# Inside Story: Ally & Microsoft Groundbreaking AI Collaboration

- **Show:** Banking Transformed with Jim Marous
- **Published:** 19 Dec 2023
- **Source:** https://www.podtrac.com/pts/redirect.mp3/chrt.fm/track/A27C8C/traffic.megaphone.fm/FPMN1544244609.mp3?updated=1702927890
- **Transcript:** https://podcasts.opennash.com/shows/banking-transformed/inside-story-ally-microsoft-groundbreaking-ai-collaboration-8c0966/
- **Provider:** OpenNash Podcasts (machine-generated)

## In brief

Ally, a digital-only bank, built its own generative AI platform, Ally.ai, on a private copy of OpenAI's models running in Microsoft's Azure cloud. The argument is that banks can adopt generative AI safely and productively if they control the platform, protect customer data, start with internal uses, and treat AI as an assistant that helps staff rather than replaces them.

The central tension is speed versus caution. Ally moved quickly, with legal contracts turned around daily, while stripping out personal data and keeping nothing in the model's memory. Both guests say culture, leadership willingness to take managed risk, and employee education decide success more than the technology does, and they stress keeping humans involved in the work.

## Key insights

1. **Ally built its own platform to control risk** Sathish Muthukrishnan says Ally needed a safer platform it fully controlled while the technology was changing fast. It runs a private instance of OpenAI models on Azure, so no one else can access it.
2. **Personal data is stripped before leaving Ally** Ally.ai removes personally identifiable information before a request leaves Ally's firewalls, then restores it in the response. Ally's team contributed this module to the open-source LangChain project.
3. **Frontline staff drove the biggest early success** Customer care associates received automated call summaries that Ally says are over 85% accurate, freeing them to focus on customers. They then proposed most of the 205-plus use cases now in progress.
4. **Culture and leadership mattered more than technology** Muthukrishnan credits Ally's 'do it right' culture and sustained board-backed investment in data and infrastructure. Priya Gore adds that educating employees on what AI can and cannot do avoids fear of job loss.
5. **Start small and keep humans in charge** Gore advises 'crawl, walk, run,' beginning with low-risk uses, and notes Microsoft chose the name 'Copilot,' not autopilot, deliberately. Banking expertise stays central, with a human in the loop.

## Questions this episode answers

### What is Ally.ai and why did Ally build it?

Ally.ai is Ally's in-house generative AI platform, built on a private instance of OpenAI's models in Microsoft Azure. Sathish Muthukrishnan says Ally wanted a safer platform it could completely control while the technology evolved quickly.

### Is a bank's customer data used to train the AI models?

No, according to Priya Gore: in Azure the bank reasons over its own data in a private instance, and nothing is pushed back to retrain the model. Muthukrishnan adds that the data is forgotten after each session.

### How does Ally protect personal information when using generative AI?

Ally.ai removes personally identifiable information before a request goes outside Ally's firewalls and puts it back when the answer returns. Muthukrishnan says this uses a technique built with LangChain, an open-source tool.

### Will generative AI take bank employees' jobs?

Both guests say AI should assist rather than replace people. Muthukrishnan says it may replace some jobs but not humans if they are helped to adapt, and Gore stresses keeping a human in the loop.

### Where should a bank begin with generative AI?

Gore recommends thinking big but starting small with a low-risk project to learn how the technology works. She suggests uses like document analysis, summarization and customer-engagement content, and involving partners for governance.

## Facts

- **Sathish Muthukrishnan** [0:02:04]: "I grew up in India and started my first job with Singapore Airlines in Singapore and got to a point where I got offered a job to move to U.S. with United Airlines."
- **Sathish Muthukrishnan** [0:02:36]: "Then had the opportunity to move into a completely different vertical industrial working for Honeywell as their first chief digital officer."
- **Sathish Muthukrishnan** [0:03:11]: "a fully native, digital native organization, the largest direct-to-consumer digital bank in the U.S., and also the largest loan originator for autos here in the U.S."
- **Sathish Muthukrishnan** [0:04:49]: "We were the first bank to introduce rewriting or deferring the loans 100 % digitally."
- **Sathish Muthukrishnan** [0:20:28]: "We have our own private instance of OpenAI on Azure, which itself is terrific because we know that nobody else has access to that instance."
- **Priya Gore** [0:19:12]: "The Azure OpenAI service is basically a secure service leveraging all of the great capability that OpenAI has developed and brought forward to the market"
- **Priya Gore** [0:19:46]: "The answer is no, because when you use the service in something like Azure, you're able to reason over your own data in a private instance."
- **Sathish Muthukrishnan** [0:23:49]: "Ally.ai allows us to strip out anything personally identifiable to our consumers or attributable to our consumers before we even send that request outside the Ally firewalls."
- **Sathish Muthukrishnan** [0:24:31]: "we have built a module that will strip out the PII and rehydrate the PII using LangChain that my team has contributed to the source code of LangChain"
- **Sathish Muthukrishnan** [0:31:52]: "It's more than 85 % accurate. It's saving me precious minutes or seconds."
- **Sathish Muthukrishnan** [0:31:52]: "We have over 205 use cases, and this is from this morning, heart of the press, use cases that my team is working on across different businesses"
- **Sathish Muthukrishnan** [0:29:25]: "I believe we were one of the first, if not the first corporate customer to use the private instance of Microsoft OpenAI."
- **Priya Gore** [0:39:15]: "The first co-pilot solution that we launched a little over a year ago was the GitHub co-pilot."
- **Priya Gore** [0:40:23]: "it took 16 years for the mobile phone to reach 100 million users. Seven years for the Internet. Facebook, 4.5 years. ChatGPT, three months."
- **Priya Gore** [0:15:36]: "generative AI is really great at four things, right? So generative AI won't be the answer to all things, right?"
- **Priya Gore** [0:12:46]: "So I joined Microsoft about six years ago."
- **Priya Gore** [0:13:16]: "we also launched an AI framework for responsible AI."
- **Sathish Muthukrishnan** [0:24:31]: "Generative AI is not able to do that effectively. And the largest open source that allows you to do it is called LangChain, which almost all companies use it."
- **Sathish Muthukrishnan** [0:50:07]: "we recently launched what we call One Ally, where we are bringing all of the experiences across all of the products into a single digital experience."
- **Priya Gore** [0:52:31]: "we're starting to roll that out with our clients that are in the early access program"
- **Priya Gore** [0:46:53]: "We're refreshing that right now. And next quarter, that will be readily available for anyone to leverage."
- **Sathish Muthukrishnan** [0:46:06]: "And we call this AI equity. How can we stick to our mantra, which is do it right?"

