I recently attended part of Optilogic’s OptiCon conference, where I sat in on several standout sessions from speakers across the supply chain and logistics space. From an inside look at how 3M is rethinking its global logistics strategy to Amazon’s approach to last-mile delivery in Brazil and Castrol’s journey building internal supply chain optimization capabilities, the sessions offered a wide-ranging view of how leading companies are using data and design to navigate an increasingly complex operating environment.
Keynote By President and CEO Don Hicks
Optilogic President and CEO Don Hicks opened with a simple but pointed reminder of why the company exists. “We’re not faking it when we say we love problem-solving. We really mean it,” he told attendees before pushing them to reconsider what actually drives their careers – time versus money, stability versus purpose – and to ask whether meaningful work matters more than a paycheck. That instinct to ask “why,” he argued, is the same instinct that drives good design, whether applied to a life decision or a supply chain strategy: “Are we accepting things the way they are, or should we push to make it better?”
Hicks pointed out that Optilogic’s mission is based on a balanced commitment to three stakeholders: customers, employees, and investors, describing it as a form of stakeholder capitalism rather than pure customer obsession or shareholder-first thinking. He thanked customers who “took a risk” on a young company with a bold new product, credited the Optilogic team’s internal culture of service as the foundation for serving customers well, and acknowledged investors who backed unconventional ideas, saying the company works to make sure “they don’t regret that evening they wrote the check.” On culture more broadly, he argued it deserves more attention than strategy typically gets: “Amateurs talk about strategy. Professionals talk about culture.” Strong internal culture, in his telling, means debating ideas rather than people, rejecting misleading tactics, owning mistakes, and sharing credit generously.
A recurring theme was design itself, not as a buzzword, but as a discipline of stepping back to see the whole system rather than just fighting daily fires. He pointed to the gap between incremental optimization and real breakthroughs as evidence of design’s power, noting that a model might yield only a 0.5% improvement on its own, but adding new lanes or products to the analysis can push that to 10% or 15%. He described Optilogic’s strategy as operating at three levels: empowering individual modelers with tools they can run independently, enabling teams so that design becomes a business function rather than a single person’s responsibility, and eventually optimizing at the enterprise level so that non-specialists such as a warehouse manager can tap into the same underlying logic and power that expert modelers use.
Hicks also traced the company’s product philosophy back to its foundation in Atlas, the cloud-native platform Optilogic built as a general-purpose coding environment before realizing its fit for supply chain design. “We didn’t set out to build a supply chain app,” he said.
“We built a coding platform. It just turned out to be the right solution all along.” From that foundation, he highlighted several other developments: Supernova, described as the company’s most powerful solver, capable of running large-scale supply chain models such as those used by General Motors; the Leapfrog AI assistant, which he said is speeding up workflows today and will expand into a team-oriented tool; and DataStar, a data transformation and analytics product aimed at cutting down the roughly 80% of modeling project time typically spent wrangling data.
He also described new professional services, an app development team, and a customer engagement team the company is building out framed not as a pivot into consulting, but as support for what he called an ecosystem: “This isn’t consulting. It supports our ecosystem. We want to be the platform everyone builds on.”
Looking ahead, Hicks named three challenges he sees facing the industry: “short-termism,” or the tendency to stay reactive rather than invest in long-term design; a cynical tech culture driven by vendor hype that he urged the audience to hold both Optilogic and competitors accountable for; and what he called “AI panic,” arguing that AI should be understood as a tool rather than a replacement for human design judgment, since it maps inputs to outputs but “can’t see the future if the past doesn’t reflect it.” He closed with a personal story about climbing Kilimanjaro with his son, using the experience, pushing through altitude sickness to finish together, as a metaphor for the design process itself: signing up for one reason, discovering how hard it actually is, and only later understanding what it meant. He left the audience with a final challenge: “You ought to be proud of everything you do. If you’re not, change it.”
