Published by
Published on
May 11, 2026
The modeler used to be the bottleneck. Your team had questions — dozens of them, hundreds maybe — and one person standing between the question and the answer. Greg Mueller, former Cargill and Best Buy supply chain leader, describes it plainly: "The modeler became the bottleneck because there were many, many more questions that you needed to answer than there were modelers."
That bottleneck is gone now. Not because the questions disappeared, but because the work did.
Agentic AI is shifting how supply chain design gets done. The organizations paying attention now will set the pace for the next decade.
Agentic AI refers to specialized AI systems that perform complex, multi-step tasks autonomously while keeping humans in control of the decisions that matter. Chatbots answer questions. Dashboards visualize data. Agentic AI does the work — classifying your data, building your models, running your scenarios, and generating your reports.
The distinction matters. Plain AI gives you speed. Agentic AI brings the domain intelligence to know what questions to ask, what opportunities to surface, what data gaps to fill, and what risks to flag.
Vik Srinivasan, Vice President of Solution Architecture at Optilogic, frames it this way: "Agentic AI has really helped us take a lot of the burden out of our user to automate areas that can be automated very easily, but still provide the human-in-the-loop experience where the user is still in full control, can steer the direction of the agentic AI workflows, can validate, override assumptions, and can continue to iterate and tweak as the agents build these models."
That last part is critical. You're not handing over the keys. You're getting a team of tireless analysts who do the tedious work and bring you the decisions that require judgment.
Most AI promises sound like this: "Leverage advanced machine learning to optimize your supply chain operations." Vague. Generic. The kind of sentence that could appear on any vendor's website.
What actually happens with agentic AI looks different. The prompt is three words: "Build a Model."
Neeru Bhopal, Director of Product Management at Optilogic, demonstrated exactly this: "The Agent Chat, which is the same interface for how this model is actually built, used the Modeler Agent, and it started life as a three-word prompt — Build a Model — to transform messy raw data into a fully functional baseline model."
Think about what that replaces. The weeks of schema mapping. The manual data cleansing. The consultant calls to figure out why your shipment file won't load. Gone.
The system recognizes what you bring it. Vik explains the shift: "Once you connect it and bring it to our systems, we're able to understand first what type of data is it. Is it a shipment file? Is it an order file? In the past, you had to do exact mappings to specific schemas. But really with the technology we have and the ontologies we've built, you can bring your supply chain data into our platform and we will be able to recognize and provide the right data dictionary around it."
The AI handles the classification. The AI handles the cleansing. The AI builds the model. Then it asks you — the human — what question you want to answer.
Let's be specific about the compression. Vik Srinivasan puts a number on it: "What would take several weeks has gone down to a few hours or days."
That's a category shift, and the implications ripple outward.
Your team stops rationing questions. When a model takes three months to build, you ask one question and hope it's the right one. When a model takes hours, you ask twenty. You explore. You test the assumptions that kept you up at night but never seemed worth the project time.
Competitive advantages built on analyst headcount shrink fast. The company with a larger analytics team used to win by volume — more models, more scenarios, more confidence. That advantage compresses when any team can produce a model in an afternoon. What matters now is how fast you can move from data to decision.
Your strategic planning horizon expands. Three-year network design projects made sense when the work took a year. But supply chains don't move on three-year cycles anymore. Tariffs change overnight. Suppliers fail without warning. Markets open and close in months. When you can model the response in hours, your planning can finally match the pace of change.
The skepticism is understandable. Every AI vendor promises to "automate" something. Usually that means automating the parts that were already easy and leaving you with the hard stuff plus a new integration problem.
Agentic AI in supply chain design works differently. The 80% it handles is genuinely the tedious part — the data prep, the model construction, the scenario generation, the output formatting. These tasks require precision and patience, not judgment.
What remains is the work that matters: the strategic decisions, the tradeoff calls, the judgment about which scenario reflects your market reality.
Greg Mueller describes the shift this way: "It's really now shifting the way that we can start to interact with these models through apps, through AI, natural language to be able to put the model at the center instead of the modeler."
The model sits at the center. Not the modeler. Not the AI. The model — your representation of your supply chain, continuously updated, always ready to answer the next question.
