How can AI improve the efficiency of loan processing in banks
What’s wrong with banking?
Lots. In advanced banking there’s such a moral hazard in simple and complex instruments that the banking industry has to be highly regulated or tends to prey on the weak or the unknowing. And sometimes unconsciously so. Given that public banks are driven by their profit motive and their dedication to their shareholders one would expect that the invisible hand of capitalism would keep banking on side in competition. Unfortunately, in smaller countries and now some of the biggest countries in the world there is a consolidated ownership in banking and there’s less of an interest in competition. However, efficiency is becoming obvious. In a retail banking operations there are thousands of people that work moving papers around or in some cases individual spreadsheets as if it was 1975. These are obvious opportunities to enhance both security and automation operations that will finally limit the outsourcing to low cost knowledge workers across the world. These people will be replaced by AI infused applications. These applications will be surprising to people because they will not look like they’re current applications. It will not look like core banking systems, or account opening systems because those are database and completely determinate. They are also brittle and require great changes when regulation changes or transforms. Think about when SOX was rolled out and whole systems had to be augmented and new reporting systems had to be built out to support the nature of the legislation and to keep CEOs and others out of jail. The same thing happened with IFRS17… there was a mountain of new software change that had to occur because of inflexible database systems. Tens of millions of dollars have been poured into the systems and they come along with heavy maintenance duties and what poor illegitimate CTOs call technical debt. Technical Debt, to me, is created by poor understanding of management teams supported by poor understanding and misinformation by CTOs and IT Architects. Goaded by techno-illiterate management teams and pushed by boards for greater and greater efficiencies, they force shortcuts or create mistakes that are impossible to unwind without recreating software or what they will call ”refactoring”.
Well this mountain of technical debt that has been created willfully will now be under the pressure of being changed and this time not by refactoring with procedural code and database technologies... There will be a new wave of frameworks that are much more complex interspersed with AI agents and other autonomous means for integration and on the fly analysis that is needed in modern organizations.
AI features are coming for your systems.
I’m not talking about chatGPT becoming the front end to your banking experience. That’s stupid. Chat interfaces are only one method to create a user experience with people and sometimes it’s super effective and other times well get serious their forms to be filled in to apply for banking services and products.. and a chat integration with 100 fields that tend to repeat themselves across accounts is not a good experience. The people that are thinking that there’s only one way and one new way to apply AI again our foolish and should not be listened to. What we’re going to do is create multimodal interfaces to Banks whether that be through video voice conversation form filling.. wouldn’t it just be nice if the bank remembered you and your contacts forever? Would it be great if your bank anticipated your needs given the flow of change in the economy or you are family instantaneously so that you’re given options to better work with the bank. And when banks have new products wouldn’t be great they didn’t tax you with surcharges and fees just because they can. Wouldn’t it be great if regulators could just watch the banks operate within the constraints of the regulation and automatically find them or adjust them necessarily as they might break rules. Well that comes from applying modern AI but in ways that are novel and new and involve hundreds of changes not just rolling out a new grammar checker or chatbot.
I thought about this this morning in 2024 and was thinking about the expertise required to change over systems like this. It’s incredibly difficult we need people with business subject matter expertise, regulatory expertise, as well as computer science infused with artificial intelligence understanding as these services and layers will evolve. In a large scale Bank like the Canadian behemoths or some of the American leading Banks like Bank of America and JP Morgan this will require the retraining and education of thousands of people before systems become viable and newly constructed. There is so much legacy morass in large companies that it’s unlikely that they’ll be able to transform efficiently using this new technology it’s too different. It’ll likely be that some of the Neo Banks or Banks spinning off new divisions with green-fields will be able to create themselves with these new technologies and have greater efficiency. This is like where I’m most in the Innovation will occur in new banking as mid-office systems change and layers of AI built to transform data based on rules and experience will be generated instead of created. Wouldn’t be fantastic that you could specify an interface and something like a spreadsheet and then automatically generate API to create interfaces with their party systems. Wouldn’t be great if you could share that space so that between trading partners you could agree instantly and easily across a simplified surface like a spreadsheet to do things like field napping between systems with trading partners?
Interfaces with regulatory bodies will be generated instead of created.
In addition, interfaces between applications like shopping applications or home inventory applications will be generated instead of created based on flexible dynamically created and semantically coherent graph infrastructures instead of flat table databases invented at home. That level of integration will take a decade maybe or maybe half a decade. But still there are modern things that can be done right now in business terms with artificial intelligence and its generative capabilities to create more flexible systems than ever before in history. Here are some areas that deserve some luck that I generated using perplexity.ai and some smart questions and it also this techniques that I know of. All that follows after this is generated by brand new algorithms that use reputable sources of information and plans to create research results that are applicable and grounded in researchable transparent information available on the internet today. Read on.
