Harnessing the power of AI to accelerate financial inclusion

There is a popular analogy comparing artificial intelligence (AI) to electricity. Just like the role of electricity during the second industrial revolution, AI has the same potential to greatly transform how industries function and benefit humankind. However, this analogy in my opinion is over-simplified. The power of AI is much harder to standardize and harness, and it requires a far more collaborative effort to bridge existing gaps between academic research and solving real-world problems.

Part of why AI implementation has been slow is that too many practitioners take a generalist approach, rather than becoming specialists. As a result, this has led to an imbalance between the demand for actual AI applications and the supply of specialist skills available from academics. It inhibits the ability to realize AI’s full potential over the long-term, especially in specialized fields such as financial services, which requires customized and continuously-optimized solutions sensitive to conditions of a given market.

As many studies have rightfully pointed out, the Covid-19 pandemic is disproportionately hurting the poor and underprivileged. While the world works to recover and heal, there is an urgent need to make finance more inclusive and accessible. This is where AI researchers and industry practitioners must come together to help the 1.7 billion adults, roughly a third of the world’s working age population, who lack access to financial services.

From both a commercial and social perspective, AI has great potential to make it more viable for quality financial services to reach the masses. For instance, the “310 model” developed at Alipay uses AI to optimize loan processes for small and micro businesses, so it takes less than three minutes to apply, one second to approve with zero human intervention, overcoming the challenge of social distancing amid the pandemic.

Based on a risk perception and management system to drive the lending process, AI powers everything from profiling applicants and assessing risks, to making decisions and intervening when there is a problem. This allows us to scale effectively while maintaining low risk levels, which cannot be achieved without specialized industry knowledge.

Specialization also allows for more fruitful collaborations between AI researchers who are good at theory and methodology, and practitioners working on industry applications. One such example is CFPA Microfinance, a company which provides financing to small and micro enterprises in rural China. Working with a team of AI researchers from Ant Group, they were able to optimize their loan origination and risk management, reducing loan approval process from two or more days to less than 10 seconds.

Another scenario that brought together AI theory and practice is Dingsunbao, an AI-driven auto insurance solution that assists car owners to assess losses using real-time video. The application works by guiding policy holders – through on-screen prompts – to capture information needed to accurately calculate damages and improve the efficiency and simplicity in processing claims. Dingsunbao is estimated to have saved over RMB1 billion in claims handling costs and around 750,000 hours of work for claims adjustors.

These examples are possible only with the collaboration of research and practice, with in-depth industry knowledge informing the development of AI technologies. This is where the complexity of AI goes beyond electricity.

For AI solutions to remain practical and meaningful over time, new models of collaboration are required. The more knowledge and experience shared between researchers and practitioners, and between practitioners and industries, the better and more customized AI technology will become.