## Transcript

**[0:00:11] Jim Marous:** hello and welcome to Banking Transformed the top podcast in retail banking i'm your host Jim Marous owner and ceo of the digital banking report and co-publisher the financial brand the launch of ally self-built ai platform Ally.ai in collaboration with microsoft represents a measured yet trailblazing foray into next generation technologies for banking. Rather than reacting to hype, Ally is taking a principled approach to focus on productivity, security, scalability, and an improved customer experience. Ally's emphasis on co-piloting change also signals the value of strategic partnerships between banks and tech players in steering AI's responsible adaption, providing a blueprint for financial institutions eyeing productive AI integration.

**[0:01:07] Jim Marous:** Joining me to discuss this exciting collaboration is Priya Gore, General Manager of Data and AI for U.S. Financial Services at Microsoft, and Sathish Muthukrishnan, Chief Information, Data, and Digital Officer at Ally. New cloud-based generative AI models for financial services must bring next-generation capabilities while emphasizing security, governance, and business innovation. Satish and Priya will share insider perspectives on Ally's deliberative approach to explore at AI, the business impact seen so far, and why financial services need a tailored framework to harness AI's responsibility.

**[0:01:54] Jim Marous:** So Satish, let's start with you. Can you give a little background as to how you got to Ally and your journey to where you are today?

**[0:02:04] Sathish Muthukrishnan:** Jim, thank you for having me. Looking forward to the discussion with Priya. Excited for it. I grew up in India and started my first job with Singapore Airlines in Singapore and got to a point where I got offered a job to move to U.S. with United Airlines. I got here right before 9-11. Worked for United for a little bit of time, going through those turbulent times, but learning a lot, and then got the opportunity to move to American Express in Phoenix, Arizona.

**[0:02:36] Sathish Muthukrishnan:** Little did I know I was joining Amex right before 2008, the worst financial time for our country. Again, a great opportunity to learn, but leverage on the learnings from the calamity with 9-11 at United. I worked at American Express, had the opportunity to lead digital transformation, helping move Amex from a financial services company to a fintech company, then had the opportunity to move into a completely different vertical industrial working for Honeywell as their first chief digital officer.

**[0:03:11] Sathish Muthukrishnan:** That was local to Phoenix, but was reporting to the CEO and being part of problem definition was very exciting. While I was with Honeywell, came the opportunity to explore Ally, a fully native, digital native organization, the largest direct-to-consumer digital bank in the U.S., and also the largest loan originator for autos here in the U.S. Took that opportunity to be part of a digital native organization and have been very happy for the last four years.

**[0:03:48] Jim Marous:** You know, it's interesting, Satish, as you describe your journey. In each situation, you've come onto the scene of your organization at a time when you could rely on your learning from the past, but couldn't rest on them because things were changing so fast during 9-11 with an airline and during American Express during a major financial crisis. So as you got to Ally, obviously, there's a lot of things going on at once, certainly in the digital space.

**[0:04:16] Jim Marous:** But what motivated Ally to build an in-house AI platform? And what capabilities does Ally.ai enable? And what came about in the decision to partner with Microsoft?

**[0:04:30] Sathish Muthukrishnan:** Yeah, that's a great question. Like you said, I couldn't rest on my learnings. I joined Ally and was a bit bored in the first month, so decided to introduce COVID. Just joking there.

**[0:04:44] Jim Marous:** You simply are a precursor to bad times. I'm just saying.

**[0:04:49] Sathish Muthukrishnan:** Yeah. I am not moving. I'm done. So, Jim, the answer to your question lies in that situation. many companies like Ally discovered what you can and what your real potential is during COVID. Everybody was facing this massive uncertainty. We were the first bank to introduce rewriting or deferring the loans 100 % digitally. Other institutions were trying to do the right things, But if you have to do it end-to-end, you may rewrite a loan or a contract, but you had to call to a call center to have it finalized.

**[0:05:34] Sathish Muthukrishnan:** And we all know what happened during the call centers during COVID. We were able to quickly build that software and do it end-to-end and showed us the potential, but more importantly, showed us what customer appreciated of us. And we always go by do it right and to a fault. We want to do what customers want first before we make sure that it also resonates with our internal customers. And that muscle was exercised when last November, OpenAI through ChatGPT showed up.

**[0:06:11] Sathish Muthukrishnan:** And we figured this was a technology that's going to be here to stay, but also will allow us the opportunity to explore new areas. And that thinking is what led us to creating what you rightly termed as the Ally.AI platform. As the technology was evolving fast and we were all learning, we needed to have a safer platform that we could completely control in leveraging this new technology. And that's how Ally.AI was born.

**[0:06:41] Jim Marous:** So Satish, sticking with you for a little bit, what's interesting about generative AI is it's a solution sometimes looking for a problem. And so not only do finance institutions have to look and say, do we want to get involved with generative AI, but we got to figure out where do we want this to be applied? And as I'm not mistaken, Ally settled on the initial focus on improving internal processes such as employee productivity and even marketing.

**[0:07:10] Jim Marous:** How did you come to where you wanted to apply the generative AI, Ally.ai model at your company?

