What’s the state of AI in the lending industry today?
Throughout the industry, AI is being implemented in many different areas of the process. From targeted marketing and lead qualification to credit scoring, due diligence and loanbook analysis.
Undoubtedly the most important component of the process is credit risk. Traditionally two different mechanisms exist to determine risk, a selection mechanism and an outcome mechanism:
- Selection mechanism distinguishes between whether an applicant is accepted or rejected for a loan.
- Outcome mechanism determines the performance of the loan – or informally, whether or not a loan is fully paid or defaulted.
What’s the biggest trend shaping AI in lending this year?
As mentioned, the most important area to a lending institution is risk mitigation, and this has been shown as a focal area for AI in academic research and is being applied across the lending industry currently.
Current applications include:
- Detecting fraudulent behaviour / anti-money laundering
- Credit risk
- Loanbook maintenance and monitoring
Voice search is also making a lot of noise across the globe. In 2018, 114 million smart speakers were sold, but 2019 could see an 80% increase taking total sales to over 200 million. Long-tail search terms, using natural language, will be the second largest use (after music) however the financial industry will have to be smart in its implementation as over 30% of users said they had no interest or were fearful of using this technology to manage money.
What’s the largest challenge to AI adoption in lending?
There many challenges in regard to lending, such as:
- Regulations such as the Basel Accord I, II and III, ensure that any model being utilised that effects lending decisions must be explainable. This completely eliminates ‘black box’ AI models, such as neural networks.
- AI can require very data hungry models, particularly in the field of deep learning. Not every business has access to the scale and quality of data necessary to leverage certain techniques.
- A company looking to adopt AI methodologies cannot expect a single technique to work for a variety of problems. Each problem will require its own analysis, design and implementation.
What’s the future of AI in lending?
In both consumer and corporate lending, artificial intelligence will enable lenders to gauge the future creditworthiness of a declined applicant, rather than only the current creditworthiness, essentially unlocking a new market that is currently unviable to the lender, enabling the financial institution to generate more business.
An integral component of the lending industry is the finance broker, which has the potential to transform into a ‘self-driving” process where an applicant applies for finance and the deal will be carefully constructed by the broker system rather than the heuristic-based approach advisers tend to take; each lender available to the brokerage will be scrutinised on the probability of acceptance by the lender based on data-driven domain expertise.