VP of Innovation & New Products Vikram Srinivasan and VP of Product And Solution Design Rebecca Janowiak
One of the sessions was about designing the supply chains that you need as part of a discussion by VP of Innovation & New Products Vikram Srinivasan and VP of Product and Solution Design Rebecca Janowiak. During this session, they discussed how organizations can move from high-level risk metrics to a more granular understanding of where that risk actually sits.
Vikram Srinivasan
Vikram noted that overall data might show a modest deviation — around 1.8% that looks manageable in aggregate, but cautioned that this top-line number can obscure meaningful variation underneath it. The point was that risk isn’t distributed evenly: it can concentrate in specific regions, client relationships, or parts of an organization that aren’t visible when only looking at summary figures. Understanding those underlying relationships is essential to knowing where an organization is actually exposed and where it needs to focus its attention.
Rebecca Janowiak
The presentation then moved into a demonstration of Optilogic’s modeling workflow, walking through how a model gets built from raw data through to a validated, production-ready state. A central feature highlighted was something called SCAP, described as a conversational AI capability that “answers before acting,” meaning it surfaces its reasoning or proposed steps before executing changes, with a human retaining oversight throughout. As users become more comfortable with the underlying data quality and complexity, the presenter noted, they can choose to grant the system more autonomy over time rather than reviewing every step.
The demo also emphasized the platform’s approach to reusing and referencing existing models: rather than rebuilding a model from scratch, users can reference a previous model’s structure and content as a starting point for a new one, carrying over validated structure while adjusting for new data. They also walked through a real example of this process running into trouble, data reloading and warehousing issues that accumulated over time and ultimately caused a model build to fail, and showed how the system responded not just by flagging that something broke, but by explaining what was being fixed and why, surfacing specific usability issues it had identified along the way (including inconsistencies between labels and reference tables, and a debugging example where an output value was traced back to a zero-valued input).
Toward the end of the session, they also introduced a feature for generating baseline projections across key business metrics: cost, revenue, profit, demand, and risk score, describing how the system creates a distribution of possible outcomes around a stated baseline, giving users a clearer picture of where their targets are likely to land rather than a single point estimate.
Felipe Luz, Senior Manager – Global Logistics Strategy & Transformation at 3M
Felipe Luz, Senior Manager – Global Logistics Strategy & Transformation at 3M, opened by outlining the scale of the challenge his team is working against: tariffs, geopolitical volatility, and a fragmented supply chain in which each region and operation runs on different systems, making it difficult to unify data at a global level. He noted that cybersecurity risk extends beyond 3M’s own systems to its partners, since an issue at any single partner can ripple through the broader logistics network, and that natural disruptions require the organization to move quickly when conditions change. Layered on top of that is 3M’s own scale and complexity, now organized around three business groups following last year’s restructuring, but still spanning roughly 2 million customers, 13,000 suppliers, about 157,000 annual air shipments, some 200 owned and third-party facilities, more than 160 countries, and over 200,000 SKUs across roughly 15 regions.
Given that complexity, Luz said the team deliberately chose to start small: build something that delivers short-term value while also laying long-term capability, pilot it, and then scale globally rather than attempting a global rollout from day one. That led 3M to focus first on its two largest export flows: the United States, which accounts for roughly 50% of international shipment volume, and Europe, at about 20% — together representing 70% of the company’s export volume. The stated goals behind the effort are to maximize customer experience through better inventory positioning, faster and more resilient shipping, and stronger cost efficiency, pursued across three levels: a strategic layer evaluating whether consolidation centers and distribution sites are located optimally and whether the network can absorb new product introductions; a process layer built around close collaboration with the business to identify bottlenecks and prioritize a pipeline of modeling projects, informed by a “digital twin” of the network; and a technical layer focused on combining disparate ERP and transportation management data sources into usable models, while grounding the results in feasibility and stakeholder buy-in — Luz cited an earlier experience where a model recommended a distribution center location that, on the ground, turned out to be impractical, underscoring why feasibility review and gate reviews follow the modeling stage rather than replacing it.