**[0:07:18] Sathish Muthukrishnan:** What a great question, Jim. You do not often have technology that can have an equal impact across businesses, across industries, really across the entire human race. I feel like generative technology can have, generative AI can have that level of impact. If you think about the last 40 or 50 years, the people that were uber successful were the knowledge workers. And for knowledge workers to be successful, you had to specialize in software development. You have to specialize in building a business and launching it and making it successful.

**[0:08:07] Sathish Muthukrishnan:** What generative AI has the potential to do is break that down. and break it down. English is something that everybody understands and you could use plain English to leverage this technology to do what a knowledge worker did over the last 50 years and have the same level or even a bigger impact. And you asked us about our partnership with Microsoft. We wanted to, understood and wanted to take advantage of this technology and had a working session with Microsoft and their campus in Seattle earlier this year, beginning of this year, we were pleased with the lean-in that we got from Microsoft and the excitement and enthusiasm in co-innovating.

**[0:08:57] Sathish Muthukrishnan:** And we found that as we were learning this technology, we have this hyperscaler who is ready to invest and take the journey along with us. And that's where the relationship started. And we were pleased we started there because of the results that we are already seeing.

**[0:09:18] Jim Marous:** So playing off that answer, you know, Ally, as you mentioned, is a completely digital organization. Microsoft, obviously, is a technology forward organization. You're comparing yourself to other finance institutions. What maybe singular or maybe two advantages do you have as an ally being digital first when you're looking at partnering with organizations that are tech first? Because I think a lot of financial institutions get stuck on, okay, how do we hire the knowledge people? How do we build our team to be able to use the tool that Genre of AI has?

**[0:09:55] Jim Marous:** Do you think there's an advantage to partnering, to Ally being the partner with Microsoft from the perspective of you both are coming from a

**[0:10:04] Sathish Muthukrishnan:** starting point of tech-first companies? My answer might surprise you a little bit. It starts with the culture that we have developed within Ally, and it rests on a very simple phrase, do it right. And you do it right by your customers, you do it right by your employees, you do it right by your stakeholders and our partners. And what that does is, allows me and my team to have a very strong and open relationship with all the stakeholders and the executive committee.

**[0:10:42] Sathish Muthukrishnan:** We are collectively able to say, yes, this is a little scary to take risk and use generative AI this early on in its cycle, but it's going to create a big impact. Let's take this risk together and let's explore what we can achieve, that inherently, that culture that is built in, I cannot put a value on it, but allows us to go do these experiments and allows us to do what is right for our customers. So that's a big advantage.

**[0:11:13] Sathish Muthukrishnan:** The second thing, obviously, you stated is a digital native company. And we all understand the importance of investing and persisting that investment in tech, which historically has allowed us to invest in our data ecosystem, modernizing our network and infrastructure, which are building blocks and allows us to use this technology much easier than any other company. And I'm so thankful for the executive committee, for the board that believes that you may not get the return on investment today, but you have to continue investing and persist that investment over a period of time, which now we are seeing the benefits and reaping

**[0:11:55] Jim Marous:** the benefits from. What a great answer. In my podcasts and webinars and writings I do, I often lean back to say, at the end of the day, leadership's going to define whether or not it's going to be a success. Investing in technology without leadership understanding the application of it and embracing it, it just will work. But you hit something that I don't hit often enough, which is really the difference between avoiding risk, which traditional financial institutions have had in their history, and managing risk, which is really a different view because it says we understand that everything's not going to be perfect, but we're going to embrace that and work from there.

**[0:12:33] Jim Marous:** So, Priya, you know, you're at Microsoft. You haven't been there a very long time in the scope of things. Can you talk a little bit about your journey and what brought you to Microsoft and what has happened since you've joined?

**[0:12:46] Priya Gore:** I thank you so much also for having me. I would love to share. So I joined Microsoft about six years ago. And prior to that, I was working. Actually, I grew up alongside Microsoft, to be honest, in their partner community, working at a few different entities, most recently prior to joining with a consulting firm that was specialized in building cloud native applications with our customers in the Azure platform. So we had a lot of work we were doing when the Azure IoT service came to life in Azure.

**[0:13:16] Priya Gore:** As the Azure AI services became available when we launched those, that was when I was prompted, if you will, to come join the company. And so my first job here at Microsoft was part of our global black belt organization as the Azure AI leader in the America's time zone. And really, my job was simply to get out there and talk to customers about Microsoft's point of view on artificial intelligence. That was six years ago. Six years ago, seven years ago, you know, when we launched our machine learning cognitive API services, we also launched an AI framework for responsible AI.

**[0:13:56] Priya Gore:** And that was another thing that I was helping bring forward. If you imagine, like, most people didn't know Microsoft as an AI company. They thought of Microsoft for the great hero products that we've had, right? Like Windows and Microsoft Office 365 and all these great solutions that we have. So fast forward to today and where we are now, it's just amazing and unbelievable. And the culture, to Satish's point, too, that we foster here at Microsoft is real.

**[0:14:22] Priya Gore:** We live and breathe our mission every day to empower every person and every organization on the planet to achieve more. And certainly we're able to accelerate that mission with this innovation that we've seen in the generative AI space and being able to not only infuse that technology in our own solutions, but bring the building blocks to our customers in Azure so they can build their own co-pilots and innovate very, you know, kind of holistically and very prescriptively for their own customers, which we're really excited about.

**[0:14:56] Jim Marous:** So, Priya, Microsoft works with a lot of different organizations across the marketplace. And it goes without saying that financial services is unique in many ways. In many ways, they have to work, but in many ways that you have to partner with a financial institution. Where do you see the most opportunity within the banking sector when it comes to generative AI? and, you know, using the crystal ball a little bit, you know, what do you see as being the upside based on where we are today to where we could be in a very short timeframe in the banking world with those firms that are digitally mature and willing to embrace generative AI?

