In this AWS Financial Services Symposium interview, PwC leaders Justin Guse, director for cloud and digital transformation, and Sud Saxena, principal for cloud and AI in financial services, join theCUBE Research’s Scott Hebner to share how PwC and AWS are modernizing contract processes in highly regulated banking environments. They unpack a real-world wholesale lending use case where AI and cloud-native architectures cut contract approval cycles from months to weeks, directly improving client experience and revenue impact.
Saxena explains how PwC combined deterministic machine learning models with generative AI to identify key contract clauses, reduce negotiation friction and ensure compliance through human-in-the-loop oversight. Guse dives into the AWS-powered architecture, detailing how services like Amazon SageMaker, Bedrock, serverless and event-driven designs delivered scalability, security and faster development. The discussion also explores trust as a top priority, balancing automation with explainability, regulatory guardrails and client confidence.
Looking ahead, the conversation touches on emerging opportunities in knowledge management, customer care and software development lifecycle (SDLC) transformation, as well as the growing role of agentic AI in orchestrating multi-agent workflows for complex, compliance-heavy industries. Both guests emphasize starting with targeted use cases, iterating methodically and tackling complex problems early to capture measurable business value.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
AWS Financial Services Symposium 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For AWS Financial Services Symposium 2025
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for AWS Financial Services Symposium 2025.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
AWS Financial Services Symposium 2025. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to AWS Financial Services Symposium 2025
Please sign in with LinkedIn to continue to AWS Financial Services Symposium 2025. Signing in with LinkedIn ensures a professional environment.
At the AWS Financial Services Symposium, theCUBE’s Dave Vellante sits down with IBM’s Global Managing Partner, Glenn Finch; Prudential’s CIO, John Glooch; and Truist’s SVP, Steven Allen, to unpack how cloud-native architectures and generative AI are transforming financial services. From core system modernization to pragmatic innovation, the trio shares real-world examples driving measurable impact.
Finch highlights IBM’s deep collaboration with AWS to build agentic AI solutions that are not just experimental, but production-ready – empowering institutions to move from legacy infrastructure to cloud-first operations. Glooch and Allen echo the importance of data trust, agile transformation and customer-centric use cases, such as intelligent claims and fraud detection, that are already proving to be of value.
The panel also dives into emerging trends – from tokenization and secure payments to regulatory readiness and AI ethics. With a candid take on skill gaps, cultural shifts and future-defining tech bets, the discussion offers a grounded look at how financial institutions are balancing innovation with compliance to drive long-term, scalable growth.
In this AWS Financial Services Symposium interview, PwC leaders Justin Guse, director for cloud and digital transformation, and Sud Saxena, principal for cloud and AI in financial services, join theCUBE Research’s Scott Hebner to share how PwC and AWS are modernizing contract processes in highly regulated banking environments. They unpack a real-world wholesale lending use case where AI and cloud-native architectures cut contract approval cycles from months to weeks, directly improving client experience and revenue impact.
Saxena explains how PwC com...Read more
exploreKeep Exploring
What is the joint value that you are delivering through your partnership with AWS in the financial services sector?add
What has been observed about the relationship between technology problems in financial services and other industries?add
What was the business problem described regarding contract approval cycles in the context of business to business financing?add
What are the key considerations and requirements when selecting and implementing a model for a project in financial services?add
What factors contribute to building trust in AI systems while balancing automation and human involvement?add
What emerging technologies are expected to have a significant impact in the near future?add
>> Hello, and welcome back to theCUBE's continuing coverage of the AWS Financial Services Symposium 2025 in New York City. I'm Scott Hebner, and thank you for tuning in. Today, I'm thrilled to be joined by Justin Guse, the Director for Cloud and Digital Transformation at PwC, and Sud Saxena, the PwC Principal for Cloud and AI that's focused on the financial services industry, both of whom are tightly aligned with AWS as a strategic partner. So Sud, Justin, thanks so much for being here today. I really appreciate you joining.
