JPMorgan Chase has introduced a generative AI assistant named LLM Suite, marking a significant step in the bank’s strategy to integrate advanced technology across its operations. This AI tool has already proven effective, assisting more than 60,000 employees with tasks such as drafting emails and reports. Designed as a comprehensive portal, LLM Suite enables users to access external large language models, including the one developed by ChatGPT’s creator, OpenAI.
According to sources familiar with the plans, the software is anticipated to become as widespread within the bank as the videoconferencing tool Zoom, CNBC reported.
Cognitive Revolution
JPMorgan Chase, the largest U.S. bank by assets, is demonstrating the rapid adoption of generative AI across American businesses, following the launch of ChatGPT in late 2022. This swift integration is mirrored by competitors like Morgan Stanley, which has already introduced two AI-powered tools for its financial advisors, and tech giant Apple, which announced in June that it’s embedding OpenAI models into the operating systems of hundreds of millions of consumer devices.
Dubbed by some as the “Cognitive Revolution,” this technology is being compared to monumental innovations like electricity, the printing press, and the internet. JPMorgan CEO Jamie Dimon remarked in April that generative AI is likely to automate tasks traditionally handled by knowledge workers, potentially transforming the workforce. He predicted that AI will “enhance virtually every job” at the bank, which employed around 313,000 people as of June.
Lifts ChatGPT Ban with New In-House AI
JPMorgan, which previously banned employees from using ChatGPT, is now offering a customised version of OpenAI’s model through a new tool called LLM Suite. The decision to initially ban ChatGPT stemmed from concerns about exposing the bank’s sensitive data to third-party providers, according to Heitsenrether. She explained that the bank’s data is a critical asset, and they were cautious about it being used to train external models. By developing LLM Suite, the bank can harness the power of AI while ensuring that their data remains secure.
LLM Suite has been rolled out widely across JPMorgan, with various teams in the consumer division, investment banking, and asset and wealth management already using it. The tool is proving useful for tasks such as writing, summarising lengthy documents, solving problems in Excel, and generating ideas.
However, Heitsenrether emphasised that simply making the software available is just the beginning. The next step involves educating employees on how to effectively use prompt engineering tailored to their specific domains. This training will help them understand the tool’s capabilities and limitations, allowing them to fully unlock its potential. As more employees delve into the technology, innovative ideas are beginning to flourish.
Additionally, LLM Suite offers JPMorgan’s engineers the ability to integrate features from external AI models directly into their own programs, providing further flexibility and customization in how the bank uses AI.
Generative AI’s Impact Far Surpasses Previous Tech
JPMorgan has been investing in traditional AI and machine learning for over ten years, but the launch of ChatGPT prompted a strategic shift. Unlike traditional AI, which focuses on specific tasks like pattern recognition and prediction based on historical data, generative AI operates on extensive datasets to create new patterns, resulting in human-like text and realistic images.
Generative AI offers “exponentially bigger” potential due to the flexibility of large language models (LLMs), according to Heitsenrether. The bank is exploring numerous applications for both traditional and generative AI, with several already in use.
JPMorgan leverages generative AI to produce social media marketing content, plan travel itineraries for clients of its newly acquired travel agency, and summarise meetings for financial advisors. In its consumer banking sector, AI analyses satellite imagery to strategize the placement of branches and ATMs and assists call centre staff in quickly resolving customer queries. In global payments, which handles over $8 trillion daily, AI plays a crucial role in preventing substantial fraud losses.
However, the bank remains cautious about deploying generative AI for direct customer interactions due to concerns over potential inaccuracies from chatbots. Looking ahead, Heitsenrether anticipates that the generative AI landscape may consolidate around “five or six major foundational models” and reveals that JPMorgan is evaluating both major U.S. tech companies’ LLMs and open-source models for future integration into its systems.
Three stages for the evolution of generative AI
Heitsenrether outlined a three-phase plan for the advancement of generative AI at JPMorgan.
The initial phase involves making these AI models accessible to employees. The second phase, which has recently commenced, includes integrating JPMorgan’s proprietary data to enhance productivity. The final phase represents a significant shift, where generative AI could evolve into autonomous agents capable of executing complex, multi step tasks. This transformation would elevate the roles of regular employees, allowing them to operate with AI assistants akin to managerial support.
This technological evolution is expected to both empower and displace workers, potentially reshaping the industry in unpredictable ways. According to consulting firm Accenture, banking jobs are particularly susceptible to automation compared to other sectors like technology, healthcare, and retail. Citigroup analysts project that AI could add $170 billion to the banking sector’s profits within four years.
Heitsenrether suggested viewing generative AI as a tool that handles routine tasks, allowing employees to bypass tedious processes like sifting through spreadsheets and focus on more strategic, high-value work.
