As global supply chains grow more complex, automated, and data-driven, the gap between theory and operational reality is widening. In this interview with Intelligent HQ, Juliet Mirambo, a rising leader in integrated supply chain operations, discusses how her background in engineering and chemistry shapes her approach to process optimization, forecasting, and manufacturing leadership.
Currently part of MilliporeSigma’s Operations Leadership Development Program, Juliet works across functions to strengthen global demand planning and material flow within a €250 million planning core model, while exploring how AI can improve decision-making in high-stakes environments. From resolving hidden data inconsistencies across ERP systems to designing reinforcement learning tools that accelerate error correction, she shares why supply chains rarely fail because of strategy alone, how automation should amplify rather than replace human judgment, and what future leaders must understand as operations become increasingly interconnected, volatile, and technologically advanced.
How did you get started in engineering and supply chain operations, and what led you to pursue a role at the intersection of process optimization, forecasting, and manufacturing leadership?
I realized that while traditional career paths in chemistry and engineering could address some of the problems I was interested in solving, they were too narrow to address them fully. As a chemist, I would spend my time in the lab optimizing reactions at the molecular level, and I would find myself frequently considering optimizations at a larger scale than just the molecular level—the large-scale systems that move materials, products, and information through global networks of operations. It was this that eventually led me to consider a career in supply chain management— not because I was leaving behind my technical education, but because I saw it as the area where the skills I have developed as a chemist become exponentially more valuable.
You are part of MilliporeSigma’s Operations Leadership Development Program, which is designed to build future leaders, what has rotating across functions taught you about how large scale supply chains actually operate versus how they are often discussed in theory?
The largest disparity between theoretical knowledge and practical application is that most textbooks assume all parties have access to the same level of information, whereas in real-world applications there may be multiple versions of the “truth”.
For example, in an operational setting, 3 ERP systems may disagree on whether you purchased cases, pallets, or individual units. Theoretical knowledge views data governance as an essential building block. Real-world application shows that data governance is often the bulk of the job. When I started my first role, I expected to optimize algorithmic processes and develop forecasting models. In my first 2 months of working at the company, I discovered that our “forecasting problem” was simply that SAP, Oracle, and our legacy system could not come to an agreement on fundamental facts regarding our ordering process.
Therefore, we were forecasting fictional data due to the numerous manual reconciliations that our data underwent, and thus it was no longer representative of our true business environment. This experience taught me that supply chains do not fail from poor strategic decision-making; they fail at the transitions that happen in between operations.
In your current role, you work on end-to-end process mapping and demand planning for configurable materials within a €250 million planning core model. What are the most common sources of friction or data gaps you see when organizations try to scale these models globally?
The largest contributor of friction is the type of invisible complexity that isn’t reflected on process maps, since everybody thinks it’s already solved. The Unit of Measure Harmonization is a perfect example. Every executive knows this issue is resolved. Why wouldn’t it be? We have enterprise resource planning (ERP) systems, master data management, and dedicated departments for Data Governance. How could we possibly not know if we’re purchasing items in cases or on pallets? However, in an environment where there are 16 separate locations; and at least three different ERP systems being utilized, each of which contains 20 + years of legacy configuration; and in excess of 1,000 SKUs are flowing through those systems – Unit of Measure becomes a blind spot.
When I developed my Unit of Measure Harmonization System, I measured its effectiveness by utilizing statistical analysis (Minitab) to document the change in performance over time. Not only did I prove the effectiveness of the system, I also was able to identify potential data issues before they became a disaster across our global supply chain.
Forecasting accuracy is a constant challenge in complex supply chains. How do you balance structured planning models with the reality of uncertainty, volatility, and human decision making?
I use the term “error correction velocity” to describe how quickly I am able to identify the difference between a forecasted outcome and an actual outcome and make adjustments to avoid cascading errors such as stockouts or overstocking. Because hundreds of individual forecasts are typically made each planning cycle, some will be accurate, but most will not. It is by identifying those incorrect forecasts in real-time, and making changes prior to them, resulting in significant issues, that the true value of my reinforcement learning system is realized. In addition to predicting demand, my reinforcement learning system monitors the gap between forecast and actual, diagnoses the root cause of the discrepancy, and executes corrective action 94% of the time.
For example, if a supplier consistently sends product in different unit quantities than what was ordered by our company, the system detects the error, makes the necessary correction, and prevents the discrepancy from cascading to other parts of the supply chain. The balance between human and machine decision-making is defined by confidence thresholds. For each issue, my system identifies six possible courses of action and assigns an expected value to each course of action: automatic correction based upon high confidence, automatic correction based upon medium confidence with associated logging, automatically initiate a request for information from the supplier, elevate the issue to a human for resolution, complete a group of similar issues in batch, or flag the issue for validation by a human.
Courses of action with high confidence values are executed automatically, while those with low confidence values are surfaced to humans with the context to make informed decisions. Structured models do not provide certainty; instead, they provide consistent decision-making across large volumes of issues. Prior to automation, whether an error was identified by a planner depended on the specific planner who encountered it and whether the planner had sufficient time to address it on the same day. With automation, planners now have sufficient time to focus on the 6% of decisions that require genuine human judgment, rather than the remaining 94% that follow well-defined patterns.
You are exploring how AI can improve demand planning accuracy and agility, where do you see AI adding real value today, and where do you think organizations need to be more cautious or realistic?
