“Our treasury manages US$100bn and we’ve got a US$50bn foreign exchange (FX) hedging programme,” explained Jason Allen, Assistant Treasurer at Google, addressing keen breakfast pre-Sibos meeting attendees at the National Hotel’s Oval Room. Hearing from Alphabet Inc – the parent company to Google – was evidently enough to get a crowd up early.

“Google’s treasury operates across 60 different countries, so we have multiple currencies and bank accounts. We handle 100,000 transactions per day, which is on a par with a European bank,” Allen continued. “Crucially, our treasury needs to show that we are investing our excess cash, as well as making it available to the business.”

AI can help with this obligation, “principally via enhancing our cash flow forecasting (CFF) engines,” said Allen. “This helps us determine the amount of excess [investable] cash. But AI can also help in other ways, such as advising us who best to invest it with and so on.”

Liquidity optimisation can also benefit from AI being deployed on smart routing payment applications or risk mitigation identification and FX hedging end uses.

What’s more, multiple AI use cases now exist in financial services (FS), from compliance and anti-fraud applications to operational efficiencies. As Allen explained, “We’ve saved a ton of time thanks to AI automation.”

AI shift demands security

Kristin Reinke, Google’s VP of Finance, took over speaking duties next, stressing the safety and governance angle. This was much debated at Sibos 2026 after OpenAI announced that it wasn’t releasing its latest GPT-6.1 Astra system after some of its AI agents had gone rogue.

“Each of our managers has full transparency and control over agents. Judgements ultimately remain with the human. Our powerful AI tools are about augmenting human capabilities,” said Reinke.

She additionally promoted Google Cloud’s Gemini Enterprise for FS agentic AI platform, which was unveiled just before Sibos 2026. It’s designed to automate complex corporate banking and capital markets workflows.

“A shift of this magnitude as AI is rolled out is just as much cultural as it is technological,” concluded Reinke. In other words, how we train and deploy AI is crucial.

Banks share their AI perspective and uses

A four-person panel of bankers moderated by Georgina Bulkeley, Managing Director of Google Cloud, also shared their AI end use stories and perspectives. The panel included:

• Sarthak Pattanaik, Chief Data & AI Officer, BNY

• Joanne Hannaford, CIO and CPO, Corporate Bank, Deutsche Bank

• Stephen Randall, Interim Services COO and Head of Liquidity Management Services, Citi

• Charles Holive, Chief AI Officer, Corporate & Institutional Banking, BNP Paribas

Holive answered Treasury Today’s question about how the panel viewed OpenAI’s recent GPT-6.1 scrapping announcement by commenting that AI is a new risk, so it was good to see caution. “I am waiting for better governance, not models, so I am very happy to see this,” he said.

Holive also outlined some BNP Paribas AI end uses, especially on trade processing, where the bank has done a core redesign with AI to improve efficiency and performance. “An experimental agent in this area was initially terrible,” he said, adding that once the agent had been refined, it learned what was required in just a couple of weeks.

“We’ve definitely gone from experimentation to production,” concurred Deutsche Bank’s Hannaford. The private bank unit’s launch of an agentic AI-enabled Source of Wealth (SoW) solution just before Sibos is proof of that.

In her own corporate banking department, Hannaford trialled Deutsche Bank’s new agentic AI framework that she is confident will provide safe, multipurpose agents that can provide good use cases.

“It is named Ada, after Ada Lovelace, one of the founders of computer science,” she explained. “As a corporate bank we are prone to macro and geopolitical issues adversely impacting our clients who expect us to help with compliance, trapped cash, FX or whatever concerns they have on client calls. AI can help our staff with this ahead of time.”

The two crucial FS end uses that Hannaford sees for AI are process optimisation and financial intelligence. “We’ve co-developed a financial insights tool [with Google] that is launching in a month,” she said. “Rapid development is possible with strategic partners.”

Citi’s Stephen Randall shared its Citi Sky use case, built using technology from Google Cloud and DeepMind. It’s a wealth management AI tool designed to help clients in the US access market insights, act on opportunities, and engage with their advisors faster in a ‘personalised’ way based on their data. Citi’s own agentic AI framework, named Arc, was also mentioned.

“Sharing use cases in this way across and between organisations accelerates adoption,” added Randall, while discussing a liquidity solution. “Document processing and onboarding can additionally be improved with AI. There are great opportunities.”