Justin Guse
>> Thanks for having us.
Sud Saxena
>> Thank you, Scott.
Scott Hebner
>> So, let's just start off with the big picture here. Can you just bring us a little bit through the history of PwC and AWS in terms of the partnership, particularly what you're doing in the financial services industry? And what's the joint value that you guys are delivering?
Sud Saxena
>> Thank you, Scott. So our history has been over 10 years long, in terms of partnership with AWS. And as you might remember, financial services was a late adopter of cloud, so they were not the first movers. The other segments like retail and media were the first movers. But financial services has gotten into the cloud journey now and quite a bit now from an AI standpoint as well. So from a financial services standpoint, I would say over the last six years or so, we've had quite a bit of traction with our FS clients when it comes to cloud and AWS. And our value prop is really complex initiatives, whether it's complex modernization efforts, large cloud migrations, or some of the AI projects, one of which we are going to talk about today, because these are the ones that really require differentiated capabilities and skills that PwC brings to bear. AWS has a pretty strong technology stack that stands well when it comes to regulated sectors like financial services.
Scott Hebner
>> Yeah, the one thing I was just going to say was, one thing I've certainly learned over my career is, what happens in financial services often happens elsewhere. It's highly regulated, very dynamic, changes all the time, very complex, Sud, as you pointed out. And if you can solve technology problems in financial services, it's the recipe for solving them in other industries. So, it's the place to go to really understand what's happening with technology and what's possible, that's certainly what I've observed over the years. Right? So, you guys have been working with AWS with a financial services firm, one of the larger ones, leading ones, to reinvent their contracting process. Can you just tell us a little bit more about that?
Sud Saxena
>> Yeah, so this one interestingly was in the business unit that does wholesale lending, business to business, finance, working capital solutions. So when you look at business to business contracts, whether it's in the lending domain or whether it's in any other domain, cycle time is always an issue because you're talking about two large customers coming to agreement or two large corporates coming to mutual agreement on terms. And for a bank, the faster you close a contract, the faster you make money. So the business problem here was really about how to cut the contract approval cycle for these business to business financing contracts, and also how to improve the customer experience. So prior to this product being in place, the entire contract negotiation happened over email and there was a lot of friction because of the nature of the negotiation, the back and forth, and then aligning on terms. So the hypothesis was, can AI be leveraged to really improve the cycle time when it comes to contract approvals and improve the customer experience as well?
Scott Hebner
>> So they were basically passing documents back and forth in email, doing the editing?
Sud Saxena
>> Right, yeah.
Scott Hebner
>> Yeah, very inefficient approach.
Sud Saxena
>> Which is still typical in most companies, that is not something unheard of it. It is the regular order of business, I guess.
Justin Guse
>> And to bring it back around, to your point, Scott, right? You just said many things you do with financial services could be applied to other industries as well. Same problem probably in many other industries as well, right?
Scott Hebner
>> Yeah. I would imagine the vast majority of businesses predominantly do contracts in that fashion. They might have a system where the contracts are sitting, but all the negotiation and all the editing and goes on, yeah, you're right, through email, and that's always been my experience, but that's a great example of where there's lost productivity. Right? You could be doing other things, things more efficiently, getting additional help, and this idea of applying the superpowers of AI to help streamline that makes a ton of sense, it really does. One of the things I know you guys had to do to make this all work, particularly in the financial services industry, is complement the LLMs to overcome some of the challenges that generative AI has in terms of trustworthy, deterministic outcomes. It's clear that's what you're going to want to have when you're kind of going through this kind of process, right? How do you guys go about doing that?