**[0:15:36] Priya Gore:** Yeah, I'll start with this, you know, just as a refresher, you know, generative AI is really great at four things, right? So generative AI won't be the answer to all things, right? And we are seeing a resurgence in classic and traditional AI as well as part of this wave and era of AI. The four things that generative AI are the best for are content generation, knowledge summarization, code generation, and semantic search. So when you translate those patterns, if you will, to, for example, the banking industry, I would say some of the things we're seeing most low-hanging fruit for banks to think about first, and I know Ally was in this boat as well with what they were doing is

**[0:16:17] Priya Gore:** thinking about client engagement, thinking about market research and content generation and how can we actually leverage these capabilities using natural language, prompted with natural language, frankly, any language, whether it's English or any language, right, to be able to reason over structured and unstructured data like never before has been incredible, right? Going down the line, we're also seeing opportunities for banks with presentations, pitch book generation, quote generation, things like that, being able to leverage this technology for that.

**[0:16:51] Priya Gore:** Market surveillance is another use case. Document analysis using natural language, of course, that's very, very popular. And that frankly is cross industry, but very popular also in the banking world. And then being able to really engage other models. So I mentioned our pre-trained cognitive service models. These are speech, language translation solutions, document, you know, reasoning solutions, semantic search services. These are all things that when you bring those services into an architecture with the generative AI APIs and build these intelligent applications at scale, the power of what you're bringing forward and the, I'll say, amplifying human ingenuity just is on fire right now in a great way.

**[0:17:36] Priya Gore:** And of course, we want to make sure that we're also helping our clients leverage all of this great capability with a sound, reasonable, responsible AI framework. And that's another

**[0:17:47] Jim Marous:** way that we're helping banks as well. Yeah, I love everything you said that there are a lot of good nuggets there. And I'm thinking, you know, not every financial institution is the same. Obviously, Ally has a lot of advantages in that they're really very digitally mature. They embrace what AI can do. They did that before Generative AI came about. But when you look at how Microsoft Azure enables ally to scale AI, how does it do it securely while managing risk?

**[0:18:17] Jim Marous:** Because if there's one thing that can be said very clearly within the financial services industry is the hesitation is always going to be around what if something goes wrong. So when you're working with all finance institutions and everybody's throwing spitballs to the wall and saying, what's going to stick? What are some cautions or watchouts that your team has seen when you look at AI,

**[0:18:42] Priya Gore:** especially in banking? Yeah, I would say this. Certainly for Microsoft, regardless of the Azure AI platform or our platform in totality end-to-end, security is the number one most important thing. and that security posture for our clients. That's the business that we're in, right? We're in the business of providing secure cloud ecosystems for our clients to innovate. So that's at the forefront. The interesting thing with something like generative AI, and I'll use the Azure OpenAI service as an example.

**[0:19:12] Priya Gore:** The Azure OpenAI service is basically a secure service leveraging all of the great capability that OpenAI has developed and brought forward to the market, but bringing those exact same models into the Azure ecosystem so that our enterprise customers, all of our customers can leverage that capability in their own secure tenant, in their own ecosystem of, you know, data repository, their ability to reason over structured and unstructured data sets, leveraging these great capabilities and these models, but doing so safely and securely.

**[0:19:46] Priya Gore:** So, for example, a lot of times we get questions like, hey, when we leverage your service, are you going to be using our data to retrain the model? The answer is no, because when you use the service in something like Azure, you're able to reason over your own data in a private instance. You're able to think of it as a one way street. So we're bringing you the capability of the model, but we're not pushing back any training of that model back out into the wild.

**[0:20:11] Priya Gore:** And so that's a simple example of how we're bringing, you know, secure, very, very, very mature, secure infrastructure, if you will, with Azure forward and leveraging, giving our clients the ability to leverage this great technology in a safe way.

**[0:20:28] Sathish Muthukrishnan:** I could add to that, Jim, very specifically what Priya stated. We have our own private instance of OpenAI on Azure, which itself is terrific because we know that nobody else has access to that instance. And as she also stated, while we use the technology, the data that has been used to trigger the technology is forgotten after the session. So it's not learning from Ally data. It is not retaining the memory of the interaction. And that's why Ally.ai is so critical.

**[0:21:09] Sathish Muthukrishnan:** So what we lose in training the model, we are able to use the data in refining our input and triggering the model and also argumenting the output from the model. And Ally.ai has become so powerful and critical in how we use generative AI for our initial set of use cases.

**[0:21:30] Jim Marous:** So, Zinia Saseesh, you actually took my next question away from me. So that was pretty good because I was going to ask you about, you know, when you're looking at intimate consumer data, how do you balance the capabilities with the security? Because sometimes, you know, we look at this and say, well, there's a lot of things we can do. And then it gets down to, should we? And then it gets back to, can we still deploy something?

**[0:21:54] Jim Marous:** So we had an interview with a gentleman named Brian Romley on around four months ago. And he talked about a potential future where we could build these individual generative AI models for each individual, where we can actually have generative AI learn on the go and build conversations and engagement with a consumer based on their data, their tech, There are conversations, there are transactions, all these things. Do you see a time when we really get that intimate with the ability to deploy an AI model to understand a consumer on an individual basis where this almost becomes your personal financial consultant?

**[0:22:37] Sathish Muthukrishnan:** Jim, this is my opinion, and I'll tell you where the consumer behavior is going. We went from a very generalized experience provided to consumers to a more personalized one. My prediction is it'll go to a personal experience from a personalized experience. That's one. The second thing that you want to think about is customer data and how customers guarded their data and what they thought was personal has changed over the last few years. Your personal phone number or email or your home address was very sacred.

**[0:23:16] Sathish Muthukrishnan:** You were probably not sharing it with anybody else other than your friends, even 10 years ago. But look at it now. Everybody has your phone number. It's easy to find if they don't have it or your home address or where you live. And you're willingly giving it to somebody that you've never met through Uber Eats or DoorDash. So how customers think about their data has changed. However, we need to protect the sacred data that allows us to interact from a business standpoint with our customers.

