AI and big data: What to know, and how to think about it

AI and big data are going to transform the banking industry: The rapid evolution of generative AI in the past year and a bit is threatening to fundamentally reshape a number industries everywhere — and finance is no exception. And with machine learning, big data, and algorithmic decision-making only growing in importance in the industry, the question becomes: What kind of transformation do we want, and how can we ensure it is done responsibly?

On the first day of the Enterprise Finance Forum, we welcomed on stage three industry insiders who have plenty of experience with AI and big data: CIB Group CFO Islam Zekry (who was earlier in his career the first head of data science at an Egyptian bank), Chief Data Scientist at Beltone Holding Basma Rady, and Synapse Analytics co-founder and COO Galal Elbeshbishy.

ICYMI- McKinsey’s Larry Lerner, who co-moderated the panel, set the scene with an introduction on what every banker needs to know about AI. You can download The promise of AI and GenAI in banking: Separating hype from potential (pdf) here.

Larry has also agreed we can share McKinsey’s What every CEO should know about generative AI (pdf) .

The 4 “V”s that make big data, well, big: Volume, velocity, variety and value. While volume is what distinguishes a dataset as big data, velocity is a business’ capacity to efficiently process and derive value from the influx of data including customer narratives embedded within datasets, said Zekry. Variety covers a wide array of data types, extending beyond numerical and financial transactional datasets to encompass elements such as social mediator interactions, preferences, videos and emails, Zekry added.

The most valuable “V”: Generating substantial and meaningful value from all of the data a bank or business gathers that could drive decision-making processes is probably the most important element to keep in mind, said Zekry. Demonstrating the return on investment in expanding big data analysis is imperative, as decision-makers may be hesitant if they do not see a clear path to tangible returns, he added.

The AI 90/10 mix in credit decision-making : Whether or not to give a client access to credit is an example of where supervised AI learning can be applied, said Zekry. CIB also uses algorithms to inform decisions about who to issue credit cards to, he said. “Financial institutions are reaching out to us for one defined use case, [and] that is instantaneous credit [decision making],” Elbeshbishy said, echoing Zekry’s thoughts. This is 90% of today’s AI use cases in the banking sector, with generative AI taking up just 10%, said Zekry. Applications of AI are currently limited in the local market, but could triple over the next 5-10 years, Rady said.

Thin credit profiles call for AI interventions: The current level of financial inclusion and credit coverage is quite limited, leaving substantial capacity for growth, said Elbeshbishy. One place where banks have been using AI to good effect is to handle clients with thin-to-no credit profiles by using alternative data as well as tapping into predefined and pre-trained datasets.

How do you secure management buy-in for AI? Demonstrate value: In the case of an institution like Beltone, the leadership team already firmly believed in the power and value of AI and data to help people at all levels of the organization make decisions, Rady said. For data scientists with management teams that still need to be coaxed on the uptake, it’s important to continuously demonstrate the value of data and AI to senior leaders. She argued that AI adoption in Egypt extends beyond financial institutions and is evident in digitally native tech startups. The adoption rates, however, vary among organizations, with some struggling to secure management support, Rady said.

The features of an AI-first institution: Data assets + talent + balanced decision-making. “If you have data and you have the talent, the sky’s the limit,” said Rady, highlighting the key pillars for AI-driven growth in financial institutions. Rady stressed the need to consider data collection in product launches and investing in talent growth, while striking a balance between machine learning and human domain expertise.

Ensuring that AI is in the right hands and applied ethically is central to any conversation about the technology, said Elbeshbishy, stressing that this involves regulation to prevent misuse. The complexity of AI systems and the inherent biases in data is another aspect that needs to be considered, Elbeshbishy added, highlighting the need to differentiate between ethical concerns and biases originating from the data itself. Businesses will need to address biases within their AI systems, even if it means refactoring code to ensure fairness and avoid discrimination.

What’s an AI-driven bank? Being an AI-based bank means being highly responsive, offering services beyond traditional banking, and tailoring the customer experience to individual consumption patterns and lifestyles, Zekry said. He underscored the need for financial institutions to adapt their business models: “Under the severe pressure we are facing from the new demographics and the new customers and all their needs, I think we are changing our business model into a more intelligent, more responsive, more customer centric business model.”

What we can do better: Personalization + cloud solutions. AI will make it possible for banks to instantly create personalized. Together with wider adoption of cloud-based solutions as the Central Bank of Egypt permits it, banks could be looking a more profitability by delivering better services more efficiently. There’s also significant potential in the local banking industry, our panelists agreed, with its relatively low market penetration, which leaves room for expansion, potentially serving up to 60-70% of the population more effectively.

What’s next? Financial inclusion for the next 10 mn people: Rady highlighted the role of data in building models for faster financial inclusion and overcoming biases through alternative data sources. “If you rely on new types of data sources that are less biased by nature, you can end up including more people into the financial services sector and have them become financially included,” she said. Within the data science and AI community, there is a heightened awareness of privacy and ethical considerations, Rady added, saying that when developing models and technologies, there is a conscious effort to assess and mitigate biases. This proactive approach ensures that the resulting systems are designed to include as many individuals as possible while minimizing any potential biases, she said.