Sibos 2026: Google Cloud reveals banks are moving from AI copilots to autonomous workflows

What happened

At Sibos 2026 in Miami, Google Cloud convened a breakfast roundtable at The National Hotel to discuss the next phase of artificial intelligence adoption in banking. The session brought together senior technology and data leaders from some of the world's largest financial institutions, including BNY, Deutsche Bank, BNP Paribas and Citi, alongside Google Cloud and Google finance executives.

The central theme, as framed by Google Cloud's financial services team, was a shift underway in how banks deploy AI: moving away from copilot-style tools that assist individual employees with discrete tasks, towards autonomous workflows in which AI systems execute multi-step processes with less direct human intervention. Participants included Georgina Bulkeley, director of financial services solutions at Google Cloud; Kristin Reinke, VP of finance at Google; Jason Allen, assistant treasurer at Google; Sarthak Pattanaik, chief data and AI officer at BNY; Joanne Hannaford, CIO and CPO of Deutsche Bank's corporate bank; Charles Holive, chief AI officer of corporate and institutional banking at BNP Paribas; and Stephen Randall, interim services COO and global head of liquidity management services at Citi.

The discussion centred on purpose-built AI for financial institutions — tools and infrastructure designed specifically for banking use cases rather than generic AI applications — and what it takes operationally and organisationally for banks to progress from experimentation with copilots to running autonomous, agentic workflows in production.

Why it matters

The framing signals where enterprise AI strategy in banking is heading next. Copilots — chat-based assistants layered onto existing systems — have been the low-risk entry point for most institutions over the past two years. Autonomous workflows represent a materially different proposition: AI systems that can initiate, sequence and complete processes such as reconciliation, liquidity management or client servicing tasks with reduced human checkpoints. That shift raises the stakes on governance, data quality, risk controls and change management, and it changes the skills and oversight structures banks need internally.

For institutions the size of BNY, Deutsche Bank, BNP Paribas and Citi, the move from assistive to autonomous AI also has direct implications for operating costs, employee roles and, ultimately, the consistency and speed of the experience delivered to corporate and institutional clients. Purpose-built infrastructure — rather than generic AI tooling retrofitted to banking — is being positioned as the precondition for making that leap safely at scale.

The Renascence take

The language of "copilots to autonomous workflows" is easy to nod along to, but it understates how much organisational work sits behind that transition. The technical capability to automate a workflow is rarely the constraint; trust, accountability and the redesign of human oversight are.

Most banks will find that the hard part of going autonomous isn't the model — it's deciding who is accountable when an AI-run process makes a judgement call a human used to make, and how that's explained to a client or a regulator after the fact. Institutions that treat this purely as a technology upgrade will stall at the pilot stage; those that redesign roles, escalation paths and client communication around autonomous decisions in parallel with the technology are the ones likely to move workflows into production responsibly. The real signal to watch from Sibos isn't the AI capability itself, but whether banks are willing to renegotiate internal accountability structures fast enough to keep pace with it.