged analytics to align themselves with such visions. A Deloitte report published just before COVID-19 emerged projected a US$67-billion market for big data in the GCC this year. And some organizations in the region are talking about an “AI-first” strategy in FSI analytics.

In theory, given their operating models, incorporating AI-driven models into FSI businesses should be relatively smooth, but many of them face a slew of challenges. Obstacles commonly begin with disparate legacy datastores linked to proprietary applications. Factor in M&A activity and you soon see a fragmented information pool that cannot be leveraged optimally.

The green-eyed monster

Exacerbating this fragmentation is a tendency across the sector towards departmental stewardship of data sources, making effective data-warehousing problematic within a single organization. This is true to an even greater extent for data sharing across different organizations. Not only are banks and insurance firms “data-jealous”, like their internal departments, but they are bound by regional and global regulatory obligations. But, much as doctors across hospitals collaborate on patient care, the digital age may demand the same of FSI entities for the enhancement of everything from simple customer experiences to anti-money-laundering (AML) efforts.

Another challenge faced when trying to implement AI in financial firms is skilling. People that are used to making risk-management decisions based upon certain models may justifiably worry that their performance may degrade when shackled to a new methodology. Organizations that adopt analytics tools must ensure employees are trained to collaborate effectively, combining skillsets in risk, business operations and data science to build effective models.

It is not difficult to imagine the value seen by investment professionals in the data-crunching prowess of AI. Normally unreachable insights suddenly come into play when advanced machine learning is let loose on the right data warehouse. Or at least, that is the perspective. In truth, many disappointments have come from this market-cracking perception of AI as it relates to the FSI sector.

Keeping it simple

But for those who steer away from “big win” use cases, towards “quick wins” — such as process optimization, customer-experience enhancement, and broader risk-management — success has often followed.

It is in these relatively humdrum areas that AI has been a disrupter, delivering more risk-conscious decisions, cost-effective operations, and robust customer loyalty. Risk management is, by now, a venerable discipline, chasing and anticipating a growing number of factors. A major one in the GCC is regulatory compliance, and AI has gone a long way towards the establishment of a stronger banking system, especially when it comes to the detection, reporting and prosecution of financial crimes.

AI has also played a huge role in the advancement of risk-pricing, leading to significant benefits for those able to leverage alternative data and agile modeling during the COVID crisis. And as more customers bank from home, expectations surrounding individualization have called for deeper data insights to enhance their journeys.

So, what has set apart the soaring maiden flights from the stalled engines? How have the agile firms scaled AI to the point where it has positively impacted costs, operations, customer satisfaction, risk management and growth? What is the recipe for smooth flight?

Best practices at a glance

First, an obvious point often made when separating out the winners from the losers in the digital transformation game: technology implementations must be aligned to business objectives. When scaling up AI capabilities, it pays to start with quick, cheap wins that demonstrate noticeable value to as many stakeholders as possible while causing minimal operational disruption.

Incremental changes can pave the way, over time, to a data-based, AI-powered culture. AI will eventually be indistinguishable from core business processes, so balancing bottom-up empowerment with top-down support will be vital. When the more ambitious changes come, many employees will undergo sweeping reskilling. They must be ready and willing to do so. Your roadmap must include strategies for change that mold employees into willing participants — curious and accepting of the new technologies.

Just as data must leave behind the siloes of the past, AI tools must not be deployed in isolation, but as part of a strategic whole. For financial institutions, the concept of ROI will loom large in any capacity-building journey. It is important that initial investments are designed to bear fruit that underpin all the investments that come after them, thereby ensuring buy-in from all stakeholders.

Assessment against promise

Carefully designed and applied KPIs will keep the AI ship on course. Do the results of an implementation meet with expectations? If not, why not? Assessment is another process that requires perspectives from many disciplines, from data scientist to risk manager. The inclusive approach is a best practice that is indispensable in successful AI campaigns — empowering the data scientist so that they in turn can empower the salesperson, the teller, the investment consultant or the risk manager.

Tools are important, to be sure, but in standing out from the crowd, you will need well-equipped and dedicated people. They will take you from vision to implementation, bringing strong AI governance, operational transparency, MLOps, and a host of other best practices that will be critical to success.