Name: AI in Manufacturing: Smarter Sourcing, Resilient Supply
Uploaded: 2026-01-29T16:16:08.829Z
Duration: 1 h 12 s
Description: AI in Manufacturing: Smarter Sourcing, Resilient Supply

Transcript for "AI in Manufacturing: Smarter Sourcing, Resilient Supply": Hello and welcome to this webinar on AI and Manufacturing in association with Intuit QuickBooks. I'm Jasmine Jessen, editor of Manufacturing Digital. Here, we'll be covering how new tools support smarter sourcing and resilient supply. There's been a lot of experimentation with AI and it clearly holds a lot of potential for manufacturers. But how can you actually integrate tools like Predictive and Agenetic AI in your operations? Well joining us today to answer this very question is Megalee Amiel, Global Industry Lead for Manufacturing and Director Corporate Services at CGI. Thank you very much for joining us today. Could you start us off by introducing yourself please, Megalee? Hi, Jasmine. First of all, many thanks for having me in today. I'm really pleased to talk about these interesting topics. So my name is Magali and I'm with CGI for six years now and in my new role since December. So yeah, the first podcast that I did, I do sorry, as a global industry lead, really pleased. So day to day, what I do is really talking with all my colleagues from around the world on manufacturing topics and of course AI is really at the core of the discussion and really pleased to exchange on this subject with you. Thank you ever so much. So I'll start off with our very first question today. When we're looking at forecast accuracy and planning capacity, what internal and external data points actually matter the most? Really interesting question and please let me introduce some elements of context before answering. So what we're seeing, manufacturing leaders are operating in a world defined by constant change. Global volatility and geopolitical uncertainty are reshaping markets and supply networks. Supply chain fragility and rising sustainability expectations redefine how products are designed, made and delivered. Cyber risks are intensifying a digital transformation exposed legacy OT system to threats that traditional IT services wasn't designed to contain. Also, talent shortage and retirement, that's a threat too because that erode institutional and operational knowledge and start progressing digital adoption. And finally, also what we are seeing that the digital leadership gap is widening as early adopters of AI and automation gain measurable advantages in efficiency, innovation and market agility. So it is quite a context, right, to talk about how it comes to forecast accuracy and playing capacity, which internal and external data points matter most. So on the top of that, we have to first, and it seems to be baked basics, right? You have to know your customers, because today they expect hyper personalized products, shorter later time and proof of sustainable practices. Listen to your employees. Why? They also expect the same intuitive digital experiences in the workplace that they enjoy as consumers. And finally, communicate with your partners and ecosystem because they expect trust, transparency and seamless data sharing. So, the biggest gains do not come from adding more data, but from combining the right signal. Forecast and planning accuracy improve when organizations stop chasing more data and focus on decision grade data. In Resume, values come from leading indicators, not lagging KPI. So that's why for the data we are seeing two major signals: your internal signals, so data coming from your demand, for example customer order patterns, order availability, point of self data about supply, your supply time, capacity, constraint and your network and your operational data, your asset availability, you are a manufacturer, can you keep the cadence or do you need more volume? So that's your internal signal and you have to combine with the external signal. We talk about geopolitical risk, war, tariff, weather, energy availability, all that have an impact also on your sourcing and on your supply and also logistics. Since 2020, I think the COVID hit us showing that the logistics is really important for manufacturing. So the biggest improvement in forecast and planning accuracy comes from combining this DICKMAN signal and constraint signal. Honestly, it's not about which data points matter most. The real value comes from connecting the signal across the ecosystem and not analyze them in isolation. So, you know, within Horse CGI Unified Manufacturing approach, we stress the importance of integration. It's not just about connecting machines or distributors or suppliers, it's about bringing people, processes, data and systems all together into a single resilient operating model. The WaveCGI experience shows that organizations gain the most accuracy when data is shared securely across an ecosystem of partners. And if I have to resume this answer, maybe it's not about having every data point, it's about connecting the right signal across the ecosystem and with AI, turn them into early warnings rather than after the fact explanation. Thank you. So my very next question is: when we do see disruption hitting, how can we use scenario planning to find the right balance between service levels and working capital? Well, in line with my previous answer, a disruption hits, if you already have connected the right internal and external signal, you may have early warnings enabled to generate different scenarios. What we're seeing is where traditional planning optimizes one objective at a time, a scenario playing with AI allows explicit trade off. So leaders who already have AI enabled scenarios are able to simulate trade off in near real time, so they get the information and they are able to quickly pivot and take decision. Really important, they understand the impact of decision across all services, cost, risk, cost and risk, sorry. So they can look at if disruption hits, they can look at service level versus inventory exposure, speed versus cost, risk mitigation versus cost preservation. So they run multiple tests and they take decision according to the framework, the boundaries that they put into the policy. For example, you are on manufacturing, you have to decide what is the minimum service level that I want for