Okay, good. Good morning. Good evening everybody. Thank you for being here. My name is Kasi, and today I want to talk about a fundamental shift in how we build and deploy artificial intelligence in the enterprise softwares. We are moving beyond simple, isolated AI tools, and we are entering into an era of orchestration. The presentation is about the. Designing multi-agent AI systems that can intelligently navigate our most complex environments, which is our ERPs and how our hybrid IT systems. For the last few years, the enterprise AI has often looked like this, a series of silent. Pilots. For example, a marketing chart bot here and a finance tool there, an onboarding tool elsewhere. They're all totally disconnected, fragmented, and it requires a ton of manual overhead to make them talk to each other. The model is. Often broken and it doesn't scale the future. And what we are building today is an adaptive ecosystem. This is a move from isolated experiments to coordinated networks of AI agents that provide enterprise wide intelligence. The they orchestrate data. And most importantly, they can autonomously optimize enterprise workflows. This is the challenge and opportunity we will be exploring. So exactly what is exact, what exactly is a multi-agent system. It is a helpful, it is helpful to stop thinking about one single ai instead, think of a team of specialized experts, autonomous agents, coordinated actions, and a goal directed steps. Autonomous agency, the agent is an AI entity with a specific role. You might have a data analyst agent, you might have a compliance agent, and you might have an inventory agent. Your data analyst agent, again, can be subdivided into specialized in bioinformatics, for example, genetics, for example, and whatnot. When it comes to coordinated action, these agents don't just have to work alone. They must be able to communicate, collaborate, and delegate tasks to solve problems that are far too complex for any single agent. And they do this with a purpose. The system is given a high level enterprise objective, reduce procurement cost by 10%, and the agents work together to achieve it with minimal human intervention. So next, what are the main four main pillars for this enterprise? AA orchestration. There are four pillars for this orchestration. The first one is a goal directed planning. Always planning is the most important one. The system must be able to take a big complex goal and break it down into small and executable tasks for each agent. Second, dynamic reflection. This is crucial. The system must be able to look at its own performance. Learn from its successes and failures and get smarter over time. Third, tool orchestration. The agents are useless if they can't interact with the real world. The pillar is about giving them access to your existing tools, your APIs, your database, and your ERPs. And finally, collaborative problem solving. This is the multi-agent. Magic agents share insights and coordinate actions to tackle challenges that cross departmental silos. This isn't just theory. These are powerful frameworks available right now that allow us to build these systems. Auto Gen is coming from Microsoft Graph is a graph based structure, which is perfect for complex cyclical workflows where agents might need to rethink a problem, what we call a cyclical reasoning. When it comes to auto gen, it is very excellent for creating conversational agents that can solve task together and crew ai. Provides a role-based framework. You literally define your agents like a team, a senior researcher, a writer, a manager, and they collaborate based on the hierarchy. Okay, let's make this real. Where is this being applied or is it just theory? These things are getting applied in procurement applications, in procurement business in finance, in hr, and in document management. For example, when it comes to procurement, I'm seeing automated supplier evaluation, contract compliance monitoring, running 24 class seven. Across global supply chains. When it comes to finance, this is a game changer. Think automated invoice processing, real time expense reconciliation, an anoma anomaly deduction that not only finds fraud, but also generates the audit trial in hr, we are automating the entire onboarding workflows using conventional interfaces to resolve employee queries instantly. And when it comes to document management, agents can classify, extract, structure, data, and route documents to the right person, eliminating massive bottlenecks. How? How are the business impact measures and the business impact is not trivial. These are real measured results from multi-agent deployments. We are seeing an average of 65, 60 8% reduction in processing time in workflows like procurement and finance, and 92% improvement in accuracy for tasks like data extraction and a 45% gain in cost efficiency by optimizing workflows and removing. Manual, repetitive work. This isn't just incremental increment. It's a step change in operational efficiency. Next comes the integration between Hi between different EA agents in ER within ERP and also in hybrid IT landscapes. Now for everyone in this room who manages enterprise it, the first question is, this sounds great, but I have a massive complex SAP system and Oracle system. How does this possibly integrate? This is the most important part we don't rip and replace. The beauty of multi-agent framework is its modular architecture. You can adapt it incrementally. The agents connect to your existing systems, both cloud and on-prem through standard APIs and middleware. The you can start by automating one small part of the process. This approach massively reduces risk, accelerates your time to value, and lets you modernize without breaking the bank or your business. Next is about the critical implementation challenges. Of course, this is not without challenges. You have to go into this with your open eyes. There are three critical hurdles, security, data, privacy, scalability, governance, and compliance. We are giving agents access to sensitive data. This demands, I'm sorry, this demands robust. Authentication and governance, and it comes to scalability. As you grow from 10 to 1000 agents, how do you manage computational resources and cost? When it comes to governance and compliance, this is a big one. If an agent makes a decision on behalf of somebody you must have an audit trial on. What does it think? Why does it make a decision? And you must be able to satisfy the regulators and build the trust based on your audit logs. Next is what are the strategies for safe adoption? So we spoke about challenges. We spoke, we got excited, we spoke about challenges. Now it's time for us to know how do we address these challenges. First is comprehensive audit logging. Second is bounded autonomy. And finally, incremental deployment. So we briefly spoke about audit, logging. Every decision, every action, every piece of data and agent touches must be logged for transparency, bounded autonomy. This is a key concept. Agents are not given free reign. We define clear operational boundaries, and most importantly, we program escalation path for human oversight. A human is always in the loop for critical decisions. Finally, incremental deployment. Don't try to boil the ocean. Start with a low risk, high value use case. Validate this results, then expand. Next is building your multi-agent roadmap, right? Your path to adoption must look like this. First is foundation, followed by piloting. Next is scaling and finally optimizing it again to summarize this right all what we are saying is don't try to boil the ocean. Access your current state, and find that the first high value use case is going to be implemented. Pilot it. Define your first agents in a re controlled environment. Measure everything. Scale, expand the successful partners across other departments. Now optimizing. This is where you truly orchestration, orchestrate. You enable autonomous agent collaboration and create self-learning capabilities. Next is the end goal is here. End goal. Here is a fundamental shift in thinking. We are not just building tactical projects, we are building a self optimizing digital ecosystem. One where agents learn, adapt to changing business conditions and coate autonomously creating enterprise wide intelligence. Okay, next is our playbook. First is Playbook is a very clean one. First is you have a design. Second is integrate and then Scale. And what is our key takeaways? The key takeaways from the last 10 minutes are Multi-Agent systems are next. Evolution of Enterprise ai, moving us from isolated tools to coordinated systems. Practical frameworks like Auto Gen, land Graph, and Crew AI are available to build today. Today, success demands a balance. You must balance between innovation and governance and security. And finally, the path to success is to start small, think big and more incrementally. Thank you so much. Again, my name is Kasi Pal. Thank you for this opportunity. You can see my LinkedIn LinkedIn, URL here. Thank you so much.