In the first post in our series on AI in lending, I began a conversation with ChatGPT, a so-called “large language model” of AI that’s made headlines recently. In that first part of the discussion, we covered 3 major benefits of AI in lending. In our second post in the series, we discussed the major risks banks should be aware of.
Assuming a bank has weighed the risk-vs-reward tradeoff, the next question I wanted to address was whether a bank should plan to develop AI technology in house vs turning to an experienced vendor in the space.
Then, if considering a vendor, what are some of the questions such a bank should ask to be sure they’re working with a reputable, regulation-compliant vendor?
Read on for the conclusion of my unedited conversation with the artificial intelligence called ChatGPT on the subject of AI in lending.
On Developing AI for Lending In House
In general, would you recommend a bank try to develop AI models in house or engage with a third party vendor, and why?
The choice between developing AI models in-house or engaging with a third-party vendor depends on various factors such as the bank’s resources, expertise, and strategic goals.
If a bank has a strong team of data scientists and AI experts, developing AI models in-house may be a good option, as it would allow the bank to have full control over the design, development, and deployment of the AI models. Additionally, an in-house team would have a deep understanding of the bank’s data, processes, and customer needs, which can lead to the development of more effective and tailored AI solutions.
On the other hand, engaging with a third-party vendor may be a more practical option for a bank that lacks in-house expertise or resources. A vendor can provide the bank with access to state-of-the-art AI technologies, expertise, and best practices, which can help to ensure the successful deployment of AI models in the lending process. Additionally, vendors often have experience working with multiple banks and can provide valuable insights and recommendations for improving the lending process.
In both cases, it’s important for a bank to carefully evaluate the vendor’s or in-house team’s approach to addressing the risks associated with AI in lending, and to ensure that they are transparent, fair, and aligned with industry standards and best practices.
On Working With a Vendor
If a bank wants to engage a third party vendor to provide AI services, what questions can the bank ask to address these risks?
If a bank wants to engage a third party vendor to provide AI services, here are some questions they can ask to address the risks we discussed:
- Data evaluation: Ask the vendor about the data they use to train and operate their AI systems, including the sources, accuracy, and representativeness of the data. Also ask about the measures the vendor takes to avoid bias in their data and algorithms.
- Transparency and explainability: Ask the vendor about the transparency and explainability of their AI systems, including how decisions are made and how customers can obtain information about the inputs and algorithms used in the lending process.
- Performance evaluation and monitoring: Ask the vendor about their performance evaluation and monitoring processes, including how they ensure that their AI systems are making fair and accurate lending decisions and how they test for bias and accuracy.
- Collaboration with stakeholders: Ask the vendor about their collaborations with stakeholders, including regulators, industry groups, and customer organizations, and how they ensure that their AI systems align with industry standards and best practices.
- Ethical considerations: Ask the vendor about their approach to ethical considerations, including data privacy, data protection, and responsible AI. This can help to ensure that the vendor has a clear understanding of the bank’s priorities and expectations in these areas.
By asking these questions, a bank can gain a better understanding of the vendor’s approach to addressing the risks associated with AI in lending and can make more informed decisions about engaging their services.
Artificial intelligence is just beginning its explosion into the mainstream. The technology evolves rapidly seemingly by the minute, and we’re only starting to comprehend the ramifications it will have across the entire spectrum of life and society. It is absolutely certain that we will see more and more integration of banking services – including and especially lending – with such technologies as they become more prevalent and accessible.
Regulators are already working hard to learn how to make sure this technology is applied safely and fairly, but it’s going to be a game with evolving rules as the data emerges and all parties digest it and learn from it.
It behooves a bank to consider carefully the when, the how, and the risks to be aware of as they move to embrace this next wave of technology for the benefit of both the bank and its customers.
It is an exciting time to be in technology and banking, even if – perhaps only occasionally – it feels exciting like a dogfight in next generation jets that you’re fairly certain you know how to fly, but you just can’t be sure….
We look forward to exploring those blue skies with you for years to come.
John is the Chief Technology Officer for Premier Insights, Inc. and Radiant Lending Technologies LLC.