Beyond Pilots & Sandboxes: Cloud, AI & the Future of Supervision

Speaker A: That subtech supervisory technology is no longer theoretical, it's happening now. Joanne.

Speaker B: Barefoot innovation starts right now.

Speaker C: Hey everybody.

Speaker B: I am excited to share today's conversation with Saket Narayan of Amazon Web Services because it goes straight to one of the biggest questions in financial regulation today. Namely, how can regulators keep pace with a financial system that is becoming faster, more digital, more data driven, and now increasingly is transforming through AI? Sakat leads industry business development for AWS's public sector work in Australia and New Zealand. And before joining aws, he spent more than a decade at a financial regulatory agency in Australia. That background gives him an invaluable perspective on both sides of this challenge, the urgent need for supervisors to modernize, and also the major constraints that regulators face when they try to do so. Saket and Ayers Nikkoch co authored one of our most popular papers called Beyond Pilots and Sandboxes, and that is the launching point for our discussion today. The paper was selected as part of the DC Fintech Week 2025 featured research series. And in today's episode we talk about why so many regulators are now moving from experimentation into production of tech tools and solutions, using cloud, AI, machine learning and modern data infrastructure to detect risk faster, reduce burden, and build the capacities needed for continuous, proactive, adaptive supervision. A key theme of the episode is that this is not just a technology story, it's a culture story. Regulators are, by design, careful institutions. They need to protect the public, safeguard sensitive data, uh, and avoid mistakes. But as uh, Siket explains, cloud technology can sharply lower the risk of experimentation, which means that agencies can now try and pre test ideas quickly. They can learn from what works, and then they can scale successful tools with confidence. We also dig into the data challenge itself. Regulators hold enormous volumes of sensitive information, but much of it is still siloed, fragmented or difficult to analyze at speed. Saket describes how modern cloud platforms can help supervisors build secure scalable data foundations, collaborate safely across teams and institutions and even country borders, and use AI responsibly, all while protecting privacy, security, sovereignty and public trust. Finally, we talk about generative AI and agentic AI. Saket says the era of perpetual AI pilots is over and that regulators are already exploring practical use cases, including supervisory dashboards, anomaly detection, document intake and triage, predictive supervision, policy simulation, fraud and scam prevention, and better stakeholder experiences. For me, this conversation captures the historic inflection point we are in. We will not get the benefits we hope for from financial innovation if regulators cannot keep pace with the change. The financial industry today needs great technology and so do its regulators. Not as luxury, but as essential infrastructure for understanding and overseeing modern markets. So enjoy my conversation with Saket Narayan.

Speaker C: Hey everyone, we have a wonderful show lined up today because my guest is Saket Narayan from aws, Amazon Web Services. Saket, I'm so excited to have you on the show. We've been trying to put this together for a while, uh, so welcome and I would love to start, start by having you introduce yourself, your role at AWS and your background. And we're going to talk about the fact that you and Nick Cook of our team at AHHR, uh, have co authored a paper, uh, and uh, talking about cloud services and the move beyond um, sandboxes and pilots, uh, and where we're headed. So uh, let me ask you to just begin by introducing yourself more fully.

Speaker A: Thank you Joanne, it's great to be here and thank you for having me. Just a little bit about myself. I joined Amazon Web Services AWS a, uh, little over four and a half years ago and currently lead industry business development for public sector in Australia and New Zealand. Before AWS, I spent more than 10 years at the intersection of business and technology in a federal financial services regulator here in Australia. Joanne, that experience gave me a close up view of how regulators manage the rapidly shifting digital financial system and the complexity that comes with it. Over the years I have had the privilege of collaborating with regulators across Australia, the United States, the United Kingdom, Canada, New Zealand, Malaysia, India and Singapore on regulatory and supervisory technologies, regtech and subtech and the challenges. And so Joanne, each of those experiences shaped how I think about supervision in a digital economy.

Speaker C: Wonderful. Uh, I want to start by talking a little bit about the paper again. It's called Beyond Pilots and Sandboxes. We will uh, link to it in the show notes obviously and people can find it on the AIR website. But what uh, prompted your interest in co authoring this paper with Nick Cook and with air?

