Sibos 2026 is taking place in Miami, Florida, from 28 September to 1 October 2026 under the theme ‘Digital finance for AI-driven economies’
Financial institutions will need to combine AI with their own data, institutional knowledge and governance as the technology becomes more widely adopted, according to Jacqueline O’Flanagan, head of financial services Americas at Microsoft Frontier Co, who is speaking at Sibos on 30 September 2026.
Ahead of her appearance at Sibos 2026, O’Flanagan argued that access to AI models alone will not provide a lasting advantage for banks and other financial institutions. Instead, organisations will need to apply AI safely across the processes through which they manage risk, move money and serve customers.
She identifies three factors that will underpin this approach: institutional reputation, proprietary intelligence and the ability to make that intelligence available to employees, clients and partners.
Microsoft describes the next stage of AI adoption as connecting context, execution and control.
Context includes institutional and market data, relationship histories, risk signals, documents and operational knowledge. Execution involves using AI agents and applications to coordinate work across workflows, rather than simply answering individual questions. Control covers areas including identity, permissions, observability, auditability, resilience and human accountability.
“Microsoft’s role is to provide an integrated foundation that helps financial institutions retain control of their data and compound their intelligence across cloud, data, AI, security, and collaboration,” said O’Flanagan.
She points to several existing deployments as examples of how this approach could be applied.
Global financial services company BNY uses its Eliza platform to bring together AI tools, agent-building capabilities, governance and learning resources. Microsoft says the platform now supports more than 300 solutions across the institution. AI-assisted onboarding at BNY is 20 per cent faster, while more than 10 per cent of client settlement enquiries are resolved or assisted by AI, with processing 80 per cent faster. Digital employees also handle more than 10 per cent of payment-repair activities globally.
“The opportunity is not just to automate work, but to also deepen collaboration through our partner ecosystem and embed Azure intelligence directly into the processes that drive operational performance,” said Deanna Lanier, global head of strategy, data and analytics at BNY.
LSEG is using Microsoft Fabric to bring systems and datasets together on a governed platform. Microsoft says this has improved data quality and reduced the time needed to develop new products from years to months.
At insurance company Genworth, Microsoft 365 Copilot is being used in investment operations to help construct complex portfolio trades. According to Microsoft, the technology has reduced the time required for the process from days to hours, while final decisions remain with employees.
O’Flanagan also points to Brazilian bank Bradesco’s use of its AILA system to support audit planning. The technology is intended to help auditors at the bank focus more on analytical work and business risks, rather than routine tasks.
Microsoft says these examples point towards a human-led, agent-operated model for financial services, rather than fully autonomous systems. Under this model, AI agents could gather evidence, reconcile information or prepare responses, while people retain responsibility for decisions requiring professional judgement or carrying significant financial or regulatory consequences.
“An agent can gather evidence, reconcile information, or propose a response,” said O’Flanagan. “A person keeps authority where professional judgment, fiduciary obligations, or consequential decisions require it.”
This means that approval thresholds, escalation processes and the ability to stop an action need to be incorporated into AI systems as they are designed, rather than added after deployment.
Microsoft also sees an opportunity to connect traditionally separate parts of financial services. In banking, for example, information gathered during client onboarding could, subject to appropriate permissions, support treasury services or liquidity discussions.
In capital markets, research, pre-trade analysis, execution, post-trade processing and custody could similarly become more connected.
O’Flanagan is due to discuss the move from AI pilots towards “AI-powered markets” at a Capgemini panel at Sibos 2026 on 30 September, alongside representatives from BNY and Société Générale.
Read O’Flanagan’s ‘Banking on intelligence’ blog post on the Microsoft website.