Adopting GenAI That Works, in a Landscape Full of Noise

The real potential of GenAI in business lending is not as a cure-all, nor is it in replacing the human touch. If anything, GenAI is a tool to empower lending teams rather than replace them.

‍

GenAI has become the buzzword of today, with almost every company featuring its "AI-powered" solutions, claiming immediate results. Unfortunately, after speaking with dozens of bank leaders in the past few months, it has become increasingly clear that this "over-promise and under-deliver" problem hurts the entire lending industry. After testing various systems and tools and investing extensive training resources, bankers have become skeptical and disappointed with the “promised” AI benefits. While AI may appear magical, the key to successful results with it in highly complex industries, like business lending, lies in thoughtful and careful implementation.

The real potential of GenAI in business lending is not as a cure-all, nor is it in replacing the human touch. If anything, GenAI is a tool to empower lending teams rather than replace them. GenAI can help teams ingest and analyze much more data, uncover overlooked patterns, and enhance decision-making processes. However, that’s true only when customized to the unique challenges of the lending industry and to the unique needs of the lenders using it.

Thinking about AI as a ‘plug-and-play’ solution sets unrealistic expectations, especially in lending. The real power of AI is grounded in the quality and context of the data fed into it. AI models are only as good as the data and context they receive. In other words, only when AI is fully embedded in workflows, supported by relevant contextual data, and operating in a closed and compliant loop does its potential in lending become apparent. When properly implemented, GenAI can bring value to areas like loan origination, underwriting, and borrower assessment. But without the proper context, knowledge, or data, AI models will always fall short of improving core areas like underwriting or customer engagement, or even providing a minimal value that will justify the effort involved in making it work.

Off-the-shelf and generic solutions just can’t address the specific challenges of business lending. Perhaps the best example would be ChatGPT. Although many of us use it for our personal use cases, such as planning vacations or doing quick research, it’s clear to us all that ChatGPT won’t be able to estimate the DSCR of a borrower without access to the borrower’s tax returns, training as an “AI Underwriter,” verifying numbers, prompt engineering, etc. On top of that, one would have to protect PII and ensure no data leaks into the AI model or is used to train it, and in most cases, that’s just not feasible. Trying to leverage or pilot GenAI technology that isn’t geared and fine-tuned for the lender’s unique needs is unlikely to be successful.

In parallel, banks should acknowledge the importance of vertical GenAI solutions. While generic tools like Co-Pilot provide immense value in areas like knowledge management or workflow automation, they fall short when applied to complex use cases like business lending. The unique challenges of lending call for AI solutions specifically built for the industry—this is where vertical GenAI comes in. Vertical GenAI solutions are designed with a deep understanding of the specific workflows, data sensitivities, and compliance requirements inherent in lending. Unlike generic GenAI models, vertical GenAI is finely tuned to handle industry-specific processes with the appropriate context, helping lenders reduce bias, improve risk management, and make more informed decisions, all while maintaining compliance with regulatory standards.

The next step in the ‘GenAI in lending’ journey is recognizing the synergy between AI and human expertise. While AI excels at automating repetitive tasks, like spreading financial documents or cross-validating information across dozens of data points, it cannot replace the human touch required for relationship-building or complex decision-making. It’s the blend of human-in-the-loop and AI efficiency that drives operational excellence and allows lending teams to make faster, more informed decisions. Banks shouldn’t expect their AI to be smarter than their team but rather to make their teams smarter.

When built properly, integrated into the lender workflow, and combined with human insights, GenAI’s ability to process and analyze vast amounts of data can help reduce bias, improve risk assessment, and expand access to capital.

As the lending industry continues to evolve, the role of GenAI will be defined by how well it is integrated and tailored to the unique needs of banks and their teams. When used thoughtfully, GenAI can unlock new efficiencies, improve decision-making, and help banks navigate the complexities of business lending in a regulated, relationship-driven landscape. To realize this potential, lenders must go beyond the hype and focus on utilizing AI solutions that work hand-in-hand with their people, processes, and data. By doing so, they will not only meet the demands of today's lending environment but also lay the foundation for a future where AI drives greater innovation, inclusivity, and growth.