Generative AI: Implementing a proactive and structured strategy for adoption

The technological landscape is evolving at a lightning-fast pace, creating a myriad of opportunities and challenges for banks of all sizes. The advent of generative artificial intelligence (AI) is creating significant waves in the banking industry, evolving from a novelty to a potential solution that is making substantial impacts and is predicted to drive trillions of dollars in GDP growth globally.

It’s no secret that banks have leveraged AI for decades to assist with loan decisioning, marketing strategies, fraud detection and data analytics. However, the advent of generative AI solutions, particularly Large Language Models (LLMs) like ChatGPT, is a more recent phenomenon.

Today, vendors are eager to integrate generative AI capabilities into their products, with notable examples including Salesforce, Alteryx, ServiceNow, and Adobe Acrobat. This makes it critical for banks and institutions of all sizes to take a strategic approach to AI integration and policy development to help ensure compliance, mitigate risk, increase efficiency and support long-term success.

Use cases for AI grow by the day

The emergence of AI platforms like OpenAI’s ChatGPT and last year’s release of GPT-4 have generated a mix of excitement and caution. Since then, companies like Morgan Stanley have leveraged the solution to enhance their internal knowledge base and handle federated search capabilities for their personnel. Adoption is inevitable, particularly with ubiquitous platforms like Microsoft Office integrating GPT’s capabilities into its products.

Awareness and disruption brought on by AI tools and LLMs are spreading throughout financial services. Technology vendors are scrambling to integrate these solutions into their offerings, and investments in new AI startups are on the rise. LLMs will continue to grow at an incredible pace, resulting in several applications emerging in the industry.

For a time, the use cases for generative AI were simply theoretical; this is no longer the case. Buy now, pay later provider Klarna, for instance, now employs AI to perform the task of 700 contact agents. This move is a significant contributor to Klarna’s bottom line.

Payments giant Stripe uses Chat-GPT 4 to autonomously monitor community and support forum posts, flagging malicious actors engaging in fraudulent activities. This method of fraud detection assists in maintaining the integrity of the platform while protecting users from potential threats. In addition to fraud detection, Stripe utilizes technology to streamline helpdesk functions by automatically routing issue tickets and summarizing user queries. This enhances the efficiency of Stripe’s technical support and ensures that customer issues are addressed quickly and accurately.

While the capabilities of LLMs in code are still developing and require oversight, they can create functional software if given high-level instructions. Resources are being dedicated to refining and integrating these capabilities into development workflows. Regarding web and mobile design, GPT-4’s image classification function can be used to create a fully functioning webpage prototype.

Furthermore, the democratization of data analytics is propelled by the LLM’s ability to create and act upon SQL queries and reports based on plain language instructions. This helps bridge the longstanding gap between business users and technical personnel.

And it has potential to level the banking landscape; smaller banking institutions can leverage LLMs to use their data for improved decisioning.

Risks associated with AI integration, including in existing contracts

Despite AI being one of the most compelling tools banks can implement, inherent risks can leave banks and account holders susceptible to bad actors. Research shows that three in four employees use AI tools secretly, and although this usage is well intended, it can leave a bank’s associated data vulnerable.

Banks should exercise concern over generative AI’s pervasiveness. The pace of progress in generative AI is accelerating, with new models and tools introduced weekly. LLMs are now integrated into social media, smartphone operating systems, web browsers and various apps.

Additionally, many vendors working with banks are building, acquiring or partnering with new AI technologies for improved efficiency. Without protective clauses in existing contracts, compliance issues could arise, and compliance could be compromised. These issues have led many banks to delay the adoption of AI tools, which means they are missing out on operational efficiencies and cost savings.

What banks should consider when adopting AI

Banks on the fence about implementing AI into their operations should consider taking foundational steps to ensure safe AI adoption. Developing a nuanced AI policy that addresses data privacy, regulatory compliance and risk tolerance is critical for safeguarding and preventing unsafe use of AI. While it might be tempting to draft AI policy using ChatGPT, it cannot create adequate policies that keep pace with an ever-shifting regulatory landscape. This approach fails to provide a tailored blueprint for AI governance and adoption.

Implementing an internal LLM assistant is the most effective approach to eliminating employee circumvention of AI restrictions. Offering internal tools that are safe for proprietary data ensures staff have secure alternatives to use. Various options and approaches are available for this type of solution, catering to all levels of risk tolerance and budget constraints.

It is common for banks to use AI through their vendors unknowingly. Banks must understand their vendors’ incorporation of AI to ensure their data remains secure across several contracts.

Other AI technologies that provide value for banks include loan decisioning engines, conversational AI, Robotic Process Automation (RPA), and biometric authentication.

AI advancements will shape the future of banking. More and more banks are considering adopting these solutions. That means implementing a proactive and structured strategy is in their best interest. This ensures banks can get the most out of generative AI without compromising security and compliance.

Connor Heaton is Director of Artificial Intelligence at SRM.