BNP Paribas Advances Agentic AI Deployment in Banking Through Google Cloud Partnership

Agentic AI is an advanced form of artificial intelligence that allows systems to independently manage and complete complex, goal-oriented tasks across multiple steps in professional environments.

Executive Summary

BNP Paribas has established a significant presence in the European banking sector and now seeks to enhance its operations through advanced AI technologies. The five-year partnership with Google Cloud represents a strategic move to incorporate agentic AI into key banking processes. This collaboration aims to improve efficiency in the Corporate & Institutional Banking division by leveraging the capabilities of Gemini models. The deployment targets the existing LLM@CIB platform, which provides generative AI assistance to a large workforce. By integrating Gemini, the bank expects to handle more complex tasks in credit analysis and trading activities. The move aligns with the bank's commitment to using AI where it delivers clear business value.

At the Nickel subsidiary, the Gemini technology is already in use for customer support functions. This provides an early example of how the AI can assist advisers in their daily interactions with clients. The overall strategy includes careful consideration of data security and governance requirements. The partnership supports deployment of agentic AI use cases in Corporate & Institutional Banking including AI agents for corporate credit memo preparation and workflows in sales, trading, research, and structuring.

This initiative builds on prior deployment at the bank's Nickel subsidiary, where it supports 200 customer advisers. The bank maintains a strategy that combines multiple technology providers to meet diverse operational needs. The decision to partner with Google Cloud comes amid broader industry trends toward incorporating advanced AI in financial services.

Background and Context

The announcement of the partnership occurred on September 24, 2026, marking a new phase in BNP Paribas' technology adoption. Prior to this, the bank had developed its own internal AI tools and maintained relationships with various providers. The multi-cloud approach allows the bank to select the best tools for specific needs without relying on a single vendor. Data sovereignty remains a key concern for financial institutions dealing with sensitive customer information. BNP Paribas has decided to exclude certain data categories from public cloud environments. This decision reflects regulatory requirements and internal risk management policies.

The bank continues to use on-premise infrastructure for critical operations. Marc Camus, the Group Chief Information Officer, emphasized that the Google agreement does not signal a complete shift away from existing setups. Charles Holive has highlighted that technology selection is driven by business requirements rather than a one-size-fits-all approach. This ensures that AI deployments fit within the established security framework. The partnership reinforces BNP Paribas' multi-cloud and multi-model strategy within its existing security and data governance framework.

Certain data categories are excluded from public cloud to maintain control. The bank plans to retain on-premise infrastructure and continue partnerships with other providers like Mistral AI for different applications. This hybrid model supports the bank's leadership in banking while adopting new technologies selectively.

What's New in Detail

The new partnership introduces agentic AI capabilities that go beyond traditional generative models. These agents can assist in preparing corporate credit memos by gathering and analyzing relevant data from various sources. The integration into LLM@CIB will make these tools accessible to the broad employee base of more than 65,000. Workflows in sales and trading will benefit from AI support in research and structuring activities. This can help teams manage more intricate processes with greater consistency.

The focus on agentic features allows for more autonomous operation within defined parameters. The deployment at Nickel serves as a proof point for the technology's effectiveness in a customer-facing role. Nickel Assist supports 200 customer advisers in handling inquiries and transactions. This initial success informs the rollout in the larger CIB division. The agreement expands access to Google Cloud's AI-optimized infrastructure and includes Gemini Enterprise features that support enterprise-grade deployments.

These elements are designed to meet the high standards of the banking industry. Initial use cases include plans to integrate Gemini models into LLM@CIB for the specified workflows. The partnership supports deployment of agentic AI use cases in Corporate & Institutional Banking.

Technical Specifics

Gemini models from Google Cloud provide the foundation for the agentic AI applications. These models are integrated into the bank's existing LLM@CIB system to enhance its functionality. The approach allows for seamless incorporation without disrupting current operations. The multi-model strategy involves using different AI providers for various applications. Partnerships with Mistral AI continue alongside the new Google Cloud arrangement.

This diversity helps mitigate risks associated with dependency on any single technology. Security and governance frameworks are central to the deployment. Certain sensitive data remains excluded from public cloud storage. The bank ensures that all AI uses comply with internal policies and external regulations. Agentic AI in this context involves systems that can execute workflows with minimal supervision.