Sud Saxena
>> Yeah, so I think you make a great point, Scott. A pure gen AI-based solutions, sometimes people think, hey, I'll just use the LLM and I'll solve the problem. That doesn't work in 99% of the cases because again, you're looking at complex problems. You're not looking at just simple things like hey, document summarization or a one-step problem. So you're really looking at something that's agentic and something that's able to put into use traditional ML techniques, as well as GenAI. So GenAI is a good complement. So in this case, what really ended up happening is a combination of deterministic approach where you use an embedding model to really compare what's been approved in the past to what's being reviewed now. So it is basically just a very deterministic ML-based approach where you convert the text into a vector and then you compare multiple vectors for a semantic comparison. So, that approach worked well when it came to comparison, but there were elements that GenAI helped with, which is really figuring out in the entire contract, where are the meaningful clauses which really need to be reviewed? So what you call as delineation, and even there, there were approaches that were used to not give the model a ton of text because if you just feed the model a lot of text, it starts hallucinating. So, we used approaches where we actually matched the document against a pre-approved contract clauses and really figured out if there were some that could be eliminated upfront so that the amount of feeding or the context that model has is limited and it allows a model to be a lot more accurate. And there were multiple cycles through which we did this, along with actually AWS help. AWS provided not just product expertise, but also expertise in terms of some of the work on a GenAI standpoint. So, I think the combination really worked out after a few cycles. I'd love for Justin to chime in on some of the architecture approaches and the stack we used.
Justin Guse
>> Yeah, yeah, I can provide a little bit more color there too. And I think one thing, one takeaway here that's really important is that it's not always about just picking any model that will just get the job done, there's a lot more to it than that. So making sure you're picking the right model or the job at hand in the use case. Also training it, right? Sud was talking a little bit about the training aspect to fine tune it along the way to make sure that we were getting the results that we wanted. So the platform itself was built on AWS, and we had all of the requirements that most organizations and financial services would have. It has to be scalable, it has to be secure and has to be performant. Right? So when we designed the architecture and we had all of those things in mind, we wanted to make it modular as well, that we were using things like platform services or services where you're not managing all the infrastructure underneath. AWS is taking care of the undifferentiated heavy lifting. So we've used things like serverless architectures, event-driven architectures, we've leveraged Amazon SageMaker service, Bedrock, for example. Some of the out-of-the-box services that you can use just to streamline the process and be able to develop faster. So we wanted to make sure we had the right balance of flexibility, in addition to the right level of control of the solution and customization of the solution so that we could get the results that we wanted.
Scott Hebner
>> Right, so it sounds like you leverage the generative AI services as sort of a gateway or into the world of AI that gives you that sort of conversational capability. And well, and I think Sud, when you said that generative AI, it's good for repeatable tasks, simple tasks, analyzing information, maybe identifying some patterns and things of that nature but you then complemented that with domain knowledge around contracts. And then I like the semantical part, which I think, correct me if I'm wrong, but you sort of provided context and meaning on how different entities relate to each other, and then you just built that all within a sort of events-based architecture. I mean, is that a summary of how you took the core LLM and generative services and built an ecosystem of capabilities to make it a hell of a lot more powerful than it would've been otherwise?
Justin Guse
>> Yeah.
Sud Saxena
>> Yeah, . We took the whole workflow process and figured out, what should be automated, where we should use traditional AI, where we should have Gen AI, and where we should have some of the other capabilities, in terms of human in the loop, which is very, very important from a compliance standpoint. So I think a combination of all these give you a safer product with the goals of cutting the contract approval cycle and improving the client experience still being met.
Scott Hebner
>> Yeah, we've been doing a ton of research obviously on adoption patterns and the state of where AI is heading, and the one thing that has consistently come back as top priority number one is enhancing trust. And it's not just technical trust, and not just trust in the sense of regulations and privacy and security, but it's trust in the fact that people that are ultimately going to use this need to trust the outcomes, right? They need to trust that the AI knows what it's doing. And you hit on the word safe, safe, trust, those words are probably some of the most important attributes that leaders are thinking as they go forward in time. And you mentioned regulatory compliance and all that, this is all enhancing that, right? The extended architecture that you built, more explainable, all that. Is that sort of a big part of this process that you have to implement?
