How to Make the Promise of Generative AI a Material Reality

Originally published in The Wall Street Journal
 

Five steps companies can take to extract measurable and scalable value from generative AI.

By Murli Buluswar, Head of Analytics, U.S. Personal Banking at Citi, and Vivek Jetley, President and Head of Analytics at EXL

It’s been a year since generative AI (GenAI) became available to the masses, setting in motion a wave of innovation and speculation that caused virtually every business in the world to refocus their tech strategies around the new technology.

There’s just one problem: Few have managed to integrate the capability in a meaningful way. In fact, according to a recent study conducted by EXL, while 91% of top U.S. financial services firms and insurance companies have implemented AI-driven solutions to some degree, only one-third (36%) have found a way to start using the technology widely across business functions. While most business leaders value the potential, few have a developed blueprint outlining how material aspects of their firms’ operations will be rearchitected to incorporate GenAI.

What companies are increasingly finding as they move from experimentation to real-world business cases for GenAI is that there is a big leap to go from building a promising-looking tool to implementing an enterprise-wide strategy. Fortunately, there is a path of best practices being forged by early leaders. We have developed the following framework for extracting measurable and scalable value from GenAI.

Step One: Seize the Moment

The first step to any enterprise GenAI initiative is understanding why now is the right time to embark on this journey. We are living at a unique moment in time when the confluence of data, sophisticated algorithmic models and computing power have created a perfect nexus for innovation. Companies that unlock that opportunity now are poised to catalyze new growth at a level we’ve never seen, but they must do so with a clear roadmap of milestones and metrics that define success. The gap between those who commit to an enterprise strategy and those who dabble will get visibly wider.

Step Two: Clearly Define the Problem You Are Trying to Solve

That means companies need to resist the urge to plow ahead on GenAI initiatives simply for the sake of it. The first and most critical step in any GenAI initiative is to identify areas where it will have the biggest impact, and work backwards to how GenAI tools will deliver measurable outcomes.

Efforts to integrate GenAI will need to address the complete picture and respect the widespread ramifications decisions made today will have on business outcomes tomorrow.

That process starts with identifying what GenAI does best and focusing solutions on those core strengths. We’ve found that the most valuable areas where GenAI can methodically solve concrete business challenges are in generating content from unstructured data, content extraction and summarization, and conversational intelligence.

Step Three: Identify and Prioritize Opportunities

When deciding how and where to focus GenAI project resources, companies should start by looking for the most obvious areas of overlap between the things that GenAI does well and the things that create challenges in their business. These include manual, repeatable tasks that are prone to human error, customer friction points, workflows that generate a great deal of data and operational processes that have significant associated costs. Then, they should prioritize development based on the relative levels of certainty they will be able to drive a material business outcome.

Step Four: Establish Guardrails

While it is critical to establish clear goals and stick to a focused product development plan, it is also important to stay flexible as tech and product teams ingest new insights and learn how these models will behave in a real-world setting. To that end, first-round GenAI applications should not be used out of the box on initiatives that have a direct customer impact. They should also be built on closed data sets with a human in the loop to monitor for unintended consequences and make necessary course corrections along the way. That is not to suggest public large language models (LLMs) will not be valuable—rather it is to suggest that internal data will always provide more context for a particular issue, and the ability to combine broader LLM capabilities with internal data will be more powerful.

Step Five: Weigh the Human Factor

GenAI is a transformative technology with truly massive potential. But its benefits will not be realized without some growing pains. GenAI will reinvent how decisions are made in an organization and how people see themselves as part of that organization. It is therefore important that any leaders running an enterprise GenAI initiative consider the emotional component and include efforts to work with people to redefine their roles, up-level their skill sets and maximize their contributions.

Accordingly, corporate efforts to integrate GenAI will need to address the complete picture and respect the widespread ramifications decisions made today will have on business outcomes tomorrow. That requires not just a tool or technology lens, but also a broader view on how these new capabilities will affect the interplay between human and machine. Firms that have a big-picture perspective that guides granular decision-making will win big.

Learn more about how EXL is making generative AI work for enterprise customers.