On results, Luz pointed to a case study involving 3M’s three U.S. consolidation centers (in the Chicago area, the East Coast, and the West Coast) and two in Belgium, where the team is reassessing when goods should route through consolidation versus ship directly from manufacturing or distribution sites when volume allows. He cited expected productivity gains of 5% to 10% this year, a 1- to 3-day average reduction in global ocean transit time, a targeted 15 percent increase in direct shipping, and roughly a 50% reduction in consolidation costs, partially offset by higher inland and ocean freight costs.
Luz closed with several lessons from the process so far: internal alignment with leadership and stakeholders is essential, since top-down decisions alone don’t guarantee execution or cross-functional buy-in; choosing the right implementation partner matters, and he pointed to 3M’s collaboration with Optilogic as an example; the underlying data will never be perfect, so the model requires trusted assumptions rather than a purely automated process; and echoing a point raised earlier that morning — technology should support strategy rather than become the strategy itself, meaning the network modeling capability needs to serve 3M’s broader multi-year logistics vision rather than dictate it. He framed 3M’s effort as still in its early stages, focused not on one-off redesigns but on building a durable, long-term modeling capability that can scale with the business over time.
Joris Wijpkema, EVP of Solutions and Strategy at Optilogic
Joris Wijpkema, a former McKinsey consultant, now with Optilogic, opened by acknowledging that the pace of change at this year’s Opticon can feel overwhelming. He noted that just a year ago, the platform centered primarily on Cosmic Frog supported by LeapFrog AI, a much simpler form of AI at the time.
In the seven months since DataStar’s general release in November, the platform has added custom apps and, most recently, ADAM, which he described as fundamentally changing how users interact with the platform and what’s possible in terms of speed and insight extraction. He framed this shift as moving supply chain work from periodic, specialist-only exercises to continuous, more broadly accessible processes, no longer confined to design teams and modelers, but increasingly connected to planning and open to people across finance and other parts of the organization. What used to require manual data work can now run through automated pipelines, and what used to be one-time projects can become always-on solutions, enabling faster turnaround on strategic questions, augmented planning processes, and continuous opportunity identification.
Drawing on his own background, roughly two-thirds of his work at McKinsey focused on planning, he argued that planning today remains surprisingly limited despite decades of progress, citing an executive who said their company was six years into an advanced planning system implementation and still only 7 percent complete, without even the ability to run quick scenario evaluations for tactical planning. He walked through several gaps he sees in most planning systems: difficulty running more than a handful of scenarios quickly, let alone extracting insights from dozens or hundreds of them; limited visibility into both predictable and unpredictable variability, versus Optilogic’s ability to run true event simulations and actual orders through a model; a tendency to produce merely feasible plans rather than optimized ones; rigid schemas that don’t reflect the real nuances of individual supply chains; and a lack of integrated network optimization and routing, which he said Optilogic combines in one system.
He was careful to position Optilogic as complementary to, not a replacement for, advanced planning systems and ERPs — those remain the execution-focused system of record for order and coordination detail, while Optilogic’s stack (DataStar, Cosmic Frog, custom apps, and Ada), connected via high-speed integration to those systems, provides the live digital twin, unconstrained optimization power, and scenario-based insight that planning systems can’t offer natively. He pointed to existing starting-point apps for demand modeling, production planning, and executive insights as early building blocks toward a full sales and operations planning or integrated planning process, spanning demand and supply evaluation through executive alignment.
He closed by turning to the organizational side of adoption — which are questions he said come up constantly, including whether teams have the right capabilities to use the new technology, how to bridge historically separate design and planning teams, and what data infrastructure and governance need to look like to support these tools at enterprise scale. He introduced Optilogic’s solutions team as the resource built to help answer those questions, describing its offerings across solution design, training, ongoing coaching, professional services (including building models and implementing solutions directly), and general partnership on business strategy.
He emphasized the team’s deep supply chain experience and outlined specific engagement types, including accelerators (quick model builds, solution design, or initial data infrastructure setup), full solution builds, help establishing centers of excellence, talent-sourcing support, and ongoing center-of-excellence augmentation. He closed with a set of stated commitments — building people-centric solutions, operating with transparency about the technology’s real capabilities and limitations rather than jargon or hype, and tying Optilogic’s success directly to the customer’s — and offered a free opportunity assessment and solution workshop to attendees interested in exploring next steps.
