Hello everyone. I'm I senior software engineer manager in payments, and also an independent AI searcher. Over the past several years, I have worked extensively in the financial services domain, leading the design modernization of cloud native system, realtime transaction platform, and AI driven risk management solutions. Today I will share how generative AI is reshaping financial services. Transforming raw transactions into personalized financial journeys and intelligent decision making. I will go to the, go to my first slide, which is our agenda. So we are going to introduce what generative BI is and how it matters and FinTech, and then we will eventually follow the evalu evaluation of gender financial services and ai. And then generative AI in action, like narrative personalization risk, then we will address the challenges and consideration. And then we will move to how do we overcome those kind of challenges? And then what is the future? How do we utilize it to show the generic pi future in the finances and then statistical case study and then take hours and conclusion. With that, I will start with my first slide. So finance is a data-driven industry, so for decades, banks have collected structured ledges and transactions, but until now, those transactions were treated as an isolated event. Generative AI changes that it can combine structured and unstructured data, like spending history of strict transaction and even voice data into dynamic contextual insight. For example, instead of just recording that I spent $50 at Starbucks generative AI can possibly mention that it has, spending $50 at Starbucks has given 18% right to rise to my expenditure, minimizing certain percentage of my savings. So basically it is a fundamental shift from transaction to financial storytelling. So how the evaluation of AI happened in finance. So it's not new. It has been there with the finance industry from 1990, but obviously at a different scale, completely at a different scale. So in 1990 to 2000 if you see the fraud alert credit scoring, that these are already part of the financial industries and that is also some kind of rule-based system integration adding a lot of value to it. It is very rigid, but it has a lot of false positive outlook. In 2010, machine learning improved personalization and risk scoring better, but still predictive. In late 2010, deep learning brought a realtime fraud detection and LP power chatbot. This is a serious style banking. So in 2020, genetic ai, it's not just detecting or predicting the fraud, or it is also creating the new output, like genetic rewards simulating risks, producing personalized recommendations. So in short, it is not only about setting the problem, but also proposing a solution of how do we minimize it. So let's see. What is generative AI in finance industry? It's not just doesn't classify your credit, it creates the insight and actions. Like I just quickly, you briefly mentioned in my previous slide, it can write personalized financial reports directly from broadband data. It can similar crisis in annual. Somewhat of what if things, right? So what if the employment is gets a spike and interest rate double. It can also propose a custom loan and investment products tuned to a customer profile. This means finance shifts from what happens to what should we do next. Like I mentioned in my previous slide, as it's not only predict service, but also proposes a solution of how do we handle that. Moving on. So transactions to narratives. So what is narratives? So narratives is as simple what explanation of what just happened, especially especially for the younger generation who doesn't, who never tries to look at the statement, but probably. If you just tell them that your expenditure has been increased or this is how you save more, probably that is something would be very easy and quick to grab their attention. So every customer leaves behind a financial data trail pending saving, investing. So generative AI teaches those into storage. So instead of statistic or static dashboard you are going to get contextual narrative. For example, the first slide or the second slide I just mentioned that spending $50 at Starbucks means that 18% higher expenditure affecting my savings this month, or based on my investment history, I might miss out some of the tax saving. So this is something is a storytelling function, so it catches more people, it catches attention, and also from the layman user, like the consumers who are actually using the finals but not into the financial industry, it is very easy to grab their attention and help them. So personalization. Traditional banking is one size fits all. So they have the limited four, five categories of the products, and we just have to customize our need to make sure which product fits with generative ai, enable sales to conventional assistant that answers plain language, dynamic loan offers interested, tailored to our realtime profiles. Adaptive investment advice that address when market shipped. Our major European bank reported that AI powered personalization increased cross sell conversion by 35%. So obviously for individual, instead of using those five, six predefined categories, it helps to tailor loan and the percentage and the usage for everything. So for example, that we have a home mortgage loan of 500 k generating an EMI of 30 three K every month. And that is supposed to be paid in 20 years. So what if you pay a thousand dollars per month extra? What if you pay a hundred k extra? So what is the duration of the loan that reduces to what is the EMI that reduces to, those are the kind of personalization that is very easy to opt in through this generative ai and give an answer to the tailored answer to the speak consumer. Last, but last, but not the list. Free scan fraud intelligence. On the different side, fraud. Fraud are using ai, right? So there are so many fraudulent activity that is trying to attack our system, and they are all using ai opting for the smartest way to hack a system. So if we use the generative ai, we can defense at the same scale that we are being hacked or at least trying to be hacked. So generative AI helps to create synthetic fraud patterns, to train models that are more robust, similar adversial attacks, like fake transaction even before they happen. It builds stress for liquidity and credit crisis. This moves banks from being reactive to a proactive guardian, and we can just think of it as a digital vaccine. Instead of the time we are hacked, how to try to get out of it, making sure our system doesn't fail. We are developing tests in, or test suit in such a way that we are actually isolating the. One in a million kind of possibilities and the age cases. And we are actually and proactively testing our system against those kind of attacks, making our system more prepared to handle those kind of AI generated attacks, which are more vulnerable than any other, any of the previous that used to happen. Yeah, but nothing is free of course, obviously any big innovation comes with a bigger responsibility and definitely Generative BI is not free of that. It definitely has some challenges and consideration to address. So bias, obviously there are people who are financially privileged, there are consumers who are not that financially privileged. So we definitely, while using the ai, we must take into consideration that it should not be the case that we are implementing security in such a way that those who are more financially stable are getting to those security versus those who are not that financially stable. If. Getting