Speaker A: Sure. So the partnership grew out of what we were hearing and seeing here at AWS and Nick's experiences and the work you do at AEA from regulators worldwide. So Joanne, as you know, financial innovation is accelerating faster than ever. Instant payments, digital assets, embedded finance and new fraud and scam vectors are reshaping financial markets, financial infrastructure and our societies on a daily basis. And so traditional on premises supervisory systems simply were not built for that speed or scale. Many regulators are eager to modernize, but sometimes unsure where to start. And so moving from pilots and POCs to fully operational capabilities in production, that's where our white paper, Beyond Pilots and sandboxes comes in practical insights, global case studies and a framework for demonstrating a bias for action in driving regulatory innovation. As you know, the paper was featured at the DC Fintech Week in October 2025 which reinforced that the questions regulators are asking in Washington, Canberra, London and Singapore are converging. It highlights real world examples like blockchain based anti money laundering tools, cloud Systems Handling, over 100 billion daily market events and generative AI regulatory use cases that show that subtech supervisory technology is no longer theoretical, it's happening now. Joanne M At AWS we often ask uh, which direction is the most customer obsessed? What is the best thing for our customers? And regulators in my view, in their own way ask the same thing. How do we lower regulatory burden while staying resilient and innovative for our uh, regulated stakeholders? So this white paper serves as both compass and a blueprint showing not just what's possible but how to get there.

Speaker C: We observe this journey, uh, regulators everywhere are moving to the cloud. I know a few that are already there and many that are just barely beginning and a lot in between. And same with banks and financial companies. That journey's underway. It would be great um, for you to share your perceptions of kind of where the whole sector is in that journey and um, what are the big challenges they're facing and how are they getting those resolved?

Speaker A: Firstly, momentum is really strong and around the world regulators are using cloud to analyze large volumes of data, market data and other forms of data in real time and to infuse AI driven insights into the regulatory and supervisory business processes. In terms of challenges, if I were to call out three firstly the data scale and speed. So like I touched on earlier, on premise systems cannot handle today's volumes. Cloud elasticity makes it viable and cost effective. Second, building real time supervisory capabilities using on premise systems is a challenge. So regulators are doubling down on cloud based analytics and AI and ML services to spot anomalies across payments or market activity as they occur in real time. Thirdly, new technologies and threat vectors, digital assets and embedded finance require fast moving supervisory processes and the rise in scams and frauds demands real time intelligence sharing across across a number of financial institutions, public sector authorities, law enforcement agencies and ah so Joanne aws has over 150,000 partners around the world including independent software vendors, ISVs who offer specialized solutions for financial crime, anti scam regtechs built on services like Amazon Neptune for Graph analytics or Amazon Bedrock, our AI service. The same dynamic holds for regulated entities, major institutions like Capital One, Goldman Sachs, Commonwealth bank of Australia and National Australia bank have made significant commitments to the cloud, recognizing it as essential infrastructure for modern financial services. So CBA Commonwealth bank of Australia for example, migrated its data platform to AWS to accelerate AI integration, building a scalable platform that connects to bank channels and drives faster, more personalized experiences using data, AI and analytics. The RE platforming completed with HCLTech, one of AWS partners, rigorously tested over 60,000 data pipelines to ensure a secure transition. NASDAQ is similarly innovating with cloud and AI using an AI powered feature in its market surveillance technology to reduce investigation time for market abuse cases by around 33%. So to come back to your question again, the financial services sector uh, is moving at velocity to drive AI and data led business reinvention. And the white paper emphasizes that this isn't just about technology. It requires a cultural shift towards a growth mindset, experimentation and agility.

Speaker C: Yeah, our listeners may know that the culture change piece is something that Nick in particular has worked on and written about it in some of our other papers. We'll link to those in the uh, in the show notes as well. Nick is the former, the former uh, director of the Innovation uh division at the Financial Conduct Authority in the UK and uh, he and our team really are working with regulators all over the world now on how you make the human shifts that need to accompany the technology, uh, modernization, uh, projects that are underway. So actually let's turn to the regulators themselves then. Um, they have a lot of huge amounts of data, more than they can uh, process using traditional methods. Um, a lot of it is very, very siloed and difficult to uh, access and use. So how are you helping regulators maximize the value of their data? And the data question?