Felipe Moraes, Executive Director of Supply Chain and Integration at Amazon
Felipe Moraes, Executive Director of Supply Chain and Integration at Amazon, opened by describing the scale and pace of growth in Amazon’s Brazil logistics network, before turning to the operational challenges that come with serving a country as geographically and infrastructurally diverse as Brazil. He described the difficulty of reaching remote regions in the North and Northeast, where paved roads are limited and deliveries often depend on a mix of transportation modes, including river and ferry routes.
He pointed to Amazon’s first river delivery in Brazil’s North region as a particularly meaningful milestone, both for the emotional response from customers receiving packages in places previously inaccessible to standard delivery and for what it represented economically for those communities. He also noted the added complexity of designing for fast delivery — in some cases 15 to 30 minutes — across such varied terrain, tax jurisdictions, and transportation modes, and highlighted a long-running partnership with local communities that allowed Amazon to build delivery capability in areas it says only Amazon currently serves in Brazil, while also creating new work opportunities for people in those communities as delivery partners.
Moraes then walked through the evolution of Amazon’s Brazil operations over time: starting in 2013 to 2015 with a network built primarily around fulfillment infrastructure; growing between 2015 and roughly 2021 through third-party delivery partners as the company built out the systems needed to support them; launching new delivery capabilities between 2023 and 2025; and, in 2025, introducing what he called the “Brazil 2.0” program to convert much of the network toward same-day and ultra-fast delivery. That shift required rethinking how the network is replenished and restructured — including the use of smaller urban facilities such as dark stores in areas inaccessible to larger delivery trucks — as well as renegotiating terms with selling partners and vendors to reduce friction in the network. He also described a partnership aimed at accelerating quick-commerce operations in both Brazil and Mexico, building on channel integration and a new inventory management model developed with vendors.
Among the constraints Moraes cited going forward: a shortage of available real estate suited to dark-store operations in dense urban areas, which often requires renegotiating with property owners and pushing for regulatory changes since these spaces weren’t originally designed for high-frequency small-scale logistics; and a broader shortage of available drivers, which he described as a challenge facing not just Brazil but Amazon’s operations globally. He noted that lessons from more mature markets like the U.S., Europe, and Japan don’t always translate directly to Brazil, and that recognizing this gap led him to bring dedicated supply chain network design capability in-house, focused specifically on optimizing the network as a whole rather than optimizing individual functions in isolation — an approach he said allowed Brazil to move faster on delivery speed than some more established markets, and one that also surfaced opportunities elsewhere, including addressing underutilized capacity in the U.S. network.
On the technical side, Moraes described data structure as one of the central challenges in network design work, noting that poorly organized data leads to failure or unnecessary cost. He described using tools including Cosmic Frog and Vista to build out more robust network design capability, and outlined a broader shift in Amazon’s operating model — moving from a national-level system, to a regional one, to, this year, a localized distribution model intended to better support fast delivery while reducing costs. He said the team has also moved from a reactive, simulation-based design approach toward a more predictive one, using accumulated data and scenario modeling to anticipate future operational needs rather than only reacting to current conditions, with an ongoing effort to integrate data pipelines more fully and continue exploring additional AI-driven initiatives.
Daniel Davis, Global Supply Chain Optimization Manager at Castrol
Daniel Davis, Global Supply Chain Optimization Manager at Castrol, opened with a lighthearted aside about the AI assistant “Ada” sharing his daughter’s name, before turning to Castrol’s supply chain transformation. He described the company’s historical approach as reliant on outsourced modeling capability with high technical standards but limited internal ownership, which meant that as models fell out of date, they were rarely refreshed to reflect changing demand or costs.