Justin Guse
>> Yeah, I would say so. To add to that, the call it the security of the platform and sort of the risks with generative AI and the guardrails have come a long way just in the last year, for example. Right? So making sure that we're architecting with those things that you mentioned in mind, and leveraging the guardrails that AWS provides to be able to make sure that we're getting the right output, we're getting secure input, secure output as part of that process. And it's just growing exponentially, right? As more and more industries are using this, it has to be safe, right? We have to be able to put the guardrails or controls around it so that we don't have security risks, for example, as it trains and we load more information into it every day.
Sud Saxena
>> Yeah, and then-
Scott Hebner
>> Yeah, and I like the way... go ahead, Sud.
Sud Saxena
>> The trust aspect also gets better because here we had human in the loop in all scenarios. So as the human in the loop starts noticing that GenAI or the overall AI engine is giving better and better recommendations, the trust increases over time. So that's where you balance automation with trust, so you don't just automate everything. And like this case, first of all, the customer didn't see it as a cost saving initiative. It was really seen as something that cuts cycle time from the contract approval and improves client experience. So that was still being met with this. And then over time, you can really strive more efficiencies, more productivity. So, it's still trust and efficiencies kind of go together. So I think it was very important to have, at least in the first few phases, a very, very important human in the loop element to allow for this, an option to happen.
Scott Hebner
>> Yeah, no, I really like the innovative approach because essentially part of the hallucination problem with the LLMs alone and GenAI alone is that they rely on statistical probabilities and correlations. And by you having human in the loop and you have the semantic knowledge and context and meeting, you're able to start to capture a little bit more of the human know-how, which then applies a better understanding of causality among elements, not just pure correlation, technically. And I think the ability to build around that core infrastructure as you were pointing out, and the added value is in everything that you guys are doing. So again, it goes back to my thought process before about what's learned in financial services being such a complex industry can be applied everywhere and it's good stuff. What are some examples of outcomes that this solution promises to deliver that weren't really possible before? I think I got the productivity part, but.
Sud Saxena
>> Yeah, so I would say, again, as I said before, this was not about cost. This was really about, if my contract approval today takes months, can I cut it into a few weeks? So this is really about the contract approval cycle, knowing that, the faster you close the contract, the faster you make money because this is in a lending scenario. And then of course, a customer experience, because if you have these business to business customers and they're seeing a lot of friction, there is always a competitive threat that somebody else could offer a more tech-enabled and AI-enabled solution and kind of start stealing business away from you, because these are now customer experiences are becoming a differentiator in terms of how you really drive value out of the relationship. So, I think it was those aspects that really are the key outcomes from this initiative.
Scott Hebner
>> Yeah, and I know from my old marketing and sales days that there's a proportional curve, if you will, on when you first start engaging with a potential partner or a client, and the time it takes to close some sort of deal or sale. And the longer things drag out, the less likely it is you actually close. So shortening that cycle has a direct relationship to return on investment. If you're a CFO or a CEO and you're seeing the time to a quality, trusted contract, and therefore a close of the business, that equates to real money, real revenue, real opportunity that could slip away if you let things continue the way they were. So yeah, I got that right, sort of the notion here beyond productivity, right?
Sud Saxena
>> Yeah.
Justin Guse
>> Yeah, trust extends beyond the solution itself too, right? In this case, the other side of the coin is building trust with your clients in the process too, that not only does the technology work, but we're moving as fast as possibly, being sufficient as possible so that we can serve our clients at the end of the day.
