Hello everyone. Welcome to Con 42 Machine Learning 2025. My name is Bala. Kris, thank you for joining me today. We all know that AI is the future, but the fact is AI is already here. It is in the apps we use and the products we buy and the companies we work for. Let me tell you a big story. There was a delivery company with a big problem. The packages were always late. Customers were complaining about the late deliveries. Then they tried something different. They used AI to help drivers choose the best routes. The system looked at traffic, weather, and distances and gave the real time suggestions. After a few weeks, the delivery times are improved and customer satisfaction went up. Is the power of ai. It is simple, helpful, and effective. Today, AI helps with so many tasks, stopping frauds in the bank, helping doctors find diseases faster, and reading job resumes and suggesting the best candidates and showing your ads that match your interests, even helping students learn better on online. So what are we going to cover today? I will show you how fast AI is going and what kind of systems companies are building examples from real time industries like hr, finance and manufacturing, and what problems businesses face with ai. And finally, where this is all going next, let's begin. Yeah. Let's talk about the area market is going fast. A is just not a cool tool anymore. It is becoming part of how businesses run every day. In 2019, the global as software market was about $28 billion. By 2027, it is expected to reach $315 billion. That is more than 10 times bigger. In less than 10 years. What is driving this growth? Let's talk some of that scenarios. Companies want faster decisions and they want to save time and money, and they want better experiences for their customers. Let's take one example. Okay. Online store had thousands of product reviews. Customers were saying all kinds of things, like some good and some bad. The companies use IA to read those reviews and understand. What customer liked or did not. Based on that they improved their products. This leads to fewer returns and happy yet buyers and companies use it. AI to remove the checkout lines in stores. This is one of the, another example. Cameras and sensors track what people picked up and charged them automatically. No cashiers, no lines, just a code that is AI in action. Let's now see how these companies build such smart systems. Let's go to the next slide. This is intelligent automation architecture. AI does not work alone. It needs a system behind it, just like a car needs an engine, fuel and a driver. Here is how it is. Usually looks data ingestion. This is where we collect the information from different types of sources like websites, sensors, forms, or anything, processing a modules. Use this data to find patterns or make predict predictions and analytics. We turn the results into charts, scores, or LX integration. We send these results to other systems like an app or dashboards presentation. People see the outcome like a warning, a recommendation, or a report. A bank, for example, might check every transaction with an A fraud system. If something looks strange, the system flags it immediately and alerts the analyst. This is a smart system in action. Fast, secure, and connected. Let's look how. This helps in HR and finance. Next. Now let's go to the next slide. Hr, analyticals and Financial Intelligence Systems. We can call it as AI in HR and finance. In human resources. HR companies are afraid with regimes. Reading them all take days. Now ai, AI tools can scan regimes and pick out the skills. Match people to the right job within minutes. They also check for bias and try to making hiring fair. One company went from 30 days hiring cycle to just eight days. Witha support in finance, AI helps in different factors, trade, market risk, and sports, suspicious payments, and rate long contracts and highlight the key tabs. One bank is an AA model to read the compliance documents. It's used to take three weeks for a team, but the AA is finishing just two days with the high accuracy. So whether it is about hiring the best people or managing the money better, AA is becoming a trusted assistance. Now let's talk about the decision making. What happens when AI helps us decide? And now let's talk about the decision support system evolution. AI is becoming more than a helper. It is becoming a decision partner. This is called a decision intelligence, and it's growing fast. In 2022, the market was about 11 billion by 2030. It may reach up to 44 billion. Let's say you work in a hospital, NAA system can look at the symptoms, test results, and medical history. They then suggest possible diagnosis or treatments. Doctors still make the final call, but they have the most support and information which they extracted from the aa. Or if you talk about in. A might tell the managers which products are running low and when to restock. This decision happens in real time. For example, a good decision system needs rules like what are all the guidance we have to follow and data like the history or the live updates and the AM models to analyze and the feedback from the humans. A financial form use AI to rebalance risky investments. They combine news reports, market signals, and the past behaviors. This is reduced, the risk loses by 25%. Let's move on and see how these systems all connect using the real time integrations. Now let's talk about the next slide. Integration patterns in modern AI systems. Now that we have seen how AA helps with decision making, let's talk about how the systems are connected. AA does not work by itself. It takes to the, it talks to the other systems and you chase the data and responds to the events happening in a real time. This is where the integration comes in to think of a smart home. The light turns on when you walk in. The thermostat adjusts when you leave. All the individuals are talking to each other. A company works at the same way. Here are some ways AI is integrated, real time data streaming. Tools like Kafka, or he even helps, sends live data to the AI models. This helps in fraud detection or monitoring missions. Even driven systems like their response when something happens. For example, when a user logs into the system checks for a risk and triggers a security scan and streaming analytics. Instead of waiting for reports, a HX data as it