Good morning, good afternoon, good evening, everyone. I'm Kumar End from Chicago, Illinois USA. First of all, thank you for investing your time to join this conversation with me here at conference 42. The topic of this conversation is beyond production, how generative AI is reshaping the future of a smart manufacturing. Over the past decade, we have heard the worst industry 4.0 and artificial intelligence so often that the risk sounding like yesterday's news, yet something fundamentally new is happening right now. Generative AI is moving us beyond prediction and dashboards into a world where machines create new designs, new workflows, even new business models every day. This talk is not a technical deep dive. Instead, I want to paint a clear business picture, why generative AI matters, what results are already being achieved, what challenges you must anticipate, and how to get started without ing day-to-day presence by the end of the next half an hour. You will have a tested framework you can take back to your team. Let us begin with the outcomes that matters most and that make executives and shareholders sit up and listen when the board asks why Now, why generative ai? There are some numbers that becomes important and which cannot be ignored. Almost 40% sort of product development timelines is achieved at an aerospace supplier. We studied digital parts. Once took almost 18 months from concept to certification. After embedding generative design, generative AI design, that cycle failed to almost 11 months. And in another case, it was almost nine months while still meeting FAAA standards. Almost 35% improvement has been seen in overall equipment efficiency. A consumer package goes CPJ, planned to use a generative AI scheduling agent. That sequence sqs hour by hour and planned downtime drop by roughly a third, and overtime costs fell right alongside it. Almost 40 per 42, 40 to 42% lift in successful innovations have been seen. Think of this as an idea to impact velocity. A tier one auto supplier did not merely brainstorm more concepts. They commercialized more of them because generative as simulation flagged and factor ability issues before the first prototype. These gains are not edge cases. In meta-analysis across more than 60 plants on three continents, we see similar order of magnitude. The takeaway, generative AI turns marginal, kaizen gains into step change performance. Two, appreciate why the gains are so large. Plus, compare conventional AI and generative ai. Conventional or traditional AI works with the boundaries of past data. It forecasts outcomes based on historical trends. It is predictive and reactive in nature. However, genetic AI, on the other hand, explores what could be. It models multiple alternative features. It synthesizes novel solutions and a number of them. In manufacturing, this means designing a new process, a new product configuration or even a new materials before a problem arises. That's for illustration. A predictive model wants that a press break is likely to fail in next 24 hours. That is valuable information. A generative model asks. What new process or press break configuration would boost throughput by 20% and minimize maintenance events in the first place. Then it produces a 3D Toolpath while you can send a straight to your CAM system. In short, generative AI collapses the gap between analysis and creation, and that is game changing for manufacturing. Environ. Now let's take this to production floor. Let's zoom into three day-to-day benefits. Every plant manager cares about number one, eliminating the bottling, realtime layout engines, EST sensors, feed, and automatically surfaces between metal flow, canon position, and even a MR that is autonomous mobile robot. Routes. Plants report almost 30% fewer micro stops, and a visible is smoothing off at times. Number two is cutting energy by almost 20 to 25%. A European glass manufacturer used a generative AI energy optimizer that predicted furnace, thermal inertia, hours in advance. Power drawn now rises and falls with the heartbeat of demand rather than staying flat out a savings of almost 2 million euros annually and a happier ESG scorecard. A third and most important aspect is hours, not weeks in changeovers. That's quick changeovers between processes or operations. During Covid, one electronics line needed to swap from automotive sensors to medical ventilator boards. Historically, that would have been at least 14 to 15 day revalidation Generative VA process planner produced new SMT recipes overnight Fast pass yield was 97% the next afternoon. Individually, each metric is impressive. Together. They create an operational operations. Flywheel throughput funds more AI plots or pilots, which uncover more savings and so on. Manufacturing strength means very little without a robust supply chain. There is some example where generative AI science, knowing about disruptions in advance, for example, by blending weather feeds. For traffic, APAs and supplier health indicators, the model can flag the material shortage week, two weeks, or even a month in advance. Prompting, preemptive action that can be taken to streamline or smooth in the supply chain process, identifying alternative vendor or supplier instead of a static approved vendor list. The generative AI dynamically compiles and risk scores, new suppliers worldwide. Complete. Complete those with the compliance documents that can be clicked through. Last but not least, logistics auto rerouting. There is a possibility that traffic is closed or there is a closed place or some accidents. The AI agent rerouted or optimizes the routes reax containers by cube key utilizations and reissues updated to ETS to customers. This all happens before a human planners logging. Ultimate result is this, all pivot happens in our not in weeks. In a world where a stuck container can cost billions of dollars, that agility is priceless. The scholars and practitioners are converging fast. Let's picture a simple batch out here. Research publications on one axis and industry implementations on the other axis. Research publications almost tripled from 2023 to 2024, and during the same timeframe, pilot project implementations nearly doubled. How has this happened? This is simply