A Tipping Point

The financial integrity landscape continues to evolve, as do the techniques illicit actors employ to avoid controls, so too must our strategies and technologies as financial integrity professionals. Generative artificial intelligence (gen-AI) promises to revolutionize how people learn, expand their knowledge, unlock knowledge in the moment, and boost their productivity, unlocking new efficiencies and capabilities in the fight against financial crime.

As an early adopter of gen-AI in financial compliance, the Institute for Financial Integrity (IFI) launched AskFIN, a gen-AI-powered financial crime assistant, in BETA in Q2 2024. AskFIN, now available to all DOLFIN users through an expanded BETA, exemplifies the transformative potential of AI when paired with authoritative resources and a focus on privacy and security.

Contextualizing Gen-AI and AI

It’s important to contextualize and baseline gen-AI within AI. AI serves as an umbrella term that encompasses a range of technologies, including—but not limited to—machine learning (ML), natural language processing (NLP), deep learning, gen-AI, and others. While AI is often casually used interchangeably with these terms, AI is not synonymous with gen-AI. This is a key distinction because the language and nuance we use when discussing these topics is important.

While AI has been around since the 1960s, with the creation of the first chatbot (Eliza), and most institutions have been using machine learning (ML) for decades, the recent buzz and surge in interest can be attributed to advancements in gen-AI capabilities—the new kid on the block.

Gen-AI made its debut in 2014 with the introduction of Generative Adversarial Networks (GANs). However, it was not until the release of ChatGPT-1 in 2018 that gen-AI began to gain broader traction and application. The tipping point came in 2022 with the launch of ChatGPT-3.5, which propelled gen-AI into the mainstream.

The adoption of ChatGPT has been the fastest of any technology in history; it achieved an unprecedented milestone of reaching 1 billion monthly users within just three months of its launch. This explosive growth can be attributed to significant improvements in NLP and computational capabilities, particularly in hardware like Graphics Processing Units (GPUs).

Today, gen-AI is everywhere, including on our cellphones. We have passed the tipping point—the irreversible moment—and gen-AI implementations and applications, and their impact on our lives, will only increase.

AI Is Not a Magical Wand, But It Is Transformative

Most institutions are experimenting with gen-AI, methodically exploring benefits and risks. One of the biggest challenges with deploying gen-AI is that the outputs are “non-deterministic.” Put another way, when used in products, gen-AI regularly provides different outputs for the same input on different executions.

To be generative and connect the dots, gen-AI must make assumptions and decisions with imperfect and incomplete information—gen-AI is dynamic by design. This introduces significant complexity to building products and capabilities with gen-AI, particularly in the compliance space. While gen-AI can deliver incredible outputs that significantly improve productivity, most outputs require human review.

The opportunity we face is how to best pair ourselves with these new capabilities. I am a strong advocate of Harvard Business School Professor Karim R. Lakhani’s thinking that “AI [won’t] replace humans, but humans with AI [will] replace humans without AI” and similarly “human-computer symbiosis.” Deploying gen-AI is about understanding how to get the best out of these capabilities. The ultimate configuration will likely vary within each institution and on each task/objective.