The true potential of AI is not in improving forecast accuracy but in helping organizations make better decisions when their forecasts fail. All organizations have complex forecasting systems already in place. Organizations do not have a lack of accuracy in their predictions, they have a lack of ability to make decisions quickly and consistently enough when reality does not match their plan. AI can identify patterns in historical data; however, supply chain operations occur in dynamic environments. What was learned by your model during relatively stable times? What happens to your model when there is a supplier disruption, geopolitical crisis or a demand shock?
The goal should be to create systems that utilize the strengths of AI for executing routine complexities at machine speed with perfect consistency and humans for addressing ambiguous situations that require business context, strategic judgement, and ethical considerations.
With your background in engineering and chemistry, how does technical training shape the way you approach supply chain problems compared to more traditional business
driven approaches?
Engineering training has one major advantage over training in the sciences — you learn to be skeptical of whether the problem you are trying to solve is indeed the problem. In engineering, you use scientific methods to systematically identify and trace symptoms back to their root cause. In chemistry, you design experiments to eliminate all other variables before attempting to test your hypothesis. The majority of approaches taken by businesses today begin by asking, “What is the solution that we can purchase?” Technical training asks “What is the problem that we are trying to solve?” This practical example of diagnosis illustrates this point. When the business forecasts do not accurately reflect future events, the first response is typically to seek out and purchase better forecasting tools. The technical training would first require you to map the data flow through the process (from raw data to finished product), and secondly, to ask yourself, “Are there inaccuracies in the data that we are inputting into our forecasting model?
Chemistry also teaches us that some chemical reactions are irreversible – once a reaction occurs, you cannot reverse it. Supply chain operations also share similar characteristics. Once you produce based on a forecast that proves incorrect, you have invested capital and used production capacity, and have produced inventory that may soon be obsolete. As such, the cost of being wrong is not always linear; it is often irreversible. Therefore, the way you consider and manage risk differs significantly from traditional business planning models. Engineering trains engineers to design for project constraints rather than ideal conditions. Business planning often assumes that you have unlimited capacity, unlimited information, and that people will behave rationally. Engineering asks: “Where are the bottlenecks? Where will this fail under stress? How will the supply chain respond to a missed shipment or a 30% spike in demand?” Experimentation is a very disciplined practice in the technical education community – unfortunately, it is relatively rare in business.
Outside your core role, you volunteer with MilliporeSigma’s SPARK™ program and lead experiments through the Curiosity Cube, why is early STEM exposure important to you, and how do you see industry playing a role in shaping the next generation of scientists and engineers?
My favorite part of working with students is seeing them discover that curiosity is a powerful force in the world. If a middle schooler says, “Why do we have to learn this?” — that is not resistance, that is a genuine question that should get a legitimate response. I had similar questions when I was in middle school. The best moments are when students make some unexpected connections. A few years ago, a student asked me why computers can’t prevent mistakes rather than correct them after they occur. That question from a kid who didn’t study computer science or supply chain led me to add preventative analytics to my system. The fresh perspectives young people bring to problem-solving are incredibly valuable because they have no idea “the way it’s always been” is the right way to solve a problem. There are a number of ways the industry can help students connect the dots between academic concepts and the application of those concepts in real life.
For example, students may understand chemical reactions, but they won’t know how understanding the characteristics of materials can be used to validate large amounts of data or how experimental design techniques can be used to test changes to processes .My hope is to provide possibilities rather than prescriptions to students. The technical knowledge students gain through formal education is transferable to many careers.
Looking ahead, what skills and mindsets do you believe future supply chain leaders will need most as operations become more data-driven, automated, and globally interconnected?
The single most important quality of the future leader is the combination of a curious mind and a skeptical eye – not cynically, but genuinely seeking answers. Future leaders should ask questions of an AI solution that claims to deliver “transformational” results to better understand: “How does this work?” “What have you tested this under?” “What would success look like six months from now?”
Future leaders must be able to think in terms of “and,” as opposed to “or.” They must be technical AND business-savvy, strategic AND hands-on, data-driven AND human-centric.
There is also something very powerful in remaining sufficiently close to the work being done. You are not a micromanager, but you have developed a deep understanding of what your team is building, the obstacles they face, and the trade-offs they make.
Communication will be more important than ever as systems grow increasingly complex. The ability to describe complex ideas in simple, accessible language will be more crucial than ever in the supply chain industry.
Leaders who will be successful in this era will view automation not as a replacement for human judgment, but as an amplifier of it. Automation will handle the volume and velocity of issues, allowing humans to focus on nuance, context, and ethics.
Finally, I believe that the leaders who will have the greatest impact will be the leaders who remain committed to their purpose. It’s all too easy to lose sight of why we do what we do as we optimize algorithms and chase efficiency metrics. The work that truly makes a difference is the work that reminds us why we are working on these problems. Maintaining that human element at the center of our thinking, even as we automate our systems, will set apart the leaders driving long-term positive change from the rest.
Peyman Khosravani is a seasoned expert in blockchain, digital transformation, and emerging technologies, with a strong focus on innovation in finance, business, and marketing. With a robust background in blockchain and decentralized finance (DeFi), Peyman has successfully guided global organizations in refining digital strategies and optimizing data-driven decision-making. His work emphasizes leveraging technology for societal impact, focusing on fairness, justice, and transparency. A passionate advocate for the transformative power of digital tools, Peyman’s expertise spans across helping startups and established businesses navigate digital landscapes, drive growth, and stay ahead of industry trends. His insights into analytics and communication empower companies to effectively connect with customers and harness data to fuel their success in an ever-evolving digital world.