because simulation to software friction is falling, and there are three main reasons for that. Number one, foundation models trained are trained on massive engineering. Data sets are now available and accessible via APIs. Number two, industrial iot maturity finally provides the volume and velocity of data generative systems. Last but not least, cloud costs are sliding, constantly making per flow scale, training economically viable. The bottom line is sandbox phases over and we are in the mainstream adoption phase, and late movers risk. A structural disadvantages. The next question is what separates the leaders from the laggards? Step four. Typical behaviors. Number one, leisure focus on high value use cases. Leaders start with one pain point worth, multimillions, not a skunk works science project. Convergence of informants and technology team and operations team. In such scenario they get the CI and the plant engineers in the same room early. So MES or manufacturing execution system data meets cloud AI without endless middleware battles. Third objective and transparent KPIs like overall equipment deficiency, OE scrap and energy KPIs are published daily on shared dashboards. Turning an anecdote into accountability. Last but not least, playbook driven scaling. Once one line works, they replicate the same tech stack, same change management rituals across plants, regions, and business units. So what is the payoff? At least 35% faster time to market and 42% higher innovation hit rates versus peers sustained over a five year, five year horizon. So are these implementations smooth without any challenges? No. Let's not sugarcoat the hurdles. Then you will generally face three big challenges. At the top of the list is data quality and governance. Governance main gap will be missing sensors on your legacy equipment. There is another possible noise that is PLC. Time stamps are out of sync, so that data needs to be corrected. Ski semantics here is, there are 10 ways to spell the part IDs. Each company has different way. Each team may have different way of work, spelling or naming apart, so this needs to be solved with the modern data fabric layer and a single owner for data towards it. Second biggest to challenge is security and IP protection. Generative models may ingest CRL process knowhow. Without a strict access controls and federated learning options, you risk leaking trade secrets. Zero. Trust architecture and secure model incls are must have not a nice to have in any genetic AI implementations. Last but not least, algorithmic bias and model shift any historical data set Under underrepresents new recyclable design, the AI may prefer old plastics sabotaging sustainability targets continuous monitoring and periodic model code reviews. Keep the system honest. Companies that invest in these foundations upfront, consistently capture higher ROI and face fewer boardroom surprises. Next million dollar question is what is the return on investment and which is most important on any project implementation? Return on investment is not just impressive. It is exponential in this kind of implementation. In the first three months, companies invest in a strategy and system set up. By month four to five, they begin saying miserable process improvements. By month seven to eight, they typically reach some kind of breakeven, and by end of the first year, many companies report a hundred percent plus ROI without with continuous compounding, as the system learns and evolves. This is intelligent automation that pays for itself quickly and continuously. Now, let's talk about some implementation approaches, to achieve successful implementation, there are four phase playbook that can be implemented or utilized. A step one is our phase one is assessment process, which can be between two to four weeks or six weeks, depending on the nature of an implementation. In this process, map data sets rank pain points by value and feasibility and secure leadership sponsorship. Phase two is data preparation, which can be between six to eight weeks in this phase, clean level and integrate sensors manufacturing, SY execution systems, ERPO, that is enterprise source planning and supplier data into common. Data warehouse. Next phase is pilot deployment, which can last for eight last between eight to 12 weeks run. One contained use case for high high, highly high important business process. For example, generative is generative. AI is scheduling on a single packaging line. With crystal clear success criteria and measurable KPIs. Fourth phases is scaling, which is ongoing flow. The stack and new use cases are basically identified. Institute AIOps for continuous monitoring. This needs to be then needs to be put some guardrails, which are basically change management training, fortnightly stakeholders demos and rollback plan. You hopefully never use need to use, but in case you have to, that comes handy. Let's talk about the next steps. If you remember nothing else from this discussion or for some reason. Take these three action. Back to the team. Number one, conduct a readiness assessment in this assessment or audit data, talent, culture are to be included. The gaps are need to be identified and what are quick wins are hiding in plain sight that needs to be identified. Number two is basically select a lighthouse pilot. That is the key business process where it can be implemented, which is of high value and low political resistance. Let's think about packaging line, not the company's flagship process on day one, and then draft a 12 month roadmap that links milestones to financial KPIs and ESG targets. Share it widely and publicly on company dashboards which basically talks about transparency and that creates momentum. You start a smart major ruthlessly scale fast. That is the recipe for durable transformation implementation. I'd like to thank you for your attention and engagement during this discussion. We are at an infection point. Where factories are no longer constrained by yesterday's data, but powered by tomorrow's possibilities, those who act now will set the pace for the next era of industrial leadership. I look forward to your questions and to discussing how we can partner on your own generative AI journey. Thank you very much.