**[0:23:49] Sathish Muthukrishnan:** And I go back to Ally.ai. Ally.ai allows us to strip out anything personally identifiable to our consumers or attributable to our consumers before we even send that request outside the Ally firewalls. And when the output comes in, we're able to regenerate it. And I apologize for getting slightly technical here. People that are using generative AI models are using a technique called RAG. And what RAG allows you to do is maintain context between the conversation and allows you to build up and connect the dots with the conversation.

**[0:24:31] Sathish Muthukrishnan:** Generative AI is not able to do that effectively. And the largest open source that allows you to do it is called LangChain, which almost all companies use it. Not only do we take our security and safety of our customer data safely, we have built a module that will strip out the PII and rehydrate the PII using LangChain that my team has contributed to the source code of LangChain for the entire industry to use. So not only are we moving fast, not only are we innovating, we are doing it safely and securely for Ally, but also being good citizens by contributing the code to Langchain for the rest of the industry to use.

**[0:25:17] Priya Gore:** No, like this is an example of like what we love in Ally is a partner and a customer. I mean, they're clearly, as Satish stated, not just, you know, innovating and developing for their own use and for their own IP development, but then, you know, taking those learnings and those best practices and those very important code bases and democratizing it back out into the market and contributing that to to the open source community so that others can benefit from that.

**[0:25:51] Priya Gore:** And I think that that's a trend that we need more developers and product builders that are using this new technology to foster. I think that that's how we all become better. Even some of the things that we're doing out of Microsoft Research, we do the same thing. We release things very early out to the market. We want people to use it safely, securely, with the right sort of parameters around it, but to help us innovate and help us make it better.

**[0:26:16] Priya Gore:** And so I love that you're doing that. And that's a great example of the partnership that we seek to have with all of our clients that are innovating with this technology at scale.

**[0:26:27] Jim Marous:** You know, it's interesting, both Priya and Satish, but Satish, in your perspective, you said the whole concept of saying, you know, the concept of what's private has changed from the consumer's perspective. I mean, I go back to a day in financial services when we were required. I'm sorry, we were not required, but we were trying to get consumers birth dates in the banking file. And not only did consumers not want to give it, but employees did not want to ask for it.

**[0:26:56] Jim Marous:** until the government required that as part of the whole PII and security issues they were looking at. But it's interesting because it really gets down to, from my perspective, value transfer. If you're going to ask for information or use information, you better be able to provide value. And I'll be okay with it then. I mean, I use this example, Amazon. My example is Amazon, where we pay $130, $135 a year for Amazon to give us digital products that may be even at a higher price, but do it quickly.

**[0:27:33] Jim Marous:** And we're okay with that because they've managed our data really well and given us value back. And when I ask a whole room of people, how many of you have Amazon Prime? Everybody raised their hand. I ask how many people have considered getting out of the whole Amazon Prime model. There's a few. And I ask how many actually did it. There's none. And that's because we believe that there's a value transfer. And from Ally's perspective, I know from a personal perspective in being an Ally customer on the loan side, you use what you need to use to provide opportunities to me that are along my life scale.

**[0:28:11] Jim Marous:** And in doing so, I'm okay with you using that. And that is so important because those organizations that actually let the consumer know how they're using data, but even more importantly, what benefits the consumer is getting from that are going to be far ahead of the game because consumers are not going to think they're just asking for data or using data and putting it to the side, George, use an old analogy. So, Satish, when you look at your early results, and it is certainly early in the game and it is changing every day, what results have stood out from an early customer care perspective and also within the Ally organization as far as how you've deployed Generative AI.

**[0:28:53] Jim Marous:** What are some success stories?

**[0:28:56] Sathish Muthukrishnan:** Jim, first of all, thank you for being our auto loan customer. I cannot wait for you to become our banking customer and blow your mind with the experiences that we're going to provide you.

**[0:29:07] Jim Marous:** Always love it. We sell on the podcast. I like that. That's good.

**[0:29:12] Sathish Muthukrishnan:** From a critical view, I would love for you to use our products and tell us where we could be even better.

**[0:29:17] Jim Marous:** Yeah. Yeah, you're probably going to get that out of this conversation. So from your perspective, what are some success stories you've had from generative AI, early success stories?

**[0:29:25] Sathish Muthukrishnan:** Yeah, I will tell you, not because I'm on the podcast with Priya from Microsoft. There is a mindset shift in using generative AI. In the initial days, and I believe we were one of the first, if not the first corporate customer to use the private instance of Microsoft OpenAI. our legal teams, we're both corporations, our legal teams were going back and forth with the contract on how and when we'll use it, how secure it is, et cetera.

**[0:29:56] Sathish Muthukrishnan:** It was completely unclear. We were working through the processes. I can tell you the lawyers were turning around the papers on a daily basis. That's unheard of. This is not a knock on lawyers. By nature, they have to be very detailed and will take a few days for them to give us an answer. But between corporations, it was happening on a daily basis. That is the first success of generative AI. But I'll bring it back to - Well, and Satish, on that Satish,

**[0:30:29] Jim Marous:** what's interesting, that was being done at a time when it was like mercury on linoleum from the standpoint of what generative AI was. I mean, you talked about early in 2023, and it only got introduced in November of 2020, 2022. So the reality was lawyers are going back and forth with paperwork while what they're trying to do overall was changing in real time. So it's not like they even knew what the landing point, in fact, we still don't know what the landing point is going to be.

**[0:30:59] Jim Marous:** But to have that kind of iterative nature of that is really unique in that nothing was going to settle. Both organizations understood that change was going to happen. and it's going to happen really quickly, and it hasn't stopped.

