BNP Paribas and Google Cloud announced a five-year partnership on September 24, 2026, centered on deploying agentic AI — AI systems capable of carrying out multi-step tasks with limited human intervention — across the bank's Corporate & Institutional Banking division, one of the largest commitments any major bank has made to autonomous AI systems to date.
The initial deployments are specific rather than sweeping: AI agents will help prepare corporate credit memos, support sales, trading, research and structuring teams, and integrate into LLM@CIB, the bank's internal generative AI assistant that already serves more than 65,000 employees. A separate tool, Nickel Assist, will support roughly 200 customer advisers handling payment accounts. Financial terms of the deal were not disclosed.
Multi-cloud, multi-model, and deliberately narrow permissions
The technical backbone runs on Google Cloud's AI-optimized infrastructure, including its Gemini models and Gemini Enterprise deployment framework, but BNP Paribas has been explicit that it isn't betting on a single vendor. "A multi-cloud and multi-model approach lets the bank select the right model" for each specific use case, said Marc Camus, BNP Paribas' chief information officer — a hedge that keeps the bank from becoming dependent on any one AI provider's roadmap or pricing.
Governance is the part of the announcement banks watching this deal will scrutinize most closely. BNP Paribas says the agents operate under enterprise-grade authentication controls and receive access only to the specific resources required for a given task, with all agent connections and interactions monitored, and certain categories of data barred from public cloud environments entirely regardless of the task. Charles Holive, the bank's CIB Chief AI Officer, framed the goal as expanding capability without losing oversight: agentic AI "can help teams take on more complex workflows while keeping appropriate control."
Why banks are moving now, carefully
Google Cloud CEO Thomas Kurian described agentic AI as "a critical shift" in how financial institutions handle operations and risk assessment — language that reflects a broader industry shift away from AI as a chat interface and toward AI as a system that can actually execute parts of a workflow, from drafting a document to flagging a trade for review.
That shift carries real regulatory exposure for a bank the size of BNP Paribas, which is why the deal's governance architecture is arguably more significant than its scope. An AI agent that drafts a credit memo still needs a human to sign off on the lending decision behind it; the value proposition is compressing the time between raw data and a decision a banker can act on, not removing the banker from the decision. Whether that boundary holds as agentic AI expands beyond drafting into more consequential parts of the lending and trading workflow is the question regulators and BNP Paribas' own risk teams will be watching over the life of this five-year contract.
