Intelligent Lending

Hugo Assagra, Group Head of Credit Risk Strategy, OSB Group

Since the beginning of the first industrial revolution back in the 1760s, humankind has explored new methods to improve efficiency and productivity, accelerate growth, and create new opportunities for innovation and development. Fast forward to the mid-20th century, we are experiencing the Digital Revolution and although the aim is very similar, there is far more complexity and ambiguity around it.

The Covid pandemic fuelled a deep and wide technological change resulting in people adopting new ways to complete tasks. We live in an era where we can use our smart devices for everything we do, including communication, shopping, health stats, having doctor consultations and even watching TV. When we stream our favourite show on TV, heaps of data are collected, and recommendations are derived almost instantly to offer us a tailored experience. Using advanced analytics, Machine Learning, and Artificial Intelligence, algorithms can customize the offer depending on the previous shows we've watched to increase the chances for a 'play' to be clicked.

Before we progress further, it's worth clarifying some basic principles as follows:

• Artificial intelligence (AI) as an overarching principle refers to a machine's ability to perform cognitive functions associated with humans. This includes activities like learning/reasoning (Machine Learning), understanding human language (natural language processing), and robotics (robotic process automation).

• Machine Learning (ML) is a branch of artificial intelligence (AI) that focuses on using data to mimic the way human brains learn and improve over time. Instead of providing specific instructions, ML techniques allow computers to be trained by examples and work out the patterns, relationships, causal effect and more. Different techniques can be employed, such as deep learning and supervised learning.

• Robotic Process Automation (RPA) is a technology that uses software robots, or "bots", to automate repetitive and rule-based tasks. These bots can mimic human interactions with digital systems and applications, such as data entry, transaction processing, and communication with other systems. RPA aims to improve efficiency, reduce errors, and free up human workers to focus on more complex and value-added tasks. It's widely used across industries for tasks that involve structured data and follow predictable rules.

• Natural language processing (NLP) is another branch of artificial intelligence that focuses on the interaction between computers and human language. It involves the development of algorithms and computational models to enable machines to understand, interpret, and generate human-like responses. NLP presence has increased considerably in the last decade through various applications including chatbots, personal assistants, automation devices, etc. An important application of NLP in financial services relates to the ability to support customers before a human agent is required. Also, NLP has been used as an authentication tool through voice recognition.

AI has grown rapidly in several areas and more recently has emerged as a pivotal force across all sectors and industries. With the advancements seen in financial services, AI has been assuming a leading role in terms of reshaping the lending journey, whilst reducing risks, preventing frauds, and improving customer experience.

Lending journey bolstered by AI

The intelligent journey begins even before the customer applies for a loan and relies on the level of information available (internal transactions) or made available by the customer (Open Banking). With the use of ML, algorithms can analyse past transactions and identify products that the customer is likely to benefit from. As an example, customers with car finance on their file could be potential buyers of a better rate finance, car insurance, or even roadside assistance. Similarly, customers with 'rent payment' transactions that also have a savings account could be offered a tailored mortgage quote where income bands and savings amounts could be used to produce potential loan size and loan to value (LTV) scenarios. If this information is further combined with a 'council tax' payment transaction, such scenarios could be tailored to the regional level and reflect the house prices of that council to determine the likelihood of that lead turning into a deal. These examples are far too basic and are provided for the sake of example, although the appropriate techniques applied to the right data could generate endless paths to be explored.

"With the use of ML, algorithms can analyse past transactions and identify products that the customer is likely to benefit from."

The ability to pay is measured through the customer's affordability (income minus spending), which is heavily based on documents and information provided by the customer as well as open banking / third-party feeds where available. Using predictive analysis (predictive modelling, AI, ML, and data mining), customer transactions can be categorized and predicted to create an income and expenditure summary based on up-to-date data. Furthermore, an interactive self-serve tool can support the customer in uploading documents and simulating different loan scenarios in real-time with the support of RPA and NLP.

Using AI for propensity to pay assessment involves leveraging ML algorithms to analyse a large amount of data and predict the likelihood of a borrower repaying the loan. These models consider numerous variables, including historical payment behaviour, credit scores, income and spending, and other relevant factors. AI enables a forward-looking assessment by identifying patterns and relationships in data, improving risk assessment for financial institutions whilst contributing to higher automation capability - quicker and more consistent credit decisions with better risk assessment.

There are clear and palpable benefits to implementing AI in financial services, however, adoption can be challenging, especially within the lending environment. The most common challenges include:

Explainability comes down to transparency and trust that ML models will continue to outperform traditional models to make them a worthwhile investment. The models must be transparent and explainable for financial institutions to meet regulatory lending requirements.

Model development and deployment can be heavily impacted by a lack of resources or expertise as it can take several months to build and deploy a ML model, especially when it involves legacy systems.

Lack of data can pose a real challenge for financial institutions with reduced customer relationships – e.g. monoline lenders in general – as they tend to have a rather scarce source of transactional data that often requires a 3rd party data feed.

AI has the potential to transform the financial industry as the use of such technologies is contributing to a more efficient, accessible, and inclusive financial market whilst generating benefits to customers, financial institutions, and the wider economy. When AI augments rather than replaces good practices, it can support positive outcomes and improve financial well-being. However, AI can also pose novel challenges and create new regulatory risks (or amplify existing ones) and the regulators around the globe are cognisant of that.

All in all, AI can improve and streamline the whole customer journey from prospection, application, onboarding, servicing, and financial support with a high level of personalisation to meet customer requirements and achieve an optimum outcome - faster and better.

Like any other maturing technology, the benefits should be explored as well as the challenges understood. This way we will continue to play a key role in contributing to a robust and sustainable financial services industry. The main goal for any lender ultimately should be supporting the customers to achieve their goals and objectives in life and AI will continue to play a key role in this quest.

Financial

Artificial Intelligence

Machine Learning

Robotic Process Automation

Natural Language Processing

Process Automation

Customer Experience

Robotics

Cognitive

Voice Recognition