**[0:31:14] Sathish Muthukrishnan:** You nailed it. You nailed it. Let me add from our first use case, and I feel like this is the biggest success story for us and a big impact, is we started to believe that the ultimate leverage for generative AI will come from every employee at Ally, not just technology employees. So when our customer care associates started experiencing the automated summary of the call they just had with the customers, they came out and said, wow, it's capturing things that I never would have captured before.

**[0:31:52] Sathish Muthukrishnan:** It's more than 85 % accurate. It's saving me precious minutes or seconds. And I am allowed to singularly focus on customers. That's making my impact on the customer much more. So those are the early successes. But let me tell you what has happened since. We have over 205 use cases, and this is from this morning, heart of the press, use cases that my team is working on across different businesses and functional areas in the company. Majority of the use cases are coming from our customer care associates.

**[0:32:28] Sathish Muthukrishnan:** They now understand the technology. They understand what it can do and are proactively influencing and forcing technology to deliver on it. We are thrilled. This is just the early days. But what we can do for Ally and our own employees initially is very, very promising.

**[0:32:49] Jim Marous:** You know, that's interesting. It's another one of those differences that you probably don't see because you're not at a legacy financial institution. But when all employees are pulling the same direction towards a common goal and not being afraid of change, I mean, working for Ally and working for Microsoft for that matter, if you're afraid of change, you're in the wrong business because it's always changing within the organization. Legacy finance institutions are usually saddled with legacy leadership that honestly has really on the surface never done anything wrong.

**[0:33:24] Jim Marous:** They continue to make profits. they continue to do well, but that also becomes their hindrance because we as human beings most don't like change. Now, we may embrace it at times and like it sometimes when it works in our favor, but because you're both working for organizations, that is part of the definition of the organization. It makes it so Satish, you brought up such a great example where employees are getting together and say, let's find other ways.

**[0:33:53] Jim Marous:** We can, in some ways, displace ourselves for the better good without being fearful of losing their job. It simply moves them forward. I think either you or Priya said earlier, the humanization of the digital transformation where early applications of AI had to do with risk models and fraud and things like that. That was a very clear indication that we were going to be assisted by AI, not displaced. But some models right now, some of the things that could happen.

**[0:34:25] Jim Marous:** Yeah, that's very true, Jim. Yeah. So, you know, some of the ways it's being used right now actually could displace jobs, but doesn't have to displace humans if they continually evolve with the marketplace. So, Priya, I think one of the things we see when we interview organizations is that we, at some point, will clam up and not want to use generative AI at all. Or in other cases, we may think generative AI can be the savior of everything.

**[0:34:55] Jim Marous:** The answer is obviously somewhere in between. But when you're looking at any type of organization, not just a financial institution, where have you seen instances where organizations have had missteps with generative AI, either in the definition, deployment, or actually in the application of it? And just as importantly, where have you seen maybe cases where there's been overly enthusiastic view of what it can do? You know, one of the and then the disappointment sometimes because.

**[0:35:24] Priya Gore:** It starts with, you know, Satish mentioned this earlier, really, I think the culture of the organization is everything. Because if you can, if you can safely and sort of, I'll call it securely, but then really just humanly describe what the technology is, take the time to educate people about what it can do and what it can't do, what it's really good at and what it should never be used for. you know then you're setting your employees and you're setting your teams up for success and you're setting them up to be part of the solution and part of the the equation versus feeling wow like leadership is coming in and bringing this ai technology that's likely going to take my job

**[0:36:05] Priya Gore:** away in three years no like there's two schools of thought right those companies that are behind closed doors thinking about those things versus those like ally that are thinking about how do we actually find ways to bring more value to our employees and bring more value to our end customers with innovation and leveraging this technology in a safe and secure way and educating people why we're doing it. What does it bring you for that incremental value in exchange for using it?

**[0:36:35] Priya Gore:** And finally, I would say like the example with client care at Ally and those teams now bringing forward their own ideas, I would say, you know, not to displace themselves and be not threatened, because they're excited and they see, as mentioned in the example, like their quality of care for their own customer went up, right? Because they were able to spend more human time engaging with their client versus like punching notes in a system because the system was kind of capturing and recapping and summarizing the call for them.

**[0:37:07] Priya Gore:** We also see examples, again, in the customer care realm where clients are, you know, they're looking for the bank or, you know, the insurance company or whatever it is. They want that agent to be like probing about them and what's going on in their life, right? What's happening right now for them and what do we need to do to help look out for them? That's what people want from their financial institutions, their insurance agents and things like that.

**[0:37:36] Priya Gore:** Generative AI is bringing that human capability, that ability to focus on people, people focusing on people forward. I think those are great examples. One last thing I'll say, the other gotcha that we've seen that we want our clients to really think about is you need to also think about your data strategy, because let's face it, artificial intelligence, like the intelligence of AI is everything to do with the data. And so if you don't have a sort of sound strategy for your data, and remember that this is a new wave of AI with generative, where now we can reason over unstructured data.

**[0:38:15] Priya Gore:** So documents, emails, images, you name it, any format, we can now, you know, reason over that information. And so it's very important to think about your overall holistic data strategy and really thinking about that so that you can really harness the power and the value of this generative AI technology and get the most out of it with this amazing and very valuable data that financial institutions have at their disposal. So sticking with you, Priya,

**[0:38:44] Jim Marous:** you work with a number of finance institutions right now. And it's funny because we're talking about this as if all this has existed for eight years and it's existed for one. But from the standpoint of what you're hearing in the marketplace, when you're visiting financial institutions and you're trying to show organizations what have been some of the really cool use cases, what are you seeing beyond Ally as far as some of the use cases that have been really like low-hanging fruit?

**[0:39:12] Jim Marous:** Yeah, I'll give you another one. You know, we talk about

**[0:39:15] Priya Gore:** generative AI being great for code generation. One of the hero, what we call co-pilot solutions within Microsoft that is out in the wild. The first co-pilot solution that we launched a little over a year ago was the GitHub co-pilot. And this is an example of a solution that we took to market where we can help developers actually have high value coding experiences at scale. And we're seeing a tremendous value there for our customers, not just in being able to be more efficient in building applications and checking in code and things like this as developers.