(btw: If you are a researcher or an analyst and you’re not using perplexity.ai or something very similar to it, today you’re wasting your own time and that of your employer. #robrule)
Modern, new algorithms and chatbots have the potential to significantly improve the efficiency of loan processing in banks in several key ways:
1. Faster processing times: By automating various steps in the loan origination process using AI and machine learning algorithms, banks can dramatically reduce the time it takes to process loan applications from days or weeks to just a matter of minutes. AI-powered systems can quickly gather and analyze data from multiple sources to assess a borrower's creditworthiness and risk profile in real-time.
2. Improved accuracy: AI models can analyze vast amounts of structured and unstructured data to gain a more holistic and accurate assessment of a borrower's risk profile and creditworthiness. This can lead to more accurate underwriting decisions and potentially lower default rates. Manual reviews of loan applications are prone to human errors and biases.
3. Operational efficiency: By automating the loan origination and decision-making process, banks can process a higher volume of loan applications more resourcefully and cost-effectively. Underwriters can be freed up to focus on just the more complex, non-routine cases.
4. Better customer experience: Faster processing of loan applications leads to improved customer satisfaction and retention. Valuable prospective borrowers will not need to wait in limbo for a decision.
5. Reduced costs: Automation of the credit process through AI enables lenders to process loan applications at a lower operations cost, and potentially be able to pass the savings along to their customers in the form of lower APRs and fees.
The risks of AI-based lending decision models, if not properly designed, tested and governed, include the following:
1. AI/ML models may lead to "black box" approaches that have stakeholders raise issues ranging from the pragmatic ("Why was I turned down?") to the potentially nefarious.
2. Potential model biases and model drift need to be monitored to ensure that creditworthy borrowers are not turned down and decisions do not have a disparate impact on protected class groups.
3. In addition, there are important adverse selection problems that AI/ML lending can not solve when models are not robust. During a financial downturn, AI/ML lending models will also predict downgraded unemployment propects for borrowers, and may cut lending despite a potential borrower making all her credit payments on-time.
4. AI/ML models are trained on back data and cannot take into account fast-changing market conditions. The past doesn't guarantee the future.
5. Documentation, knowledge sharing, and setting up of IT processes and IT controls need to be iterated and implemented.
6. Aspects of Explainable AI also need to be considered proactively in model development. The model outputs need to be continuously monitored for bias and fairness.
7. The AI models created to help determine creditworthiness have an immense finacial and social impact of individuals, in terms of determening loan eligibility and rates, and there are increasing calls for techniques that can provide explainability on how these "black-box" models make decisions that impact this aspect of people's lives.
In summary, the development and use of AI/ML models in the financial industry presents both opportunities and challenges. It is important for practitioners to carefully consider the potential risks and biases associated with these models, and to develop robust processes for monitoring their performance and ensuring fair and responsible use. This requires ongoing collaboration between data scientists, compliance experts, and business stakeholders to ensure AI is used in an ethical and accountable manner.
Citations:
[1] AI in Lending and Loan Management: Impact & Challenges - Plat.AI https://plat.ai/blog/ai-in-loan-processing/
[2] Harnessing the Power of AI in Lending: Artificial Intelligence for ... https://www.lightico.com/blog/harnessing-the-power-of-ai-in-lending-leveraging-artificial-intelligence-to-revolutionize-loan-originations-servicing-and-document-management/
[3] Revolutionizing Ways AI in Loan Processing is Transforming ... https://www.businessnext.com/blogs/ai-in-loan-processing-malaysia/
[4] Leveraging AI in Lending and Loan Management - Docsumo https://www.docsumo.com/blog/artificial-intelligence-based-loan-management
[5] How AI is Changing Lending Practices - ezbob https://www.ezbob.com/ai-changing-lending-practices-in-banks-and-financial-institutions/
[6] How AI in banking is shaping the financial industry - Liquidity Group https://www.liquiditygroup.com/resource-funding/how-ai-in-banking-is-shaping-the-industry
[7] How AI Is Adding Faster Funding And Efficiency To Small-Business ... https://www.forbes.com/sites/forbesfinancecouncil/2024/01/19/how-ai-is-adding-faster-funding-and-efficiency-to-small-business-lending/
[8] AI in Lending Guide: Impact, Challenges, Use Cases [2024] - App0 https://www.app0.io/blog/ai-in-lending
[9] Key Challenges Banks and Financial Institutions Need to Consider ... https://www.pennanttech.com/blog/key-challenges-banks-and-financial-institutions-need-to-consider-for-large-scale-lending-platform-transformation/
[10] Challenges in Loan Origination in Banks, & Financial Services [2024] https://www.app0.io/blog/challenges-in-loan-origination
This article is generated by Rob Tyrie because he’s interested in banking and insurance and it’s obvious need for modern processing that is highly automated. Generative AI represents a leap forward in Automation in the creation of interaction and conversation between Banks and customers. So here we are trying to create the universe in the future once again.
Rob is the founder of IronstoneAdvisory.com as well as a co-founder of the Grey Swan Guild a virtual Think Tank.
He has a fondness of books and interesting curious people. He is a builder. 🧰👷♂️