arrives. This is used in stock trading, weather alerts, and even food delivery Timings, hi hybrid setups. Some models runs in real time, others updates overnight. This keeps the systems faster and accurate. A quick example for this scenario, a flight company checks the engine data during the flights. There is a strange, significant, the AA alerts the ground team before the plane even lands. That is how powerful a the realtime AI can be. Before we go to the next slide, let's explore F1 and usual part of aa, which is chat bots. We have seen them on websites now. We will look at how they actually work. Now let's talk about enterprise chatbot architecture. Chatbots have become our first line of communication with many businesses. When you ask a question on a website and a bot replies, it is more than just pre-written answers. It is a whole system working together. Let's see how it is built. The user interface. This is where you type your questions. It could be a website, a WhatsApp, or a even voice, and it understands the languages. NLU, we call it as NLU. The AI reads your cushion and figures out what you mean, and the next one is the dialogue manager. It keeps the conversations going and remembers what you have said earlier. The backend connections. This is where it takes to the company systems like checking your balances or booking an appointment. And the next one is security and tracking. Everything is logged and kept safe. Chatbots are used in different industries like hr. Asking for a leave are checking the salary informations and it supports resetting the passwords, opening the tickets. In healthcare, booking the appointments with the providers, sending the remind reminders about your future appointments and retail checking orders are giving the recommendations. Example, a quick example here. A telecom company launched a chart board in four languages. It handles 68% of customer requests. Without a human, it save a lot of time. And reduce the pressure on the supporting team and give users a faster experience. Let's see what makes the systems hard to manage and how businesses are solving it. And now let's talk about the technical challenges in enterprise. Aa a AI sounds amazing and it is, but it also brings some serious challenges, especially in big organizations. Let's talk about some of the top issues we are seeing with AA data privacy. AA systems often use personal data. It must be protected. Rules like GDPR and HIPAA are very strict about this and model rate. Over time, the world changes multiple last month may not work today. AI needs to be adjusted for this one. Scaling problems when 10 users becomes 10,000 systems must be fast, reliable, unstable, and bias in models. The training data has bias. AI may give unfair results. We must test and monitor carefully and human review. Sometimes a should not make the final call. Human should step in for complex absence to issues. For example, a financial company tested a new model for approving the loans in shadow mode. They decide it was giving the lowest scores for the recent immigrants, and they passed it and fix the issue and launch it later with better rules. And coming up, we talk about the infrastructure. How do we manage all the data models and try in a smart way. Now let's talk about the next slide. AI fracture and resource management. Behind every smart AA system, there is a powerful setup. AI needs data pipelines, model training and deployment tools, and people who monitors it all. Let's look at the key bots automated pipelines. Example, tools like Airflow and Q Flow help more data and trying the models regularly. CSCD for aa. Just like software get updates, AA models also need to update it. The essentials, they stay accurate and safe. A, some AA runs on small devices like phones, cameras, or even drones. These models are smaller and faster future stores. This holds common data pieces that model uses like location, age, or last purchase, human in the loop. In some cases, people check or adjust. What I suggest, this helps with training and trust. A quick example for this scenario, a drone company uses small models and the drone itself to detect cracks in the pipelines. It work without needing to send the data to the cloud, and it saves the cost and time. Let's no talk about the future. What is coming next in the AA and why it is matters. Now let's talk about the future directions, RIML, and expandable a. The two big things are shaping the future of the aa. One is Rml. We call it as automation learning. This is a AI that builds other ai. It helps teams quickly test and pick the best model, even if there are no experts and expandable ai, we call it as XA. This helps us understand why a AI made a decision. It is very important in banking, healthcare, and loss examples. A bank is not to create a credit approval model. Then it used explainability tool like shop to show customers why they were approved or denied. It builds trust and reduce complaints. This tool get together Auto ML and expandable AI makes a faster to build and easier to trust and know. Let's wrap up it. Let's bring all it together. Now let's coming to the conclusion, let's recap everything we have talked about so far. AI is changing how companies work from decisions to automation, and it is going fast and it is being used in all industries. Smart systems need good design, integration and monitoring. We must think about privacy. Fairness and keeping humans involved and the future easier. A without IM, L and more trust with expandability. And one final thought. A will not replace people, but people will, who uses a will replace those who don't. And thank you all for your time. It's Bill aa. That is not only smart, but also useful, fair and trusted. So thank you all so much for being part of this session. AI is not just a technology. It is a way to solve the real problems, make better decisions and support people in their everyday work. As we move forward, it will say systems. They are smart, yes, but also simple, safe, and helpful. If you would like to connect, ask questions, and continue this conversations, feel free to reach out to me and my email ID is contact bama@gmail.com and my LinkedIn is available in this slide. I'm always happy to share the ideas. Helping projects are, just talk more about this exciting journey. Thank you and enjoy the rest of the conference party. Thank you.