my plant or my factory? What is the maximum working capital exposure I can take? What is the acceptable risk threshold? What So we are seeing, when in your scenario it's really clear in your governance and your policy and with AI you have this power to really have scenario and take decision real time. Really important, AI does not replace decision makers. It surfaces the consequences of choice clearly and early. So the shift that we are seeing that the conversation from what is the right plan to which trade off are willing to make. So what we are seeing with the industry and it's really interesting, right? Interesting time That AI allows teams to quickly model the impact of disruption services, inventory and cash, compare alternatives and make decisions based on risk appetite of the organization rather than intuition. This is especially powerful in crisis time when speed and alignment are more important than precision. So what we see in conclusion is that resilience is the new ROI. So accepting that forecasting is no longer a single number, but a set of scenarios. It's the ability to absorb disruption with confidence. Brilliant, thank you. I wanted to ask next about sourcing. Where do we see AI deliver the fastest results? Sourcing is becoming quite an exercise in augmenting economic intelligence. Today we have access to many information, buyers and sources have a lot of information and you know when you have too much information you kill the information in the end. So AI delivers the fastest results in procurement where decisions are frequent, rule based and where structured data already exist. So especially in intake to pay, compliance and risk census, that's where we'll see the fastest results and then move upstream into strategic sourcing. So what AI allows for organizations is really to move beyond historical suppliers because they have access to faster indication of alternative suppliers, early detection of risk, financial, geopolitical risk, climate, if you have an impact and also on the negotiation. AI helps because with all the information you have, you focus more on cost drivers, capacity, resilience rather than price alone. So it's a big shift, right, for the smarter sourcing. Well, the areas, you have many areas where you have results, but the first one is really intake to pay automation and it's true for manufacturing but also in other industries, it's really cross intake to pay automation. That's where when you get results from AI, you have the trust and you have the credibility to expand in other parts of the organization. What So make it take to pay automation? AI reduces cycle times, errors and manual efforts. So as you have high transaction volume of this kind of job, clear rules and policy, it's really, really easy, we think it's easy but that's where you can have gain really easily. So some examples of what AI does on intake to pay automation. So you have auto classification of requests, for example, you know, it's time consuming for people when you operate plant and when you have a network of plants to operate, so with AI you accelerate that. Also policy guidance before submission, so AI can easily put all the information together and send to a human to validate. And finally, also invoice anomaly detection. You know, when you have to deal with invoice, sometimes that could have an impact on the cash flow and on your cash. So that's the kind of example where we have fastest results. And really interesting, we had some conversation and one of the clients say to us, if AI cannot simplify intake to play, it will struggle to earn credibility elsewhere. Absolutely. So that's your first. That's another one. Then also on spend classification and visibility, we see that AI accelerates and same that for intake to pay. It's because it's structured and you have information. What's interesting with spend classification and visibility that helps sourcing become fact based and more proactive than reactive to situations. Another example maybe where AI delivered the fastest result, supply risk monitoring and early warning because AI works twenty four hours. So, continuous scanning of supplier and sub tiers have the ability, the detection of the weak signal, for example, financial stress, geopolitical exposure, lead time variability, so bringing this result, so AI identify for the human where to look and the human decides what to do with this information. So maybe let me another I have many examples on this one. One interesting is also supplier discovery in constraint category. When you work on manufacturing, depending on the kind of product you made, you have some constraints. So, AI really helps to direct material with supply concentration, region with geopolitical and capacity risk and look at what is innovation driven sourcing. So provide information that rapid scan the long tail supplier, the capability matching between the certification you need, the capacity of your sourcing and the geography and your exposure. So bringing all the information and to make faster decisions. So that's the fastest AI win in procurement operation, as I say, intake to play, compliance, anomaly detection, because they immediately reduce friction into the organization. For supplier discovery, it delivers strong value as well, but typically you need to have data quality and governance in place before running this kind of project. And I want to finish with one quote that our leader at CGI AI, my colleagues Diane Gottig say to executives, you should think beyond fast. The real leadership question is not where AI is fastest, but how value compounds over time. What we see on CGI twenty twenty five voice of a client, the voice of the client is the exercise that we do annually with the C level across the world to get the trends and what they are doing in their business. And our research results for 2025 show AI maturity is accelerating. 