State Street at Sibos 2026
Miami, September 28 – October 1

The financial industry’s push into agentic AI is shifting the focus from experimentation to accountability.

Agentic AI offers more than task-level automation. It gives institutions an opportunity to reimagine entire workflows and modernize processes designed decades ago.

But ambition alone will not deliver results.

Manoj emphasized that a successful transformation begins with a trusted data foundation. Without the right architecture, ontology, and context, even advanced AI investments may produce limited value at significantly higher cost.

“None of the investment that you’re doing in AI will be of any fruitful value without creating that deep layer of foundation and understanding,” he noted.

Most importantly, the path to adoption should balance innovation with control.

Manoj outlined a three-part approach to AI adoption:

1. Empower employees with AI for personal productivity
2. Promote experimentation in a controlled environment
3. Scale proven solutions through strong governance

This structure gives employees room to explore new possibilities while preventing what he described as a “tsunami of agents” operating without informed orchestration. In higher-risk settings, governance must include rigorous evaluation, observability and guardrails aligned with both the criticality of the process and the autonomy given to the agent.

That distinction is particularly important in post-trade operations, where institutions move money and act for clients. AI agents may serve as digital workers, but employees must be able to interpret their actions, examine their reasoning and intervene when needed.

At the center of this approach, he said, is the firm belief that “Trust is something that cannot be negotiated.”

For Manoj, the path forward is clear: AI will deliver its greatest value when operations, technology and control functions evolve together, replacing fragmented automation with a connected operating model built for resilience, accountability, and measurable returns.