Generative AI, a groundbreaking technology, is reshaping the landscape of artificial intelligence (AI) and its applications. Ahsan Shah, Senior Vice President of Data Analytics and AI at Billtrust, highlights the pivotal role of large language models (LLMs), such as OpenAI’s ChatGPT, in this transformative journey. The fusion of natural language processing (NLP) with conventional AI functionalities is creating a new paradigm, particularly impacting the payments sector.
The rise of generative AI in payments
Ahsan Shah underscores the significant impact of generative AI, emphasizing its ability to write, communicate, and generate content. This human-centric interface, seamlessly integrating NLP and conventional AI, is proving complementary from a technological standpoint. Applications such as fraud detection, forecasting, anomaly detection, and recommendations have existed for years, but generative AI introduces a novel dimension by understanding language multimodalities. Shah sees this as a catalyst for innovation in the payments domain, bringing efficiency, accuracy, and security to financial transactions.
From predictive to generative AI: A complementary approach
While AI, including machine learning, has been an integral part of back-end systems for years, Shah cautions against rushing into the new ecosystem without proper consideration. He points out that AI models are knowledgeable about the world but lack business-specific insights. Addressing security parameters, infrastructure, and the nuances of feeding data to language models becomes critical. Shah’s recommendation is to focus on known problems and low-hanging fruit initially. Immediate wins can be achieved in customer support, sales, marketing, and anomaly detection in payments, streamlining processes and reducing manual overhead.
Building technical infrastructure for AI in payments
Shah likens the ubiquity of AI applications to a “new electricity” that can be harnessed universally. Despite the broad applications, he anticipates the emergence of domain-specific AI applications fine-tuned for specific use cases within the payments landscape. Notably, he emphasizes the importance of data organization and suggests being deliberate in addressing data fragmentation, an issue predating the existence of AI. Shah foresees a future where domain-specific AI systems, particularly in payments, will be trained on their own data. This shift requires careful consideration of data sharing between third parties, especially as foundational models may only be built by large companies like OpenAI, Anthropic, and Google.
The future of AI in payments
In Shah’s view, taking no action is the least desirable option, but he advises against hasty approaches to AI adoption. Acknowledging the conversational element of AI, he predicts a swift adoption due to the significant process improvements achieved through natural language interfaces. Manual submission and data sifting may become outdated when compared to the substantial efficiency gains offered by AI, particularly in payments.
Navigating the AI frontier in payments
As generative AI reshapes the payments landscape, businesses must tread carefully, aligning their workforce to become proficient in providing context to AI models. Shah’s pragmatic approach advises addressing known issues first, building confidence in proven wins, and gradually expanding into more complex workflows. The prospect of AI agents becoming multi-agent ecosystems is on the horizon, opening up possibilities for cross-domain and cross-functional applications with careful consideration of necessary guardrails.