35% of organizations now implementing traditional AI, up to 9% from twenty twenty four, twenty six percent implementing Gen AI, up to 13% year over year. So, and the organization that see ROI do three things early: define intended business outcome upfront, design measurement into the use case, not after deployment but before, really important and ensure value is realized in production or quickly adjust if not. AI doesn't replace buyer judgment, it protects it by focusing attention when it matters. Thank you. So my next question, obviously you've talked a little bit about how AI can kind of get us risk and ESG monitoring. How can you combine those with the expertise of your buyers? You know, that's a when I see this question, when I saw this question, sorry, I see. Wow. That's a really good question. Where do I start? Because you have different angle, right, where you can look at it. But we agree that you are core principle when you look at this question. First, buyers contextualize, policy governs and AI recommends. So blended works only if each element has a clear role and formal authority in the decision flow. So the key point to it is that's really AI should be treated as a decision support layer, not an automated gatekeeper. It's really important to separate and having the protection around to be sure when we blend all the information, we have something really strong. So first, for example, AI risk and ESG, AI helps to detect, score, predict and stimulate. Then you have the bio judgment where you interpret, prioritize and you override with context because sometimes AI doesn't have all the context, right? And you have the policy, really important, set boundaries, threshold and escalation rules. All that, the objective for all that is to avoid black box decision and keep accountability human, really important. When you have this that in place, what we see is we look at use multi dimensional indicators. So, you know, as I say previously, my answer, how you combine the information internal, external, that's the same. So you look at risk probability, risk impact, time horizon. And what we are seeing with the best performing organization, they really use AI to continuously score suppliers on risk and energy performance, but already at the beginning, they embed internal policy directly into AI models and constraints and thresholds. So meaning that you have all the information and at the end you keep also the buyer in the loop for exceptional strategic decision. And you already have, and I want to highlight two initiatives that are existing and are really interesting to follow because sometimes we don't have to reinvent the wheel, right? If something is working, we have to look at the initiative and see how it can fit with the organization. So you can take, for example, the Catina X initiative. Catina X is a collaborative, open and federated data ecosystem designed to enable trusted data sharing across industrial value chain. Catenex enables concretely end to end supply chain visibility, ESG and carbon footprint traceability, information about risk propagation and compliance reporting. So that's an industry, they work together on this blended information across an industry. The other one was really interesting also, that's the digital product passport. It's a digital record attached to a physical product that contains standardized information and lifecycle wide information about the product. So it follows you from the design to the end of life and can be accessed by different stakeholders and with the DPP for example for buyers and procurement you have what we call a supplier transparency, we can get all the information and you can take this information and you know combining with your own indicators and you have results and in the end, really important, your buyers, your humans take the decision. So what is interesting on this question is that AI brings consistency and scale, human brings context and accountability and resilience, depending on both. Thank you ever so much. So if we're looking at a manufacturer who's looking to kind of implement AI for the first time, what are those first use cases to kind of go with and how can you set the right metrics for success to make sure you're hitting the right boxes? Question two: back to basics solve a real pain point, have measurable business impact and fit into existing workflow. What we see with AI is sometimes a really interesting project, people think outside the box and how AI can bring and they want to start some use case but you know when you are in manufacturing and other industry, you know, you have accountability on your financial results, right? So really, when you want to pick up an initial use case and especially if you look at smart sourcing or resilient supply, you have to look at three things. First, when do you have rich and accessible data across your supplier, for example? Does your decision that you have to make are repeatable, high frequency and time crucial? You know, it's really because that's where you can have some results really easily and also the outcomes. Really important to say, okay, we started a new use case, but what I'm expecting and what is going to bring to my organization and you have to link the outcome to your business KPI such as total cost, service level, lead time and reliability, really important. When we look at the successful use case in AI coming from, you know, use case to operational industrialization, classically adopters in smart sourcing and resilient supply chains start with supplier discovery, segmentation and qualification, demand forecasting and supply demand matching, lead time prediction and variability reduction, risk sensing and disruption anticipation and inventory optimization across multi tier networks. So if you are in the manufacturing area and you have not yet started this kind of project, you can pick one of them. Then define success metrics before you start. Non negotiable I think that you can choose a project but non negotiable. What will be your success metric? So you have to look at the metric directly tied to what the business value most for you. So for example, if you are on when you look at demand forecasting and supply demand matching, sometimes how you reduce supply disruption frequency, also how you can improve your supplier in full performance, how you can shorter and having more predictable lead time, so you look at what is important for you and you set the metrics. Also important, before the deployment, you have to align on baseline supply and sourcing performance. It's not jazzy, right, what I say, but it's really important if you want success in the end when you implement AI. Target improvement on acceptable trade offs and the time horizon over