For example, in credit memo preparation, the AI can compile information from multiple sources and suggest content. Human oversight remains in place for final approvals and complex decisions. The five-year agreement expands BNP Paribas’s access to Google Cloud’s AI-optimized infrastructure, Gemini models and Gemini Enterprise.

Key Changes in AI Deployment Strategy at BNP Paribas

Aspect	Before Partnership	After Partnership
Primary AI Providers	Internal tools, Mistral AI	Internal tools, Mistral AI, Google Cloud Gemini
LLM@CIB Access	Available to more than 65,000 employees	Enhanced with Gemini integration for more than 65,000 employees
Subsidiary Example	Not specified	Nickel Assist supporting 200 customer advisers
Data Handling	On-premise and multi-cloud	On-premise retained, sensitive data off public cloud
Use Cases	Existing generative AI	Agentic AI for credit memos, sales, trading, research, structuring

Market & Stakeholder Implications

For executives in the banking sector, this partnership illustrates the potential for targeted AI adoption without full cloud migration. Other institutions may consider similar hybrid approaches to balance innovation with compliance. The quantified scale of LLM@CIB usage demonstrates the readiness for broad AI integration across large organizations. The focus on CIB workflows suggests opportunities for productivity gains in high-value areas such as credit analysis and trading support.

This can lead to faster decision-making and improved service delivery to clients. However, the bank has not released specific metrics on time savings or revenue impact at this stage. Stakeholders including clients and regulators will note the emphasis on data sovereignty. This can build trust in the bank's handling of information. The retention of on-premise capabilities addresses concerns about data location and control.

The strategy supports the bank's position as a leader in European banking by adopting new technologies responsibly. Peer banks can learn from the selective integration of agentic AI into existing platforms. The approach prioritizes business needs in technology decisions.

1. Assess business needs first when considering AI deployments to ensure alignment with operational goals.
2. Retain flexibility through multi-cloud and on-premise infrastructure to address data sovereignty concerns.
3. Scale successful pilots from subsidiaries to core business divisions for broader impact.
4. Continue diversification of AI providers to avoid over-reliance on any single partner.
5. Monitor and maintain governance frameworks to support compliant AI usage in regulated industries.

Expert Reactions

Thomas Kurian, CEO at Google Cloud, described the partnership as a way to combine banking leadership with secure infrastructure for intelligent agents. The statement highlights the potential for agentic AI to transform financial institution operations.

Agentic AI marks a critical shift in how global financial institutions streamline operations, assess risk, and serve their clients. Through this five-year partnership, BNP Paribas is combining its banking leadership with Google Cloud’s secure infrastructure and Gemini Enterprise to deploy intelligent agents within platforms across the bank.Thomas Kurian, CEO at Google Cloud

Charles Holive, Chief AI Officer at BNP Paribas CIB, stressed the importance of starting with business needs when deploying AI technologies. This ensures appropriate control levels are maintained during complex workflow handling. The approach helps teams manage more complex workflows while maintaining the appropriate level of control.

Marc Camus, Group Chief Information Officer at BNP Paribas, clarified that the agreement does not eliminate on-premise infrastructure. This statement reassures internal teams and external observers about the continuity of existing systems. The fact that the bank signed the agreement with Google does not mean it will get rid of its on-premise infrastructure.

What's Next

The bank plans to continue expanding the use of agentic AI across additional workflows in the coming years. Further integration of Gemini models will occur as use cases are validated through testing and feedback. The five-year duration of the partnership provides a framework for sustained collaboration and development. Retention of on-premise infrastructure will remain a priority for sensitive operations.

This hybrid model allows the bank to leverage cloud benefits where appropriate while keeping control over critical assets. Additional partnerships may be explored to diversify the technology stack further. The success at Nickel provides a model for future deployments in other subsidiaries or divisions. Scaling the technology will require ongoing attention to governance and performance monitoring.

The overall strategy supports the bank's position in the competitive banking landscape. Executives at peer institutions should observe how BNP Paribas balances innovation with risk management. The multi-model approach offers lessons in avoiding vendor lock-in. Continued focus on business-driven AI adoption can guide similar initiatives elsewhere in the financial sector.

The partnership is expected to yield improvements in operational efficiency over time as the AI agents are refined. The integration into LLM@CIB positions the bank to handle increasing volumes of complex tasks with AI assistance. This development reflects the evolving nature of AI in enterprise banking environments.