Scott Hebner
>> Yeah, again, I mean, that's another observation that I think I'm learning as I talk to more and more people is that generative AI has clear ROI to it, it's just very hard to measure because it's a little bit more about personal productivity and being able to do things quicker, faster, but it's hard to translate into actual dollars, either dollars saved and particularly new revenue. And what you've done by extending beyond that, I mean, you have a very clear use case that you can directly correlate to real revenue, right? Closing of new deals because you're speeding the cycle and all that. And I would think that that would spark leaders, whether it be CEOs or CFOs or whatever to say, "Hey, it's working here, let's invest even more in AI across different use cases," because there's probably an uncountable number across a business like this, right?
Sud Saxena
>> Yeah. It's really top-down, that's how we see it. There has to be top-down vision on how or what outcomes you will achieve. And going in with clarity around there, the teams will find a way. So, I think that's the real learning here, the top-down view and the top-down belief on driving certain outcomes.
Scott Hebner
>> Right. Well, I want to comment about vision. So, what is your vision or your view of the evolving AI and cloud landscape going forward here? I mean, what emerging technologies are exciting to you that maybe a year from now if we're talking that you might be putting into place?
Sud Saxena
>> So it's hard to pick a horse, but I can talk about the domains where we see a strong impact due to AI and GenAI. And the impact could vary from productivity, to customer experience, to the example I gave in really cutting cycle time or friction from certain transactions. So the three domains we are seeing really rise to the mainstream are knowledge management. So this use case would kind of be in the knowledge management domain. Customer care, and SDLC. So if you look at, for example, financial services customers, a lot of processes that are very document-centric and have a lot of friction. Take mortgage processing for example. I think there's an opportunity to reimagine all of those with AI now, in getting these types of tooling that drive friction out of the process. SDLC, there are a bunch of coding tools out there, startups that are playing this space, the large hyperscalers, including AWS, have their own solutions. Huge, huge productivity improvement potential, but to the point you made earlier, is right now not a single client can say, "Hey, I leveraged this AI tool in SDLC and I'm driving 20% savings." Not a single client can say that because it's still at a task level. The question is when you start really driving a top-down vision and then building workflows that are somewhat complex in nature, like the example I gave, that's when you really start driving productivity. So the message we have for our clients is, yes, provide AI tooling to your people. I think they need to get comfortable, they need to start using it. And the statistics go from, I think it was about 40% to almost 50, 60% usage now. So a majority of the enterprise that we are giving, going back to the SDLC or coding example, is using it, which is good. Then it's really about, how do I take some of these big transformational programs and how do I make them more efficient? So I think the message really here is that other than providing AI tooling for your people, look at these three domains which are pretty , and pick specific initiatives that can be used to show value. And then once you start seeing success in a few of those, then you will see the cream rising to the top or the effect where other people would like to do the same to really drive productivity and value in their initiatives. So, that's really the lesson for large enterprise customers.
Scott Hebner
>> Yeah, one thing that's consistent across those three domains is the notion of being workflow driven, which usually involves more than one individual being involved in solving a problem or completing something. Which then gets us to the question on your view of agentic AI and AI agents, and how do you see that starting to get infused as we go through time here, right? Because I think the difference between AI assistant based on generative AI, again, doing repetitive tasks, analyzing information, things of that nature, creating content. When you get into an agent, you start focusing on its ability to help you solve a goal or make a decision, and then agentic is when you wire these things together to partner in a workflow along with human in the loop. Right? So, you see that becoming a big deal, particularly in financial services?
Sud Saxena
>> Oh, 100%, right. So that's why I think you might have heard of the PwC Agent OS platform, it's basically an orchestration mechanism for multi-agent workflows. And the reason we did it is because as we were leveraging GenAI internally, we found there was an opportunity to really drive these multi-agent workflows, and you need an orchestration mechanism to do that. So yes, what we really see happening is there is an agent feeding to another agent, to feeding to a human, and then coming back to an agent. So, I think that's where you really start building these agentic workflows. And in the domains that I talked about, that's the opportunity we see. So I think, and ultimately, what is an agent? It's just AI that takes action, or in some cases it could be an AI that provides input. So I think the opportunity here, as I said, really is to take a complex workflow, decompose it and figure out what tasks can be done by AI, what tasks can be done or should still be done by humans. And how do you really make the whole process efficient by connecting it in a fashion where it's safe and it also provides trust and value, and you have an opportunity to improve it over time. Yeah, so I think that's really the potential for FS or large enterprise plans.