which value is expected to materialize. The time horizon is important because we think that with AI it's, you know, you put AI model in your data and you have some results, it's not working like that and sometimes we have to set up that before the test and you know clear metrics ensure AI initiatives are measured on real supply outcomes not technical sophistication because sometimes, yes, you implement something, it's really technical and provide information but the outcome is not there so that's why it's really important. So you picked up one of the case, you define your metrics, you align before outcome, before going to the outcome, to financial and operational value. So the leaders today are not asking, should we use AI? It's not the question. Here they are asking, how does AI protect revenue, margin and continuity? So what are the executives are expecting from AI? Reduce dependence on manual expediting and firefighting, increase resilience to supplier logistics and demand stock, improve consistency and speed of sourcing and allocation decisions, enable proactive rather than reactive supply chain. So this expectation should directly shape use case prioritization ensuring each initiative has a clearly understood value proposition. So what we can say to that? So when you have the outcome and your operational value attached, you have an early win and then when you have your win, you go to the bigger resilient journey. So for that, for example, when you have digitized all your sourcing, so you have a smart sourcing, so you can extend AI, for example, from local sourcing decision to cross category, cross region and multi tier networks. So that's why it's really important at the beginning, having a really short and controlled use case and you test it and when you get it, you can, you know, fly to the moon with, right, you can expand. So when you when if I resume all this answer, I like to resume in one sentence, it's that pilots scale when users see value in their daily work, not when the model is technically impressive. So all you do with the fastest result is because the people on the work floor, at the office, they see value on their daily work and help them to concentrate on other tasks. And for that, I just want to mention that my colleagues Marcel Moritz, who is one of your experts on manufacturing too brought on blog, who is available on steered.com, really interesting about the top two strategies to secure investment for AI powered asset optimization in manufacturing. With a lot of examples, what we will do with a client and how also to follow the different steps to identify the fastest AI use case. So if we're looking at starting to embed these AI tools across planning and sourcing and workflows, how can we make sure that they are embedded, but also that they have space to scale? That's another good you have a lot of good questions, Jasmine. That's great! Thank you very much, I'll do my best! We define success in business outcomes. So for us, when you talk with the client, it's also what is the business outcomes, not the model performance. Interesting, mean the model performance and we have people working on really interesting models, but what's the outcome attached to? So for example, accuracy matters, but adoption matters more how we adapt inside the organization. So the key shift is from AI as an insight to AI as a decision support, finally to AI as execution trigger, right? What we see. So if I translate that in function from sourcing and supply, so for example for planning, when you have to plan your operation, you embed AI into your S and OP, your cycle, your MRP runs, your scenario planning. So that's the first point of entry where AI is an insight and started to support the decision and how you scale it. So AI recommendation become inputs, not optional reports. Really important. AI is not a PPT or is not a really nice slide that you present, right? So AI become by default inputs, not operational report on your planning. Now look at the sourcing, for example, so you embed AI onto your RFx, so RFP, F5, FQ, the selection of your contractors, the contracting allocation also, so do you embed AI into this workflow and how you scale it? That AI narrows choices, so a flag risk, a suggest action, so really you take the information from all the information that you put in your FX, look at all suppliers and identify some recommendations. And so planning, sourcing and I go to the shop floor there, so you embed AI on the scheduling, no matter on release, the exception, the handling because that's a loop coming to the planning. AI anticipates constraint before the disrupt execution. It's a very important factor in manufacturing when you have the supply chain is very tight and you have the impact of social political or impact of what's happening, for example, the weather in Canada for this weekend will have an impact on transportation because it's going to be really cold. And that's where AI anticipates and how to take decision. So maybe I can explain a concrete example and I take supplier discovery, segmentation and qualification because it's so on sourcing is really important. So where to embed AI? So first on your procurement platform, CRM sourcing suites, on your RFQ, RFI, RFP and supplier onboarding workflow. So you embed your AI, so you say I want AI to help me in this process and how do you scale it? So AI is going to scan internal performance data, external signal, is going to scan also supplier, dynamically and resegmented. So AI is going to say, okay, your supplier is resilient or strategically you have a new supplier that you can look at. They are going to look at the constraint, the risk and also because about sourcing, it's always about risk management. And then also AI is going to help the qualifications check to become more and more important. That's we call them the new when you have a new risk, you have a requalification trigger. So that's meaning if something happened to one of your customers, AI can easily send the information to you and you can make the decision to change AutoPay bot. So the workflow impact, buyers no longer search manually, so meaning that you have more time. When you open your laptop, for example, in the morning, AI screen all the information and trigger you, okay, your attention must be concentrated on