**[0:39:53] Priya Gore:** But also what we're hearing from our clients and the sort of findings and using these technologies is, hey, employees are actually expecting this technology to be leveraged. Developers are in high demand and customers are looking to employ and retain top talent in the space. And if you're not using the latest and greatest technology and tools for them to do their job and do it well, they're going to look for another place to go, right? So it becomes an advantage, right, when you start leveraging this technology.

**[0:40:23] Priya Gore:** I mean, you think about, you know, generative AI and sort of the era that we're in with this technology, as you mentioned, that's still very young. You know, if you think back, so it took 16 years for the mobile phone to reach 100 million users. Seven years for the Internet. Facebook, 4.5 years. ChatGPT, three months. So like it's an incredible moment in time, an inflection point, I think, societally. And, you know, because this is natural language prompted and Satish mentioned this earlier, I'm sure he has comments on this, like it's something that is capable of becoming democratized.

**[0:41:04] Priya Gore:** So I expect generative AI technology to be infused into every application experience that a company brings to market for consumers. You know, in consumer world, certainly we see it already with sort of even just the applications we use in our personal lives. And then certainly on the business side and in financial services, the opportunity is endless. Think about every application experience that you bring forward for your clients or that you use internally for employee enablement is going to likely have generative AI capability as part of it.

**[0:41:34] Priya Gore:** And the question you'll have to ask yourself is, do I build it myself or do I buy it? And that'll be a question that we'll constantly think about together. And the Microsoft team is geared to help our customers wade through that decision and help guide you along the way. Because it is new and we're learning as we go and it's an exciting time, but it's pervasive and it has the opportunity to be completely democratized across the human race, which I think is exciting.

**[0:41:59] Priya Gore:** But also, you know, a very serious matter that we need to make sure we're taking responsibility for together as technologists.

**[0:42:09] Jim Marous:** So one more question for you, Priya, is that we're seeing a lot of organizations, small organizations, doing just amazing things in the marketplace. But to do so, they have to find partners that can deploy solutions at speed and scale. I think reaching out directly, and I may be wrong, but I think if a small financial institution wants to reach out to Microsoft directly, that's a little bit overbearing from the standpoint of I'm over asking what I can actually use.

**[0:42:41] Jim Marous:** But I would imagine there's a secondary market of solution providers out there that are partnering with Microsoft on the financial institution's behalf where smaller financial institutions can make partnerships with those. Does that secondary market, for lack of a better term, exist today?

**[0:42:57] Priya Gore:** It does. We have amazing partners in the space of data and AI. We have amazing partners engaged with Microsoft along every part of our platform to help our customers embrace that technology safely and securely and give you that extra set of hands, if you will, an extension of your own team to bring that expertise forward so that you can go faster and engage with technology more quickly in your business. So we do have a mature partner ecosystem.

**[0:43:25] Priya Gore:** And I would say that we have some newer partners that have been really, really kind of game changing in the generative AI space that we're also working with. So I would say, too, like any company out there that's looking to leverage this technology, it's okay to reach out to Microsoft with a person within Microsoft that can then share with you this list of great partners that are capable of helping you scale the capability that you are building in-house to infuse this technology in what you do.

**[0:43:55] Priya Gore:** And that's not new for us. We're very proud of the partner ecosystem that we've fostered over the many years we've been doing this. But certainly in the generative AI and the data and AI space, it's something we look at regularly and we're constantly vetting our great partners to make sure they're ready to go so that we can make those recommendations and get those connections made for our clients.

**[0:44:15] Jim Marous:** You know, it's interesting to both of you that every day I wake up, it's been happening quite a bit now. I, in my mind, am going to come up with another idea on how GenreVe AI tools can help me. trying sometimes to find which tools, which platform is going to help me the best, I end up testing. So I think on my computer now, I think I have four tabs that have some type of AI tool that I test to see who's going to be the best at moving my thoughts forward.

**[0:44:47] Jim Marous:** And when you consider that that's happened on the consumer level, you know, Priya, you mentioned this, that the acceptance of generative AI has just been so astounding that you can't get to a cocktail party. We can't get to a podcast or a webinar without it being part of it. So Satish, part of it also is looking at those leaders and saying, where are they looking? What are they doing next? So if I want to get some specifics from you around what is on your to-do list in 2024 with regard to deployment of generative

**[0:45:24] Sathish Muthukrishnan:** AI, what would I see? There's numerous use cases that I'm particularly excited about, but I would like to talk about something that is different. And you mentioned this, AI, generative AI specifically, may replace jobs, but will not replace humans. but we have to help in that journey. This is an assistive and a technology that arguments anything anybody does on a daily basis. But not everyone has the ability or the opportunity to learn what generative AI can do for me.

**[0:46:06] Sathish Muthukrishnan:** People may not even understand what a tab is, Jim. You have four tabs of generative AI helping you out. And we call this AI equity. How can we stick to our mantra, which is do it right? How can we do it right for our customers and communities in helping them move along in this journey? I don't have a specific answer, but I will tell you, we are spending our brain cells, we're spending time, we are collectively thinking about how to make an impact.

**[0:46:42] Sathish Muthukrishnan:** And maybe we will call on Microsoft to help partner with us in educating the larger society and helping them understand how generative AI can play a role in uplifting

**[0:46:53] Priya Gore:** their well-being. Yeah, I just would add a quick comment on that. I agree with you completely. I I mean, we have a wealth of content and self-paced learning material out there in our AI business school curriculum. We're refreshing that right now. And next quarter, that will be readily available for anyone to leverage. But I would say beyond that, you know, we also owe it to ourselves to, like, share best practices of things we're seeing.

