On May 7, 2026, financial services leaders gathered in New York City for the Amazon Web Services (AWS) Financial Services Symposium. Under the theme “Rethink Everything,” the day featured customer-led breakout sessions, immersive demonstrations spanning Banking, Capital Markets, Insurance, and Payments, and conversations about what it takes to move Artificial Intelligence (AI) agents from experimentation to production at scale.

Scott Mullins, Managing Director of Worldwide Financial Services at AWS, opened the day with a keynote that invoked an observation about the transformative power of AI: “Electricity wasn’t just a better way to light a room – it was a way to reorganize every industry on Earth. The main difference is that it took electric light 40 years to do it, whereas today’s technological changes are moving much, much, much faster.”

The evidence that AI is rapidly reshaping financial services was on display throughout the day. And nowhere is the speed of adoption more visible than with AI agents. The complexity and duration of tasks financial institutions can solve with Agentic AI is growing exponentially while the cost to do so is plummeting. Research agents can now propose strategies for traders. Underwriting and lending are becoming more automated. Customer service is becoming more personalized. Payments are moving autonomously between machines.

The keynote challenged the audience with five questions:

1. Do we have the right governance structures to scale AI responsibly?
2. Is our data AI-ready?
3. Are we orchestrating our agents or still working in silos?
4. How is our software development process evolving?
5. Is our workforce ready for the new way that work works?

“Some of these questions are technical in nature, but most of them are not,” he said. “Implementing AI across the enterprise isn’t a purely technical effort. It includes people and process changes that are fundamental to how companies operate. You can’t build tomorrow’s applications and products on yesterday’s dogma and legacy technology.”

To help financial organizations act on these questions and explore what “rethink everything” means in practice, the Symposium focused on three themes. Build foundations for the future examined the governance, security, and tooling required to scale AI responsibly. Re-imagine critical systems tackled the modernization of legacy infrastructure that remains a barrier to deploying AI at enterprise scale. And Elevate every experience showcased how institutions are using AI to change the way they serve customers.

Theme 1: Build Foundations for the Future

The first theme examined whether organizations have the right foundations to scale AI responsibly. Not just the technology, but the governance, security, and development practices around it. Depository Trust & Clearing Corporation (DTCC) took the stage to discuss how it is using Kiro to rethink the entire software development lifecycle and highlighted Kiro’s potential to enable “thinkers to become builders.” The message: before racing to deploy agents, invest in the structures that make scaling them safe. Throughout the day, breakout sessions brought this principle to life, with institutions showing what it looks like to build governance, safety, and responsible AI practices into agentic systems from the start.

• S&P Global: Deployed agentic solutions on Amazon Bedrock AgentCore with an Agent Safety Control system, providing near real-time observability across all deployed agents, enabling teams to detect misaligned behavior and terminate rogue agents before they affect operations. It also implemented a FinOps circuit breaker that automatically constrains unpredictable costs as agent deployments scale.
• Nubank: Serving millions of customers across Latin America, Nubank shared how its Defense IO team processes more than 450 million events daily, making real-time fraud scoring decisions on every transaction while maintaining high availability. Operating with a lean team, Nubank showed that fraud detection at scale does not require massive headcount when the right AI and machine learning infrastructure is in place.
• WEX: This B2B FinTech, which handles over 40,000 tickets annually across fleet mobility, healthcare benefits, and corporate payments, shared how it moved from a virtual engineer concept to a full agentic platform on Amazon Bedrock AgentCore. AgentCore’s identity, memory, gateway, and observability primitives allowed WEX to democratize agent development safely across the organization.

Theme 2: Re-imagine Critical Systems

The second theme tackled the modernization of legacy infrastructure that remains a barrier to deploying AI at enterprise scale. As the keynote noted, “AI applications are still applications. Like all applications, they require actionable data, strong security, comprehensive resiliency, and of course high availability.” The Hartford took the stage to discuss how AWS Transform is compressing years-long modernization timelines into months. Across breakout sessions, customers showed what this modernization journey looks like in practice and the new capabilities it unlocks.

• Itaú Unibanco: Latin America’s largest bank, migrated a 50-year-old mainframe-based bank statement platform to a cloud-native architecture, achieving approximately 80% operational cost reduction and a drop in response times from 2 seconds to 191 milliseconds. Built on 30 Amazon OpenSearch clusters with 285,000 shards serving 70 million customers, the platform unlocked natural language search powered by Amazon Bedrock, enabling customers to query their statements conversationally.
• Transamerica: Optimized tens of thousands of core nodes used for actuarial modeling by migrating to purpose-matched chipsets and co-locating workloads on a low-latency backbone, cutting total infrastructure costs by 50%. Then built an agentic AI layer on top, including an agent that bridges legacy mainframe semantics and modern Oracle ERP, translating between old and new systems so teams can transition without losing institutional knowledge.
• Gemini: The cryptocurrency exchange cut tick-to-trade latency by 80% using AWS Outposts and Local Zones. Evolving from a crypto exchange into a broader markets company with its own clearing house, Gemini deployed infrastructure at the edge to support high-frequency trading and sub-second execution.
• HCL Tech and Fiserv: Fiserv, which processes $10 billion daily across 25,000 transactions per second, presented its approach to migrating a core mainframe using agentic AI-powered refactoring – what it described as “changing the engine of a flying plane.” Key drivers: shrinking mainframe talent pool, demand for innovation velocity, and globalization requirements.
• Fifth Third Bank: Shared its migration to Amazon Connect Customer and deployment of conversational AI to improve Net Promoter Scores and reduce average handle time.

