BNY is deepening its AI push by teaming up with Google Cloud and wiring the latest Gemini 3 model into its in-house AI platform, Eliza. The goal: turn today’s piecemeal automations into “agentic” AI systems that can take on entire workflows, not just individual tasks.
The partnership, announced on Monday, is the latest sign that Wall Street’s AI race is shifting from experiments and chatbots to production systems that touch core, highly regulated processes.
Eliza gets a Gemini 3 upgrade
Eliza is BNY’s internal AI platform, named after Eliza Hamilton, the wife of bank founder Alexander Hamilton. It already pulls from multiple large language models, but the bank is now embedding Google Cloud’s agentic AI stack, including Gemini 3, directly into that system.
Sarthak Pattanaik, BNY’s chief data and AI officer, told Business Insider that the integration is meant to help employees move faster through everyday work by letting AI agents coordinate complex, multi-step tasks. Instead of just answering questions or drafting emails, these agents can break down a workflow, route each step to the right tools, and return a neatly packaged result.
Gemini 3, which debuted in mid-November, can interpret text, images, tables, PDFs, and audio together. In a bank full of dense documents, that multimodal capability means staff can feed mixed financial materials into Eliza and have the system pull out the critical pieces.
Turning client onboarding into an AI-assisted assembly line
One of the clearest examples is client onboarding, a process that normally involves a long list of manual steps: collecting documents, checking IDs and tax forms, locating key details, pulling risk information, and then logging everything into internal systems.
Pattanaik says agentic AI can orchestrate these steps, splitting the work into smaller components and handing the right pieces to the right agents. The result is a more streamlined flow, with fewer repetitive tasks for humans and a clearer view of what still needs attention.
BNY has been building toward this for a while. Its generative AI rollout accelerated in 2023, and Eliza now supports more than 120 automated tasks across the firm. Nearly the entire workforce has completed training on generative and responsible AI practices, according to Pattanaik.
On a recent earnings call, CEO Robin Vince said the bank is already using agentic AI to deploy over 100 “digital employees” that work alongside staff on jobs like payment validation and code repairs — a hint at how far the automation push has already gone.
A high-profile win in Wall Street’s AI arms race
BNY is far from alone in trying to industrialize AI. Major banks are stitching together a mix of homegrown platforms and outside tools: Goldman Sachs has been expanding its internal systems while experimenting with startups like Cognition Labs, and Morgan Stanley has deployed OpenAI technology to assist its financial advisors. Executives at JPMorgan, meanwhile, have talked publicly about junior staff managing teams of AI agents.
For Google Cloud, landing BNY is a showcase deal for Gemini in one of the most heavily scrutinized industries. Earlier this year, BNY also announced a separate partnership with OpenAI and describes itself as “the first major bank to deploy an AI supercomputer (powered by Nvidia)” on its website, underscoring how aggressively it wants to be seen as an AI leader on Wall Street.
The Gemini 3 integration effectively turns BNY into a live testbed for agentic AI in finance — the kind of environment where success stories can quickly influence rivals.
Safety, guardrails, and model-risk reviews
All of this sits under a tight regulatory spotlight. Embedding AI agents into core banking workflows raises obvious questions about data privacy, model risk, and who is accountable when an automated system makes a bad call.
Both BNY and Google emphasize that the deployment is wrapped in strong governance. Pattanaik says every agent has to clear an internal model-risk review before going live, and agents are constrained by strict access controls that define what information they can see and what they’re allowed to do. Once deployed, performance is monitored daily and fed into a continuous feedback loop to catch issues early.
On Google’s side, Rohit Bhat, head of financial services for Google Cloud, describes a system of “development kits” and “protocols” that governs how agents talk to each other and what data they can access. Those kits are designed to enforce boundaries: which agents can communicate, for what reasons, and under what rules.
The message is clear: agentic AI in a bank isn’t just about clever automation; it has to be auditable, predictable, and tightly fenced in.
Why Google sees banks as ideal testing grounds
Google argues that financial institutions are a natural fit for agentic AI. Banks run on complex, rule-bound workflows and mountains of documentation — precisely the kind of environment where large models that can reason over long, mixed-media inputs can shine.
Gemini’s pitch is that it can digest lengthy materials while staying grounded in a firm’s internal policies. If that holds up in production, it paves the way for agents that can safely operate across custody, markets, onboarding, and other sensitive domains.
But that grounding is also the hard part. As Bhat puts it, the models need to understand not just generic finance, but the specific rules and policies of each institution — and then reliably follow them. That bar will determine how far, and how fast, agentic AI really spreads across Wall Street.
For BNY, plugging Gemini 3 into Eliza is a bet that those agents can move beyond hype and start delivering tangible efficiency gains without crossing regulatory red lines. If it works, expect other banks to follow quickly.