That left insights becoming obsolete quickly and prevented the team from making proactive decisions rather than reactive ones — a point he illustrated with a Tony Robbins quote about leaders anticipating and losers reacting, noting that Castrol had historically fallen into the latter category. Three years ago, Davis said he made the case to Castrol’s supply chain leadership to invest in building internal optimization capability rather than continuing to rely on external partners, and 18 months ago the team began its work with Optilogic.
He gave some background on Castrol as a global manufacturer of premium lubricants for the automotive, industrial, energy, and marine sectors, operating 22 manufacturing plants along with joint ventures, hundreds of warehouses, roughly 80,000 SKUs, and about 2 billion liters of product moving through a supply chain that runs from raw material sourcing through blending and bottling plants to a distribution network delivering to customers either through packaged goods or bulk distribution directly into customer tanks. His global optimization team includes two modelers — a senior modeler and a junior modeler — supported by regional network managers who, while less technical today, represent an opportunity for broader skill-building across the organization.
Davis then walked through the team’s key projects since adopting the platform. The first was a global product flow optimization model — the company’s first end-to-end view of its supply chain showing which manufacturing plant produces which product for which market, along with total landed costs — which he said unlocked millions of dollars in savings and, because it’s a repeatable model, can continue to be used for ongoing network optimization.
That initial success led to further use cases, including a cost-to-serve study for Castrol’s European industrial business that identified several areas of network inefficiency and, because much of the underlying data had already been collected, fed directly into a follow-on strategic network design project at the European level. The team also worked on a multi-source study in Australia focused on bulk deliveries, looking less at cost reduction and more at improving delivery speed and frequency within a one-to-two-day window. Most recently, the team used the platform for resilience planning — running alternative sourcing configurations last year to help mitigate the impact of new tariffs, including rerouting supply sources away from China to avoid higher tariff exposure.
Reflecting on 18 years of work, Davis said the shift has moved Castrol from one-off, project-based modeling to models that are repeatable, refreshable, and continuously maintained in what he described as a kind of model library. He cited three main benefits: faster answers across a broader range of use cases, and reduced dependency on external partners to run the system.
That success has also created a new challenge — demand for the team’s services now exceeds its current capacity to deliver. To address that, Castrol is working toward what Davis called a “community of excellence” model, shifting ownership of modeling capability out to the regions so that regional teams can run models, maintain them, and execute scenarios independently, with support and guidance from the central team rather than doing all the work themselves.
He said newer platform capabilities — particularly around data science and AI, including Ada — should help make that shift possible by allowing less technical users in the regions to collect and manipulate data and run useful scenarios on their own, ultimately democratizing modeling capability across the business. He closed by noting Castrol is still early in exploring these data science capabilities but intends to lean into them as much as possible to keep delivering value to the business.
Designing for What’s Next
Across every session, one thread tied the day together: the shift from reacting to disruption toward designing for it. Whether it was 3M rethinking a fragmented global network one export lane at a time, Amazon rebuilding its delivery model around Brazil’s unique geography, or Castrol moving from outsourced, one-off modeling to an internal capability it’s now working to scale across regions, each speaker described the same underlying transition — from treating supply chain work as a series of fires to put out, to treating it as a system that can be deliberately designed, tested, and improved. Don Hicks’s framing from the opening keynote turned out to be the through-line for everything that followed: design isn’t a one-time project or a buzzword; it’s a discipline, and the organizations getting the most value from these tools are the ones building that discipline into how their teams operate every day, not just how they respond when something breaks.
It was clear from these sessions that the technology itself, whether that’s Ada, DataStar, or the broader platform, is advancing quickly, but the more interesting story was how differently each company is choosing to put it to work. 3M is using it to bring discipline to a fragmented global footprint. Amazon is using it to solve problems without precedent in more mature markets.
Castrol is using it to shift ownership of modeling capability out of a small central team and into the regions. That range is probably the best evidence that this isn’t about one tool solving one problem, but about giving more people across an organization the ability to ask “why” and act on the answer. If there’s a single takeaway to carry out of OptiCon, it’s that the companies pulling ahead aren’t necessarily the ones with the most data or the fanciest models, but the ones building a lasting habit of designing for change instead of waiting for it to arrive.