Theme 3: Elevate Every Experience

New technologies are giving financial institutions the opportunity to enhance how they serve customers by making service more personalized, financial advice proactive rather than reactive, and enabling payments to move autonomously. U.S. Bank discussed how it is reimagining customer experience with Amazon Connect Customer. It unified hundreds of contact centers serving millions of customers and advanced toward agentic self-service that resolves inquiries across voice, chat, and SMS without human intervention. Breakout sessions showed what elevating every experience looks like in practice.

• Rocket: Built a unified data foundation that connects context across the customer journey thus reducing the siloed, disconnected experience that has historically defined home buying. This led to a 20% increase in the refinance pipeline, 10% conversion lift, and three times industry-average recapture rates. When Rocket acquired Mr. Cooper, a large mortgage servicer in the US, that same foundation enabled 40,000 leads to be onboarded within nine days, with the first client closing in just three. Rocket’s VP of Engineering stated that unified customer understanding is far more important than isolated AI initiatives, because the shared context of business entities fuels AI.
• Coinbase: Framed AI agents as a new client archetype that financial infrastructure must be prepared to service. It introduced the x402 protocol, an open web standard now under the Linux Foundation, that enables agents to pay websites per request using stablecoins, creating the foundation for enterprise-ready agentic commerce with sub second, sub cent transactions purpose-built for machine-to-machine payments.
• Hazel by Altruist: Wealth advisors spend 45% of their time on operational tasks across 8–10 software solutions. With a projected advisor shortage by 2034 and 60–70% of heirs leaving their parents’ advisor, that’s unsustainable. Altruist presented a purpose-built AI platform that automates the operational work, freeing advisors to focus on relationships that matter.
• Insurance Panel (Baldwin Group, Ameriprise, Hudson Insurance): In a panel moderated by Deloitte, insurance leaders discussed where AI creates the biggest impact in brokerage, wealth, and specialty lines. Baldwin Group shared that by reimagining existing workflows rather than automating them as-is, it rolled out a new insurance product in three days; a process that normally takes three to five months. Ameriprise described building a data foundation that serves both AI and operational workflows, while Hudson Insurance emphasized an MVP-driven approach: win fast, lose fast, iterate in one-to-two-month cycles to beat larger carriers to market.

Launch Announcement: Amazon Bedrock AgentCore Payments

Announced at the Symposium, Amazon Bedrock AgentCore Payments represents the first managed end-to-end payment capabilities purpose-built for autonomous AI agents. Built in partnership with Coinbase and Stripe, AgentCore Payments enables agents to discover, access, and pay for web content, APIs, Model Context Protocol (MCP) servers, and other agents using stablecoin micropayments and the x402 protocol. The infrastructure layer enforces spending limits deterministically, giving enterprises the control to let agents transact autonomously without the risk of runaway costs.

A subsequent session, called “How AI Agents Pay” provided additional context for why this matters now: over 51% of web activity is now bot-driven, and companies building wallet infrastructure are seeing significant growth in agent-managed wallet creation, a figure that was essentially zero a year ago. The conversation has shifted from whether agents can make payments to how enterprises govern, audit, and control those payments. AgentCore Payments is available in preview in US East (N. Virginia), US West (Oregon), Europe (Frankfurt), and Asia Pacific (Sydney).

Conclusion

Cutting across all three themes, leaders from TCS, Anthropic, and CardWorks shared a unifying insight: moving from Proof of Concept (POC) to production remains the industry’s defining challenge. Citing McKinsey’s 2025 data that 88% of organizations have implemented AI in at least one line of business, they noted the top blockers are integration complexity, AI-ready data, cross-functional alignment, and centralized governance. Anthropic recommended starting with low-complexity, high-impact wins and appointing a single-threaded AI leader to build confidence before scaling. Their message reinforced what every session showed: the gap between experimentation and production is not closed by technology alone; it requires the governance, infrastructure, and organizational change the Symposium’s three themes addressed.

Three takeaways emerged from the day: (1) Governance and safety are prerequisites, not afterthoughts. Financial institutions reinforced that scaling agents requires purpose-built observability and controls from day one. (2) Modernization unlocks AI. Institutions are proving that migrating legacy systems isn’t just about cost reduction; it’s about creating the foundation for capabilities yesterday’s architecture cannot support. (3) Unified data powers everything, and across sessions, institutions showed AI puts superpowers in the hands of organizations that can bring together a complete view of their customer, and that a shared data foundation across the business is more valuable than any single model.

The AWS Financial Services Symposium continues in Sydney (August 4) and London (September 30) and we’re excited to see what customers across these regions are building and invite you to join the conversation. Save the date for the Sydney Symposium.

Learn More

• Watch AWS Financial Services New York City Symposium 2026 session recordings on the AWS Events YouTube channel.
• Visit aws.amazon.com/financial-services or get in touch with an AWS Financial Services expert.
• Follow us on LinkedIn.