Justin Guse
>> And I think just to add to that and kind of stitched it all together in the solution that we built. While we were not using agents specifically for the solution, I see many opportunities to improve it going forward using agents. Right? So as we had talked about for knowledge management for example, or even training, right? Perfect use case to stitch the pieces together to train the models, learning on the fly, for example. Streamlining the process of how we were using it and uploading documents and there's a step to each part of the process driven by the business rules, but you stitch that together with agents and it makes the experience a lot more streamlined for the clients. So a perfect example, I think of a use case where just our own solution that we built for the client, agents have in place.
Scott Hebner
>> Yeah, and hopefully we'll look back someday and say look, we spent three years with generative AI. I think it gave people, everyday human beings, both in the business world and in society in general, it gave them a flavor what AI can do and taught them how to use it as sort of the gateway. Right? And then companies like you, in partnership with AWS, help real businesses build extended architectures around that generative AI capability to solve real problems that have real ROI that can be measured. And then we move into the world of agentic AI where you get into workflows and organizational level kind of stuff, and like you said, acting more autonomously and taking action. And I think we're just getting started with all this, which sort of brings me to our final question here before we have to wrap here. Your experience, everything that you've gone through with this project, what have you learned that can help other financial firms think about what to do now, what to do going into the future? What's the learnings for others that are maybe watching?
Justin Guse
>> My mantra in life is, start small, start simple, with a particular use case in mind. Right? We solve problems with technology, it's usually not the other way around. So starting small, finding a particular use case, and then building around that. Right? And it doesn't mean that you start with a big bang, like, oh, we have to build all the pieces at once. Right? Even the approach that we took was a very methodical approach to building the solution. Right? You've got the front end, and you've got the middle where, you've got the back end, and then you stitch that together and you prove your use case incrementally over time. So, I think you have to start simple, you have to have a use case. Otherwise, you're just going to kind of be going down a path or down a road that might not lead you anywhere at the end of the day.
Sud Saxena
>> Yeah, yeah. So the other thing I would add is, start, don't sit on the sidelines, and I don't think any client is. But the more complex problems you try to tackle, the more you learn. So, don't just start by taking the simplest problems, take some of the more complex ones, use your innovation studio, use your innovation labs, or use AWS funding mechanisms to experiment a little bit with some complex problems where you can actually show value to the organization.
Scott Hebner
>> Yeah, I think that's a really important message. My old mentor, Irving Wladawsky-Berger when I worked at IBM, always used to say, "The game is played on the field, not in the dugout." You got to get out there and start doing things and learning, and you will find your way as you go through time. The worst thing you can do, particularly today with these innovation cycles moving at warp speed is, it's going to be too easy to fall behind and you may never catch up. So you got to get out and start doing things, don't wait, which I think is a really good message. Along with this idea that you got to pay attention to what's happening in financial services, right? It's definitely a leading indicator of what's going to be possible elsewhere. So Sud, Justin, thanks so much for coming on theCUBE. It's been a fascinating conversation and certainly has given us a lot to think about here and been an absolute pleasure talking with both of you.
Sud Saxena
>> Thank you, Scott, pleasure talking to you.
Justin Guse
>> Thanks, Scott. Appreciate it.
Scott Hebner
>> And for all of you, thank you for tuning in. You're watching TheCUBE's coverage of the AWS Financial Services Symposium 2025. Visit thecube.net, and Siliconangle.com to watch all the interviews that are part of the broadcast. We'll be right back right after this short break. We are the leader in enterprise tech news and analysis. We'll see you soon again.
Scott Hebner
