Remember the cobrador del frac? Spain’s frock-coated debt collectors took an elaborate name-and-shame approach that emphasized recovering money by any means necessary. It summed up the biggest problem with debt collections: efficiency and customer experience have too often been secondary considerations at best.
Debt collections practice is modernizing around the world, though. Italy, Spain, Germany, France, and Belgium show the greatest propensity to up their game among the Eurozone nations, while China, India, Indonesia, Mexico, and Brazil are leading the pack beyond. Customers are becoming the key consideration for banks, and technology is improving efficiency across the board.
EXUS' Financial Suite already tackles this problem to an extent: it can rate customers based on the stage of the debt recovery process they're in, and collections teams can decide whether to send them a letter, follow up with a call, or make a personal visit.
Artificial intelligence will streamline and automate this process still further, allowing more informed and nuanced decisions to be made.
AI can tell banks and agencies who is most likely to pay and under what circumstances, and provide automated portals for debtors to make and negotiate payments - but it can’t modernize the entire industry all by itself.
Machine learning, the underlying process that makes AI systems work, is all about pattern recognition - assessing statistics and predicting likely outcomes based on them. This has happened, so that should happen. It’s a significant gain for debt collections because it means AI can spot trends for defaulting loans and refine early response processes, including more detailed and personalized approaches to customers.
However, AI can ‘overtrain’ on past events, missing out on new types of risk or events, and it’s not capable of making exceptions. There are also concerns that the ‘black box of AI’s actual functioning may create a lack of transparency for customers, who may feel they’re feeding their data into a computer which automatically rejects their application or demands a payment, based on a process they don’t understand and can’t be shown. In general, even leading innovators like Metro Bank’s Alex Park agree that data security and regulation are priorities for the emergent AI banking technology.
AI is already being applied in the front office of financial institutions, detecting early risks at the credit scoring stage of loan applications. Credit scoring - a task that involves collecting data and comparing it to strict criteria - is ideally suited to AI. AI can also bring in consumption and transaction data to establish willingness to pay and responsibility from other sources, helping consumers with less-developed credit records.
‘Computer vision' can scan faces to assess moods by looking for tiny changes in expression, and cross-referencing them against a list of their implications. It’s a powerful tool for customer service, and it’s already being used in financial services. Chinese Insurance company Ping An uses the tech to screen customers for reliability, assessing the tells that reveal their disclosures are incomplete. In the UK, Metro Bank has already used computer vision to improve the performance of their chatbots, making them more responsive to the facial expressions and moods of human customers, switching scripts as customers become impatient, unnerved, or frustrated.
Other aspects of customer service can also be automated, or enhanced by AI’s data processing power. The pre-legal action stage of collections involves a marketing campaign’s worth of contact points, channels of communication, and decision-making about how best to reach a given debtor. AI can streamline that decision-making process by highlighting successful trends in similar cases and recommending courses of action. If the courts are involved, AI can create, check and submit legal documents in a fraction of the usual time.
AI also has the power to improve existing practice, by analyzing data and highlighting areas of concern. Analysis of recorded calls and the data points they provide - specific uses of language by successful collections agents, common objections from debtors, and how they’re best overcome - generates shareable best practices for the agency as a whole. Compliance violations can also be highlighted and predicted with machine learning, with particular language uses tagged for intervention and targeted coaching.
Customer-first approaches
AI is a tool, not a solution. The businesses who put it to best use, and make the most significant improvements in their debt collections operations, will be those who remain focused on their customers rather than their own systems. A process and system which uses AI to understand, predict, and empower customers to pay their debts will outperform one that’s used to catch them out.