Speaker A: Yeah, so we focus on three things. First, scalable data foundations. So Amazon S3 a uh, simple storage service. Amazon S3 alone powers over a million data lakes globally. A decade ago Joanne, only about hundred AWS customers stored more than a petabyte. Now it's thousands. As we celebrate 20 years of AWS, it's worth remembering that Amazon S3 was one of the very first AWS services launching in 2006. And it's now the foundational data layer underpinning both the AI driven world we are talking about and the regulatory data platforms supervisors are building today. As you say, regulators store some of the most sensitive and comprehensive data assets. And finding the needle in the haystack requires cloud scale computing power across petabytes of supervisory information. Second, secure collaboration. Our uh, services let cross functional teams share data and models Safely with fine grained access controls and built in governance. Essential for supervisors working across departments, but also other entities in that industry, sometimes even across the industries. And third, openness and interoperability. So AWS supports open data formats and integration across sources so regulators can modernize analytics while keeping data sovereignty, security and governance are embedded at every layer. Encryption by default, strict controls and features like Amazon SageMaker Catalog that help discover and manage data assets using natural language. So Joanne, a modern secure data platform allows regulators to store, process and analyze data in a governed scaled way to deliver their core supervisory objectives faster.

Speaker C: So you called out security, uh, in that uh, analysis and security is the secret to solving so many problems. As you said before, we need to be able to share data, uh, and make it accessible for a lot of kinds of purposes to a lot of kinds of people. But to do that we have to keep it private, we have to keep it secure. What is paramount in the security challenge?

Speaker A: Great question. Security is job zero at Amazon. The protection of sensitive regulatory data requires a multi layered approach. So let me highlight three critical elements. First, security must be architected from the ground up. The white paper emphasizes that innovation and security aren't trade offs, they must coexist. AWS's infrastructure is designed to meet the highest security requirements, including for financial regulators. And we support 143 security standards and compliance certifications. Second, automation is essential at scale. AWS offers 300 security tools and features, including tools such as AWS Shield Advanced and Amazon Inspector to automate protections and reduce human errors when handling sensitive supervisory data. Third, uh, there is the shared responsibility model. So AWS is responsible for security off the cloud and our customers maintain security in the cloud. And we provide tools like AWS Key management service so customers retain full control over the encryption keys. The white paper highlights secure reading rooms and strict protocols for confidential information. And so AWS supports this with features like fine grained access controls and comprehensive audit capabilities. And so this approach lets regulators innovate while maintaining the highest standards of data protection and public trust in regulatory oversight.

Speaker C: So you used uh, the phrase helping regulators innovate or enabling them to. As you know, the name of AIR is the alliance for Innovative Regulation. As I think you know Saket, I'm a former regulator myself and we know that innovation does not come easily to regulatory bodies. I like to say that's a feature, not a bug. They are designed to be extremely careful and get things right and be risk averse, uh, and all of that is critically important. But it also makes it hard for them to Be agile and innovative. How do you think about that challenge? How do regulators meet the challenge today of being innovative and agile, uh, and still making sure that they're not making mistakes?

Speaker A: Good question. When I think about cloud, cloud makes experimentation low risk. And to your point, we frame that with a simple mental model. One way door versus two way door decisions. And so one way doors are, uh, big irreversible bets that require deep analysis and sign off. Two way doors are small reversible experiments you can try, inspect and iterate on quickly. If they fail, you can shut them down. If they succeed, you can scale them rapidly. The generative AI proof of concept by Australian regulators, as the paper highlights, delivered 93% model confidence for certain use cases in just five weeks with minimal investment. That, to me is a textbook bias for action. Two way door example when it comes to cultivating that culture of experimentation and innovation. Cloud also levels the playing field. So any regulator, big or small, can access advanced AI and machine learning and analytics services on a pay per use basis. So just like at, uh, Amazon, our, uh, two pizza teams, small agile squads of fewer than 10 people, small enough to be fed by two large pizzas, supervisors, are forming empower teams around focused regulatory outcomes, around the challenges, testing, proving and scaling fast. This sort of iterative approach helps regulators move past pilots and POCs to full scale implementations with confidence. So essentially thinking big, working backwards from key business regulatory priorities, experimenting boldly, evaluating rigorously and then scaling what works encouragingly. Joanne, Regulators globally are using tech sprints, hackathons, innovation sandboxes to tackle some of society's most complex and gnarly challenges collaboratively.

Speaker C: Yeah, absolutely. I share your, um, optimism and definitely the enthusiasm. These techniques are transformative. So I want to talk a little bit about generative AI. Even in the time since we put the paper out, actually there have just been so many leaps in the whole area of generative AI, including agentic AI. Um, how are regulators working with those new technologies and doing so responsibly and carefully?

Speaker A: Absolutely. The era of perpetual AI pilots is over.

Speaker C: Yeah.