that, that and that. That's where that helps the buyer in their daily routine, but also provide them information to be more proactive on decision, on tactic and strategical decision. And really important, governance rules define when humans must approve or override. When we talk to AI, you see it's more and more important all around the governance. So when you have this governance in place, you have faster sourcing cycle and you have more resilience into your overall strategy and so that's really important. And maybe three conditions that are really important to embed. First, AI is part of the workflow, not a dashboard. Second, clear decision ownership, human own judgment and expectation. AI owns speed, scale and consistency. That's to be very clear. And then you close the loop by learning because every decision outcome feeds back into the model so your AI is more and more proactive but the governance defines when models are trusted between or constraint. So AI scale in planning, sourcing and operation when it is embedded into how decisions are made, govern and execute, turning insight into action and action into learning. So you've touched on kind of governance and oversight there, but really my final big question for you is where do humans sit within these processes? Oh, wow. Yes, let me start with a remark about, from our Voice of the Client about governance because it's really the core subject. One of the biggest misconceptions we see is that governance and speed are intention, but in our Voice of Client 2027, the research shows the opposite. Organizations with holistic AI strategy perform six times higher than Gen AI maturity. Those with robust data strategy have five times higher maturity and in practice governance is that allow AI to scale safely and confidently into production. So when you start with a clear governance and ownership, AI must be governed like any core operational capabilities. Executives set the outcome, cost, service, resilience, risk. Business leaders hold the decision, AI support, for example, in planning, sourcing operation and you have one accountable AI product owner per use case that you answer the value delivery into the organization. So, for example, CGI responsible use of AI framework embed governance across the entire lifecycle. We envision, so meaning that define the human AI future aligned to values and outcomes, Explore, select ROI led use cases under clear and guard frame. Engineer, so embed governance into operating models, data pipeline and security and expand with scale responsibility with continuous monitoring and improvement. So the key message that governance by design accelerates deployment because it's about rework, risk and loss of confidence later and system integrators help to define decision rights, the RACI and operating cadence, so AI does not come as an IT only initiative. You have to make security and compliance non negotiable into your governance. So the trust starts with enterprise grade control, so you have the role based access to data and AI recommendation, all the full traceability of the data, the security of the model deployment, really important. You know, the system information ensures AI is embedded safely with your different system and planning platform. Also, you have to treat model as managed assets, not black boxes. In AI sometimes, people put AI on the oh I'm going to do AI on this part and we see what we have on the results. No, AI is not a black box, it's a model and you have to monitor accuracy, the BS, the drift continuously. You have to provide clear business explanation beyond recommendation and a clear control, versioning and rollback because in the end you want your AI that is predictable, auditable and trusted. Now we have to define human in the loop decision checkpoint, right? Because it's important, AI is there to support the human, to empower decisions, to empower what we can do. So not all the decisions should be automated. Well, low risk decision, you can have automation but with human override is needed. For material decision, AI recommends and humans approve. And for strategic and reputational decision, AI informs and human decides. So these checkpoints are embedded directly into planning, sourcing and execution workflow and into your governance and when you have that, you are ready to go. So really important also to say that the goal is not to replace people, it's to treat them to focus on decisions that truly matter, on a subject that truly matters for the organization and the growth of the organization. AI scale, when it's governed like a mission critical operation, secure, explainable, accountable, and always anchored to human judgment and business outcomes. Thank you ever so much, Megalie. Now if you're to leave the audience today with one final piece of advice, what do you think that would be? It's really difficult just to one advice, right, in the end, but in conclusion, I think the next decade will reward manufacturers who treat AI, data and emerging technology as levers for reinvention, not incremental improvement. Those who unify their foundation, scale intelligence, responsibility and prepare for what's come next will create stronger customer value, more resilient ecosystem and sustainable growth. So at CGI, we stand with our manufacturing clients to design the future of business and it's not one but three takeaways from the conversation. It's not about having every data point, it's about connecting the right signals across the ecosystem and the AI turn them into early warnings. Use AI to bring consistency and scale where human brings context and accountability and resilience depends on both. So that will be a good one. And in the end, one is important, you can build AI model quickly, but you can't scale trust without governance. So maybe data, human governance for the future. Thank you ever so much. Well, that's all we had time for today. Thank you ever so much, Magaly. It's been a pleasure to speak with and thank you to you all for watching. If you've got any questions, do feel free to pop a message in the chat and we'll send them all to Magaly so she can get back to you. If you've enjoyed this session, please do check out our webinar section on manufacturingdigital.com. I hope you've learned as much as I have today and we'll see you next time.