To prepare for gen AI tomorrow, financial institutions should prioritize data organization today

From Time to Google, 2023 was declared a standout year for the development, integration and popularity of generative artificial intelligence, or GenAI. However, the rise of GenAI as a buzzword across industries has not led all financial institutions (FIs) to embrace its direct implementation.

Rather, many regional and community FIs have not yet begun leveraging GenAI for meaningful evolution in their business. Research indicates over half of banking leaders have assessed their own AI initiatives as poor or fair with cost and implementation challenges named as complications.

One notable obstacle to GenAI’s implementation has been regulatory concerns, which have stalled implementation in its tracks as leaders grapple with how to handle matters like AI hallucinations. In these instances, where GenAI presents falsehoods as facts, financial institutions need safeguards in place to protect against missteps that can influence nearly every area of critical decision making.

Federal regulators have offered guidance on model risk management, and AI experts have weighed in with options to bridge the gap including quality control measures to mitigate exposure to misguided data. While this advice often falls short when it comes to GenAI in particular, the model risk management guidance can provide practical insights to better understand the current climate and hindrances for the road ahead.

Yet one area often overlooked by analysts and leadership is data organization. FIs can better prepare for a future with GenAI by exploring how their organizations manage architecture and access for both structured and unstructured data.

According to McKinsey, “Gen AI’s heavy reliance on unstructured data adds another layer of data-related complexity, and banks’ current data strategies and architectures may not be up to the task… Data quality—always important—becomes even more crucial in the context of GenAI.”

AI and FIs Today

As a logical first step, most regional and community FIs are looking to deliver chatbot experiences for their customers. This conversational AI approach relies heavily on decision tree scripts and highly structured data to provide timely, accurate inroads for, as an example, a self-service customer access platform. Historically, these solutions have been adequate to meet the needs of smaller FIs that have been slower to modernize their tech stacks.

However, the rise of more robust AI uses and applications such as ChatGPT making headlines has some FI leaders wondering what’s next. As AI becomes more generally accepted and used by consumers, a reset of expectations is likely to slowly drive a sea change in the finance industry. While customers may not be asking for GenAI offerings by name yet, a consistently more tech-savvy consumer base is raising expectations for personalization, speed and convenience when it comes to accessing everything from customer support to account details and product purchase options.

These and more can all be catered to with more powerful AI capabilities which, in turn, rely on well-conceived, organized data structures.

For GenAI to effectively roll out in the regional and community FIs of today, proper inception and planning, including system design and a thorough understanding of data sources and architecture, are absolute requirements. While the investment in more robust IT support systems and staff may seem daunting, the benefits far outweigh any near-term detractors.

GenAI is different

Those FIs who are convinced of GenAI’s value and are ready to take on the next phase of their business development should embrace large language models, or LLMs. These dedicated AI solutions are designed to use considerably sized data sets to learn specific business needs as they train over time. The data needed for LLM implementation is much larger than that needed for more straightforward chatbots since the system is looking to teach itself how to naturally process business needs and contribute unique insights.

To this end, how data is organized within a financial institution’s existing structure can have a direct impact on how accurate and valuable the LLM can be. Just like humans, LLMs have to sift through large amounts of data in order to answer a given question. When that data is incorrectly or unclearly organized, an LLM will have the same struggles as a human when looking for correct information in an efficient manner.

With this in mind, it becomes clear quickly that financial institutions need to be wary of “garbage in, garbage out” scenarios with respect to their data. Any FI even remotely considering the future implementation of GenAI solutions should prioritize the organization of their data today. Tomorrow’s AI systems will be built on a foundation of the data available for access, and structured, curated data is the essential bedrock for orchestrating accurate, efficient GenAI deployment.

Next steps FIs can take today

Financial leaders can make progress towards their future GenAI journey by evaluating who is overseeing their data curation in the here and now. With this will come a likely emphasis on more roles for staff in charge of content creation and data administration. Leadership and human resources should work together to outline talent roles dedicated to leading data management initiatives. Smaller FIs in particular may want to consider working with an external consulting firm to define their goals, establish skill set requirements and source appropriate talent.

To prepare for the inevitable future demand for GenAI solutions, FIs should also start establishing robust data management practices. This can include everything from incorporating zero trust policies to auditing existing systems and data sources to ensure only essential integrations are established. Indispensable integrations should subsequently be evaluated to confirm data is organized, synced and seamless across platforms.

Ensuring compliance with industry regulations when it comes to data collection, use and access is also an essential part of preparing to implement GenAI. Dedicating a human resource to oversee governance for AI use, or leveraging a firm dedicated to the field, can help keep organizations current with shifting recommendations and regulations. As AI capabilities evolve, so, too, does the governance surrounding it. Recently, an Executive Order from the White House required the Secretary of the Treasury to provide a report regarding “best practices for financial institutions to manage AI specific cybersecurity risks.” Remaining up to date with changing state, federal and international laws will provide guidelines for how and what data to structure for the road ahead.

Establishing procedures and gatekeepers, both human and digital, will help streamline data input and classification, laying groundwork for more robust data applications and analysis later on. The result will be the highest quality data possible for a financial institution, and one fully ready to embrace GenAI and its benefits.

Mark Leher is Director, Product Management at Alkami.