reprint from ING there. So we just have to maintain a balance and kind of implement in such a way that cost should not be a bias at the consumer level, making those kind of facilities available to all level of consumer. Privacy banking data is amongst the most sensitive data. And it has to go through so many laws. It has G-D-P-R-C-C-P-A, so we just need to make sure that we comply with those kind of laws. Otherwise miss you utilizing of those kind of PIs can cause bank a lot of mail. Explainability regulators won't accept this bag box answer like modern LLS. Most of the times we just keep command and things get executed. So we do have to deal with all the regulations at the bank admin level, and we need to make sure that we have answers to any kind of justification on any kind of output. Security model sensor can be attacked, obviously because that's a system. There like we used to have in old time R-D-B-M-S, we used to have the SQL injection, so like that we can have a prompt injection. So we need to make our models prepared for all this kind of setup. So success is not about just the accuracy, but it's also about trust and fairness. So it is not an impossible to overcome those kind of challenges, but while designing our system, we definitely have to make sure that we take all of this into consideration. So privacy preserving generative ai. So obviously with that this is also again, a part of the risk and especially with the banking data. There are lot of security issues. There are a lot of privacy related issues. So how do we do that? We have to implement the federated learning so we can have, we can let the bank collaborate without sharing the raw data. So we can definitely think through that direction to protect the privacy. At the same time, use such a net. B, we can use differential privacy privacy that ensures output can't expose individuals. Explainable ai. So this kind of the, in the previous slide, this is one of the challenges that we need explanation for the regulatory and compliance things. We cannot accept the black box answers. So to do that, there is explainable ai xai that helps regulators and users understand the decision that this is what happened, this is what is rejected, why this has happened or rejected. Without exposing the direct rules, but also harm some user understand format of the rules. So when these are combined we would definitely get systems that are not only powerful, but also ethical and reor regulatory ready. So how do we use it in future like autonomous finance? Let's imagine where this technology is heading. We are on the path towards autonomous finance, but many financial DI decisions are optimized in real time by intelligent systems. So here are now three emerging directions that our finance industry is heading to. So multi-agent AI systems. So instead of using a single model making decisions, there are specialized agent work together, farming and agent TKI. This Azure system that has one agent for risk, another for liquidity, another for compliance, and they can negotiate balance, trade off, self optimize. For example, a risk agent might signal an alert and liquidity agent might automatically relocate funds to mental stability. So that is fully automated, right? Without any kind of manual intervention. So personal finance copilot, so just as many as. As many of us uses the GPS today for any kind of navigation system, even for searching a gas station, the same way we can have an AI copilot to for balancing our finances. For example, the it can remind us for the bill pay, it can remind us of the credit card payment. It can schedule the payment for us in a more advanced way. There is some account that is running out of balance. It can transfer funds between that so it can actually start or manage the finals in an autopilot mode. From optimizing our monthly savings to choosing the, choosing the right tax advantage investment to automatically adjusting the portfolio to as the market shift. And the third and the biggest one is the large scale financial and simulation regulators and central banks could be AI driven simulations of entire economics to credit systematic risk. Imagine being able to stress test, I release of a new currency, like a digital dollar or a euro of hundreds of scenario before launch. So this is finance that is resilient and in process, but again, success depends on embedding fairness, privacy, and transparency on every stage. Moving on to bring this down from vision to practice, let me share a concrete case study. Personalized wealth management assistant, we piloted. So we will have an input process and output. So from the input perspective, it we will, it would be a system ingested customers, real world data spending, pattern setting behavior and investment preferences. So what do we process in it? It would be, it would use the generative bi. It'll create the customized report projections and even what if scenarios. For example, if we increase monthly savings by 5% maybe we will reach our retirement three years already again. So what do we do? What is our goal and what do we do to achieve that? Personalized proposals for a solutions and output. Generated tax optimization strategies. For example that we should have there was a tax investment scheme, like we should have invested in four one K more. We haven't. So that kind of ended up with this much of tax tax credit, which we could have easily avoided by using this, and we can probably plan that next year. So what do we input, based on our real time data, the process is the what type analysis and output is. What do we actually do to avoid what we didn't the same. So the results were remarkable. The customer's engagement increased by 40%. The financial literacy score improved significantly because the customer finally understood the money in plain language, which I say that from non-financial consumers payments were, it is, it was never more than just a number. Just the numbers and some balance, but they never used to understand finance by looking at it statement over last 30 days or 60 days. It is probably just random numbers for them, but with this generative AI and simplified language and the storytelling, like financial storytelling, that helps the name and consumer to understand the financial language and plan accordingly. So this demonstrate that the generic PPI isn't just password, it can drive business KPI's, customer satisfaction and regulatory alignment. Moving on key takeaways. So as we, there are four messages that I would like to carry forward. Generative AI is moving finance from data to experience. We are no longer limited to raw transactions. We are building personalized financial journeys. It enables both customer value and institutional resilience on the front end. The hyper-personalization improves engagement. On the backend, the synthetic fraud scenarios and stress, strength, and defenses responsible AI is non-negotiable. Privacy, fairness and explainability are the nice they're not just nice to have. They're regulatory, ethical, and strategic requirement. The future is autonomous finals. Multi-agent system, personal finance, co-pilot global financial simulations will transform not just the banks, but also society, not also how society interacts with the money. If I had to summarize the winner, sub next ticket in finance will be those who can balance innovation with responsibility. So with this note, I just would like to conclude my presentation. So generality, BI is not just transforming finances, it is humanizing it, turning core transaction into meaningful journey. Thank you.