How AI can power equitable and inclusive risk assessments
/
Abhishek Bhasin is head of product at Uplinq, a global credit assessment platform for small business lenders.
Artificial Intelligence is proliferating in the financial services industry by enabling use cases from customer facing front-end applications to back-office operational efficiency initiatives. How it can help shape and efficiently deliver products and services that solve real-life customer problems — especially small businesses’ need for financing — is key.
A December 2024 U.S. Congress Bipartisan AI Task Force report focused on responsible AI innovation. It called for an approach that balances innovation with the need for regulatory oversight, in order to realize the full promise of AI.
Responsible AI innovation calls for balancing multiple stakeholder objectives: small businesses, lenders and regulators, which can be at odds with each other. Small businesses need a constant supply of working capital in a timely manner while lenders need to ensure that they are able to fulfil that demand in a risk prudent way. Regulators, on the other hand, emphasize safety and soundness of the system, while enabling equitable access without bias. Lenders can have a tough time juggling these priorities and given their risk-averse nature, often choose regulation and risk over the needs of the small businesses.
From machine learning to generative AI, these technologies can be helpful if deliberately designed with these objectives in mind. Instead of chasing quick wins through short cuts, a natively designed, sound AI approach considers both intended and unintended consequences. For example, AI can bring objectivity in decision making by leveraging large datasets, but any tool or solution’s success depends on varied and authenticated datasets. This approach satisfies the objectives of lenders and regulators, while ensuring the timely flow of funds to small businesses.
Many small business owners belong to marginalized communities and often have their loan applications declined due to design deficiency in lending systems. AI can therefore expand access to financial products, particularly for underserved communities through customized solutions and streamlining lending processes.
Staying ahead of regulation
Given the highly regulated nature of financial services, firms are expected to follow all laws irrespective of technology advancements. They need to be able to trace the outcomes of AI-supported decisions. In this way, a lender should be able to explain an adverse customer outcome based on AI to ensure compliance with the Equal Credit Opportunity Act (ECOA).
By contrast, regulators facing time constraints may be using AI to identify non-compliance with regulations, potentially raising the bar for scrutiny.
Lenders have a big opportunity to create systems that can get ahead of regulation.
A winning risk assessment toolset always comes back to design: Companies need to be clear on how they’re using AI beyond unspecified claims. Simply claiming a solution is powered by AI and machine learning isn’t going to cut it. As expectations evolve, lenders will face a higher bar for regulatory compliance. They will need to balance inclusion with profitable growth. Assessment solutions that succeed should be adaptable to changes in the economy, market conditions, demographics and other market dynamics. Lenders will also need to conduct experiments and tests prior to product deployment to ensure the AI software offering meets regulatory standards. It should also be flexible enough to accommodate the evolution of regulations over time.
AI has the potential to transform the financial services sector to serve the needs of all customer types, including small businesses. AI system designs need to accommodate multiple stakeholders. Additionally, adequate controls around the security of data are becoming table stakes given the proliferation of cybersecurity incidents. These controls need to be a part of the core architecture design of AI systems and not a superficial layer at the top. AI can be a force of good if designed in the right way, leveraging human expertise where necessary.
