The race among major Wall Street institutions to implement artificial intelligence systems capable of independently managing complex, start-to-finish business processes has entered a new and more competitive phase. On Monday, the Bank of New York Mellon (BNY), one of the longest-operating financial institutions in the city, announced that it will incorporate Google Cloud’s cutting-edge agentic AI framework—featuring the recently unveiled Gemini 3 multimodal model—into its proprietary internal AI platform known as Eliza. Named in honor of Eliza Hamilton, the wife of the bank’s founder and early American statesman Alexander Hamilton, this platform represents the core of BNY’s digital transformation ambitions. The decision to embed Google’s agentic technologies into Eliza marks the continuation of the bank’s long-term strategy to modernize its technological foundations and further its leadership in AI-driven financial operations.
Through this integration, BNY seeks to achieve a significant leap in speed and operational fluency across routine workflows. Sarthak Pattanaik, the bank’s chief data and AI officer, explained in an interview with *Business Insider* that Eliza already utilizes multiple large language models, yet the inclusion of Google Cloud’s advanced framework introduces new levels of autonomy and orchestration. For instance, tasks that have traditionally been laborious—such as onboarding new clients—involve a multitude of small, detail-oriented steps: staff must collect identification documents, confirm tax information, extract critical details from various forms, assess risk factors, and ultimately register all the data into several internal systems. By leveraging agentic AI, BNY anticipates a transformation in this process. The AI agents would deconstruct the entire task into discrete, manageable parts, execute them efficiently, and then seamlessly integrate the results into a coherent workflow, reducing friction and the potential for human error. According to Pattanaik, the outcome is a faster, more synchronized process that feels almost effortless to the user, allowing human employees to focus on higher-value analytical and client-facing responsibilities.
At the technical forefront of this initiative is Gemini 3, Google’s most recent foundation model, released in mid-November. Unlike earlier models limited to specific input types, Gemini 3 is capable of understanding and processing text, imagery, tables, PDFs, and even audio data in unison. This multimodal capability means that financial professionals at BNY can feed in mixed sets of information—such as regulatory filings, scanned contracts, and customer statements—and have the model interpret, prioritize, and synthesize what truly matters to decision-making. Such an ability has the potential to reshape the way complex financial materials are navigated, summarized, and acted upon within highly regulated environments.
BNY’s journey toward a fully generative AI-enabled enterprise began accelerating in 2023, when the bank intensified its investments in automation and data intelligence. Today, Eliza supports more than one hundred twenty distinct automated tasks, ranging from document handling to data analysis. The institution’s commitment to technological advancement is underscored by the fact that nearly all of its workforce has undergone training not only in generative AI but also in responsible AI practices—ensuring staff understand how to use these tools appropriately and ethically. During the company’s third-quarter earnings call, CEO Robin Vince emphasized that BNY is already operationalizing these technologies at scale, noting the presence of over one hundred “digital employees”—AI-driven agents working directly alongside human staff members on essential tasks such as validating payments, updating code, and performing routine maintenance of internal systems. “We believe our AI opportunity is significant,” Vince told analysts, “and we are pursuing it with urgency.”
This latest partnership with Google follows BNY’s earlier announcement of a collaboration with OpenAI, the developer behind the well-known ChatGPT language model. On its official website, BNY highlights that it was the first major bank to deploy an AI supercomputer—powered by NVIDIA hardware—to augment computational performance in its operations. These moves collectively position BNY as one of the sector’s most assertive early adopters of large-scale AI infrastructure.
However, the rapid deployment of agentic AI in a heavily regulated field like banking inevitably raises pressing questions about governance, security, and data privacy. Both BNY and Google Cloud have stressed that the introduction of such autonomous systems into a tightly controlled financial environment must be guided by rigorous oversight and predefined boundaries. Pattanaik explained that before any agent can go live within the bank’s infrastructure, it must undergo an internal model risk review to confirm compliance with operational and regulatory standards. Access restrictions are carefully defined, determining which data repositories and decision-making processes an individual AI agent is permitted to interact with. Once active, each system is subject to daily monitoring, and insights from this continuous evaluation feed back into refinement mechanisms designed to maintain accountability and transparency.
Rohit Bhat, Google Cloud’s head of financial services, elaborated further on the safety architecture underpinning this initiative. He noted that Google provides built-in safety mechanisms that govern how the agents communicate and the extent of their data exposure. According to Bhat, every agent is equipped with a set of “development kits” and “communication protocols” that explicitly limit interactions between AI entities. In his words, the purpose of these mechanisms is to establish a secure communication pathway defined by strict boundary conditions—essentially ensuring that one agent can only interact with another for specific authorized reasons, and never beyond that scope.
The broader context of this collaboration reflects a wider movement across the financial industry. Major institutions on Wall Street are increasingly deploying a hybrid mix of proprietary and third-party AI solutions. Goldman Sachs, for example, continues to enrich its internal platforms while simultaneously testing innovations from emerging technology startups such as Cognition Labs. Likewise, dealmakers and investment professionals are turning to newer firms like Hebbia, which specializes in deep search and advanced prompt libraries that facilitate more powerful data retrieval. Morgan Stanley, on the other hand, has integrated OpenAI’s technology directly into its advisory workflows to support its financial planners, while executives at JPMorgan have publicly speculated about a near future in which junior employees might direct teams of AI agents as part of their managerial responsibilities.
In this evolving landscape, Google sees financial organizations as ideal testing grounds for agentic AI. Their operations inherently involve large volumes of documentation, strict procedural oversight, and complex risk controls—all conditions that benefit from intelligent automation. Google’s central argument is that Gemini’s strength lies in its ability to think through extensive, multifaceted material while simultaneously adhering to an enterprise’s internal rules. This alignment between reasoning capability and regulatory discipline is essential for any system operating in areas such as custody management, trading markets, or client onboarding. As Bhat summarized, it is critical that “whatever these agents are doing is grounded in the business context and business specificity.” Achieving that grounding, he explained, requires models that not only interpret data but also faithfully respect institutional policies and compliance standards—ensuring that every AI decision meets the stringent benchmarks a financial institution demands.
In essence, BNY’s alliance with Google demonstrates both the potential and the prudence underpinning the next generation of AI adoption in finance. It reflects an industry-wide recognition that automation, when implemented responsibly, can enhance efficiency, accuracy, and innovation without compromising the core tenets of security and trust that define modern banking.
Sourse: https://www.businessinsider.com/bny-ai-boost-google-gemini-3-agentic-ai-system-eliza-2025-12