Speaker A: To accelerate the path to production, um, one of the mechanisms we offer is the Generative AI Innovation center, uh, which essentially it pairs organizations across industries with AWS scientists, strategists and engineers to implement practical AI solutions that drive quantifiable business outcomes. So in 2025, 65% of its projects moved from concept to production, some launching in just 45 days. And to your earlier point, given the breakneck pace of AI development, the line between early and feature use cases blurs. From days to weeks we are seeing customers building AI powered supervisory dashboards and anomaly detection, automated document intake and triage staff productivity, whether it's developer productivity, regulatory staff supervision staff, UM processing document processing information faster through coding assistance. And you mentioned the genti capabilities and also better stakeholder experiences through services like Amazon Connect, which is our cloud based AI infused contact center solution. We also see regulators and other government customers iterating on predictive supervision and risk based targeting, embedding AI into regulatory innovation sandboxes and working with the industry on solving some of these complex challenges. Some of the other use cases around generative AI revolve around policy simulation, brief preparation, fraud and scam prevention and governance focused supervision. So the rapid growth of AI and um, intelligent agents, it brings both promise and new challenges and responsible adoption is crucial. So on responsible adoption we at aws we take a people centric approach that integrates responsible AI across the end to end life cycle, helping customers maintain control of their data. No customer data is used to train our uh, models, ensure transparency, implement appropriate governance and deploy security controls that are specifically designed for generative AI. And finally, as Amazon CTO Werner Vogels says, generative AI lets us generate code in seconds, but if you put garbage in, you get really convincing garbage out. Uh, so the role of data quality cannot be overstated.

Speaker C: Yeah, oh, I'm into that. Ah, yeah, that's great. Um, Saket, uh, is there anything you want to add about the future of cloud and financial regulation? Um, as we stand at this, this uh, inflection moment, with so much changing around us,

Speaker A: the future will be data and AI driven, proactive and adaptive. And so three shifts that we already seeing from periodic to continuous supervision. Again powered by that elastic computing, storage, analytics and AI services, including large language models and AI agents. From reactive to predictive, with AI surfacing risks with human in the loop before they materialize and agents helping regulatory authorities accomplish more with less and from rigid to agile with the approach we have discussed and a small empowered teams driving innovation with a culture of experimentation, culture of learning. And so ultimately cloud gives regulators the same agility as the institutions they oversee. By working backwards from customers, faster risk detection, reduced regulatory burden and impost, and greater resilience, they can modernize confidently. This is just a tech upgrade. Uh, it's a new mindset for contemporary regulation. Balancing innovation with public trust on a modern secure foundation.

Speaker C: Thank you so much for spending time with us and for working with us on the paper. We like to say that for the innovators who are listening and for the regulators who are listening. We'll never get the things we hope for from new technology and innovation in the financial sector if the regulators can't keep pace with the change. They need to understand what's happening in the marketplace. They need to use great technology themselves and they have a critical role to play. And sometimes it's overlooked. I think people underestimate the criticality of it. But, uh, helping them keep, uh, up with this changing world that we're in is just crucial, crucial work. Uh, is there a place you want to cite for people to get more information other than Again, they can come to our air website@regulationinnovation.org for the paper and more information in the show notes. But is there a place, um, they should come at, uh, AWS if they want more information? Well Saket Narayan, thank you so much for being my guest today. It's been fantastic.

Speaker A: Thank you so much for having me, Joanne. I really appreciate the time and the thoughtful conversation.

Speaker B: I really hope you enjoyed today's show. Again, there's much more information, including the transcript@regulationinnovation.org before we close close Let me flag a few things coming up at air on the podcast, we have terrific conversations ahead, including Yahya Fanuzi of aleo, Greg Rupert of finra, and a special live episode that we recorded at the American Bankers association recent Risk and Compliance Conference, the rcc, capturing a session that I did there with Raj Date and Conrad Alt on how AI is changing bank compliance. Our team is also out speaking at several events this summer and fall. Shelley Anderson will speak at the center of Fintech annual event at the University of East London on tokenized money and the future of Finance and regulations. And she'll speak later at AH Perspectives on Financial Inclusion, uh, which is a webinar focused on bringing the Handbook for Women's Financial Inclusion Policy, published by Women's World Banking, to life. I'm going to be speaking at the AWS Summit in Washington, D.C. on July 1, and in October, Nat Weber and I will both be speaking at Money 2020 USA in Las Vegas. As always, you can find full details on Ayres website along with links to our latest papers, events and podcast episodes. Thank you for listening to today. If you enjoyed this show, please rate us on your favorite podcast platform and share it with your friends and peers in the financial ecosystem.