**[0:47:23] Priya Gore:** I was with a customer a few days ago last week, and they were talking about how, as they were engaging generative AI into some of their solutions they've taken to market, they've actually launched a couple of products, much like, you know, same pattern that what Ally's done. And they made sure that they, they actually told me that they produced like TikTok-like videos of what is generative AI and then certainly like building on those modules. of building out material that talk about, okay, what is this solution?

**[0:47:50] Priya Gore:** How should it be used? You know, because it is, it's going to be a new skill, I think also for people to engage these generative AI infused co-pilot solutions. And to Satish's point, these are designed to be assistant in nature, not to, we named our stuff co-pilot, not autopilot for a reason, right? And that's the intent, right? Is to augment and to bring value, to work alongside you, next to you, at your desk, not do it for you and replace you.

**[0:48:19] Priya Gore:** The human in the loop is going to be very, very important. The expertise in financial services, great example, the expertise of the banking professional will absolutely need to be at the center of everything that we do, client facing, to have credibility. And so the work we do will change and how we go about getting work done will change with these assistants and these co-pilots. But I do think that the ingenuity of the human and the expertise will still need to be harnessed and the value you'll bring, you'll be able to up-level the work you do every day if the adoption cycle and sort of the learning cycle of how to use these tools is cared for and thought

**[0:48:56] Priya Gore:** about by enterprises as they bring this technology in. I think that will be, it's really what we need for, frankly, the human race in order for us all to really get our heads around what's happening here and do so safely

**[0:49:08] Jim Marous:** and responsibly. You know, that's interesting because I think we forget where we each started in this journey. When people initially would say, you know, I think gender of AI is a bunch of, you know, crock, it's not good. I finally, you know, you very quickly get to the point going, rather than damning the technology, maybe you have to look in the mirror and say, are you asking it the right questions? And we all realize as we've gone forward that in the structure of questions, it provides you just amazing answers.

**[0:49:39] Jim Marous:** And when you get that aha moment that says, oh, I asked this a little differently and I got like massively what I wanted, you realize you have to think a little differently for talking to technology. And so in our last question, I'm going to give you each something different. Satish, from your perspective, what really excites you about what could happen next year in financial services and in Ally with

**[0:50:07] Sathish Muthukrishnan:** Genre of AI? Jim, we recently launched what we call One Ally, where we are bringing all of the experiences across all of the products into a single digital experience. That connected to what I stated earlier about customers' expectations of moving from more personalized experience to a more personal experience, Gen.AI can help power that. So we treat every customer of Ally as if they are our only customer. The power of segment of one can be realized through generative AI.

**[0:50:48] Sathish Muthukrishnan:** And collectively as an organization, how we can differentiate ourselves in serving our customers is exciting me tremendously for 2024.

**[0:50:58] Jim Marous:** Great. And Priya, from your perspective, if somebody's just looking to get started, where do they start?

**[0:51:07] Priya Gore:** Yeah, I mean, listen, you think big and you start small and give us a call. No, I'm kidding. That was just a little... No, but seriously, I think we are in a time of what we call the era of AI. And I think that as you're just getting started, You know, I think you really do want to brainstorm these use cases. Think about, you know, nothing's off the table, right? But then I would say, you know, start with something like crawl, then walk, then run, right?

**[0:51:35] Priya Gore:** So start with something low risk that you can do to get familiar with the technology and get familiar with how it works and how you infuse it into, again, application experiences. Remember, these generative AI models are APIs. So you bring them into an architecture of an application along with other services. and that's how they come to life. And so learning a bit more about that and we can help you with that to get familiar with how they work and what they're good for and what they're not great for is the first step.

**[0:52:05] Priya Gore:** And then I think also like, you know, looking at some of the existing, for example, co-pilot solutions, whether it's, you know, something custom that allies brought to market and you're a banking customer of theirs and you're leveraging their tools, great way to learn more about how this stuff works by, you know, having a personal experience with a technology and an application like that. And then certainly like our co-pilot, our Microsoft 365 co-pilot solution, which touches many, many knowledge workers, right?

**[0:52:31] Priya Gore:** We're starting to roll that out with our clients that are in the early access program and getting a lot of great feedback on how that's really improving and empowering people at the knowledge worker level. And again, that's a great way to get started. So I would say, you know, think big, start small, do your research and really do call us and we'll engage partners that are really able to bring you along and help you create centers of excellence and governance teams that can help make sure that you're also thinking about this from an ethical and sort of responsible AI perspective.

**[0:53:06] Priya Gore:** We've got resources there to help with that as well, which obviously is very, very important in the regulated world of financial services.

**[0:53:14] Jim Marous:** Satish, Priya, I cannot thank you enough. This has probably been my favorite. I've said this before, but things happen. My favorite podcast I've done, not just because of the knowledge you shared, but the leadership that you each individually have shown and your organizations continue to show, and you reinforce everything that we keep on trying to bring forward in this podcast, which is it all starts with leadership. You need to have the people to support that and make it so that it's democratized what we're doing and that it can be deployed against any size organization today.

**[0:53:55] Jim Marous:** Again, thank you so much. I'm going to have Leah contact you very quickly because we're going to set up another call a year from now and see where everybody is because I know can be very different than it was today. Thank you so much for being on the show.

**[0:54:08] Priya Gore:** Looking forward to it. Thank you for having us. Thank you very much.

**[0:54:12] Jim Marous:** Thanks for listening to Banking Transformed, the winner of three international awards for podcast excellence. We really appreciate the support we've received to make this an endeavor a success. If you enjoy what we're doing, please take some time to show some love in the form of a positive review. Finally, be sure to catch my articles that I'm doing for the financial brand and check out the research we're doing at the Digital Banking Report. This has been the production of Evergreen Podcast.

**[0:54:40] Jim Marous:** A special thank you to our senior producer, Leah Haslidge, audio engineer, Chris Fafalius, and video producer, Will Pritz. I'm your host, Jim Marous. Remember, the future of AI-driven banking will often involve the collaboration between banking and tech leaders. Thank you.
