Skip to content
AI Automation

BNP Paribas Signs A Five-Year Agentic AI Deal With Google Cloud, Starting With Credit Memos - As Sibos Opens On Banks Putting AI Into Production

On 24 September BNP Paribas and Google Cloud announced a five-year partnership giving the bank access to Google's AI infrastructure, Gemini models and Gemini Enterprise, with the first agentic deployments in Corporate & Institutional Banking - starting with agents that help prepare corporate credit memos, then extending to sales, trading, research and structuring. Gemini will also be integrated into LLM@CIB, the bank's generative AI assistant already used by more than 65,000 employees, inside a multi-cloud, multi-model strategy and the bank's existing security and data governance. Days later Sibos 2026 opens in Miami under the theme 'digital finance for AI-driven economies', with the week's argument about how banks move AI, tokenised deposits and Swift's new ledger into production. The credit memo is the right place to start, and here is why.

AlchmAI Editorial11 min read

5 years

Length of the BNP Paribas - Google Cloud partnership on agentic AI and cloud, announced 24 September

65,000+

Employees already using LLM@CIB, the bank's generative AI assistant, which will integrate Gemini models

Credit memos

The first agentic use case, in Corporate & Institutional Banking, before sales, trading, research and structuring

500+

Speakers at Sibos 2026 in Miami, where the theme is digital finance for AI-driven economies

Big-bank AI announcements tend to promise transformation and specify little. BNP Paribas' announcement with Google Cloud on 24 September is more useful because it is specific. The five-year partnership expands the bank's access to Google Cloud's AI-optimised infrastructure, Gemini models and Gemini Enterprise, supporting a move toward broader agentic AI. The first planned deployments are in Corporate & Institutional Banking: AI agents assigned to targeted workflows, beginning with helping teams prepare corporate credit memos, with further applications planned across sales, trading, research and structuring. Gemini models will be integrated into LLM@CIB, the bank's generative AI assistant, which more than 65,000 employees already use.

Two further details matter. The agreement reinforces a multi-cloud and multi-model strategy the bank already operates - Google is a supplier, not a sole dependency. And deployment happens within the bank's established security and data governance framework. Those are the two properties that decide whether a bank's agentic programme survives its first supervisory conversation, and BNP Paribas put them in the press release.

What Makes This A Template For Other Banks

  1. 01Start where review already exists. The memo has a mandatory human reviewer. Agentic AI is safest where it accelerates work that someone accountable already checks.
  2. 02Build on an existing assistant with real adoption. LLM@CIB already has 65,000-plus users. Plugging agents into a tool people already use avoids the most common failure in enterprise AI: a clever pilot nobody opens.
  3. 03Stay multi-model. Adding Gemini to a multi-model strategy keeps the bank able to switch or combine providers - increasingly important as the leading AI labs visibly move together on pace.
  4. 04Govern inside the existing framework. Running agents within established security and data governance, rather than in a separate innovation sandbox, is what lets them reach production.
  5. 05Sequence from drafting to action. Credit memos first, then sales, trading, research and structuring - moving from document assembly toward more consequential workflows only as controls are proven.

Sibos: From Sandboxes To Production

The announcement lands just before Sibos 2026 opens at the Miami Beach Convention Center, running to 1 October with more than 500 speakers across 250-plus sessions. The theme - digital finance for AI-driven economies - and the programme point the same way as the BNP Paribas deal: the conversation has moved from experimental sandboxes to deployment. Citi's Jane Fraser opens, Google Cloud is due to host a roundtable fresh from the BNP partnership, and the running argument of the week is how banks put AI, tokenised deposits and Swift's new ledger into production. One session frames fraud as 'AI versus AI' - the challenge is no longer detecting more fraud but building controls that adapt faster than the fraud itself.

“The most important AI conversation in banking is not about chatbots. It is about which workflows an agent can own end to end, and which controls make that safe.”

What This Means For Banks That Are Not BNP Paribas

The Bottom Line

BNP Paribas and Google Cloud's five-year agentic AI partnership is notable less for its size than for its sequence: credit memos first, then sales, trading, research and structuring; Gemini inside an assistant 65,000 employees already use; a multi-model strategy; and deployment inside existing security and data governance. It arrives as Sibos 2026 makes production AI the industry's central question. The lesson for every bank is that the safest first agent drafts work an accountable person already reviews, in a tool people already use, with a model layer you can change. That is the AI and workflow automation work we do for financial institutions in London, and BNP Paribas has just published a very good template for it.

References & Further Reading

AI AutomationWorkflow Automation AgencyAgentic AIBNP ParibasInvestment Banking Trading ApplicationsClient-Facing & Internal ToolsAI Agency London
Share Email
AI

AlchmAI Editorial

Research and analysis, London

The AlchmAI team writes about the markets, technology and regulation we work with every day. We build trading platforms, real-time charts and AI analysis tools for brokers, prop firms and fintech teams from our office in Mayfair, London. Every article lists its sources. Nothing we publish is investment advice.

This article is general information and commentary. It is not investment advice or a recommendation to buy or sell any investment. Important information

Keep reading

All articles