Your team doesn't disappear in this world. They become more valuable. They spend less time wrestling with data and more time interpreting results, challenging assumptions, and making the calls that shape strategy.
The agents don't stop at building baseline models. Consider what happens with demand.
Optilogic's demand modeling agent — called Pulsar — does something manual analysis rarely achieves. Neeru Bhopal explains: "Pulsar analyzes shipment history across multiple hierarchies and models growth at different customer product levels across all possible combinations, but it also surfaces causal drivers."
That last phrase matters. Surfaces causal drivers. Not just "here's what demand looked like" but "here's what drove it." That's the difference between a report and an insight.
Manual demand analysis typically examines the dimensions you think to ask about. The agent examines all of them — every customer, every product level, every combination — and tells you which ones moved the needle.
For executives who've watched demand planning teams struggle to explain variance after the fact, this is the shift: the explanation arrives before you have to ask.
Here's a problem agentic AI solves that rarely makes the feature list: organizational access.
Traditional supply chain models are built for control, not collaboration. One team owns the model, one team runs it, and everyone else either waits for the output or quietly builds their own version with different assumptions. You end up with competing analyses pointing in different directions — and no clean way to know which one to trust.
The alternative is what Optilogic calls always-on what-if modeling: "Hundreds and hundreds of users in your company should be able to leverage the model, but in their own sandbox, play around with as many scenarios without having to change what the planning teams are planning on today."
Each user gets their own space. They test their ideas — the regional manager wondering about a new distribution center, the procurement lead modeling a supplier switch, the finance team stress-testing a capacity decision. None of them break the production model. None of them wait in queue.
Supply chain design becomes an organizational capability, not just a technical one. The bottleneck wasn't only the modeler. It was the access.
The narrowest version of supply chain design lived in one team. The analysts who knew the tools. The consultants who built the models. The specialists who could translate business questions into mathematical constraints.
Agentic AI opens the aperture. Vik Srinivasan describes the expansion: "The AI allows access to many more personas who are probably more in the tactical horizons, like planners and analysts who can get the same value of scenario planning within their capabilities."
This isn't about simplifying the analysis. It's about removing the technical barrier between good questions and good answers. The people who needed the insight were always there. They just couldn't get to it without going through someone else first.
That changes what's possible across the whole organization:
The agents don't stop at building and running models. They tell you what the results mean.
As Neeru Bhopal describes it, the output analysis agent generates an executive-level report — headline scenario outcomes, cost driver decomposition, facility hotspots, and prioritized recommendations.
That used to be weeks of work after the model finished running: interpreting outputs, building the narrative, translating optimization results into language executives could actually act on. The agent handles that structured layer immediately — the decomposition, comparison, ranking, and formatting that always had to happen before any human interpretation could begin.
Your team reviews the draft, adds the context only they know, and presents something meaningful — instead of spending those hours assembling raw materials for a report that might never get written.
The practical path from raw data to working model looks like this:
Connect your data. Optilogic supports Snowflake, Databricks, Google BigQuery, Amazon and Azure data factories, AWS S3, CSV files, and planning system extracts — about nine different connection types. Bring what you have. The system figures out what it is.
Let the agents classify and cleanse. The data agents profile your files, identify what type of data you're working with, map it to the right schema, and fill gaps where needed. No manual mapping. No consultant engagement.
Build your baseline model. Build a Model. The modeler agent constructs a working network optimization model from your data. You review it, adjust assumptions, and iterate through conversation — not through weeks of project work.
Run scenarios. You can run scenarios, ask questions, and test assumptions directly — without waiting for someone else to set up the sensitivities manually. The what-if agents handle that.
Get your answers. The insights agents summarize what matters: the headline outcomes, the cost drivers, the recommendations. You make the decision.
What took months now takes hours.
Ask yourself what your team would do with the time back.
The hours spent cleaning data. The weeks spent building models. The months spent waiting for scenarios to run. What strategic decisions would you make if you had the capacity to model them first?
Your competitors are asking the same question. Some of them are already answering it.
Agentic AI in supply chain design is available now, not somewhere on a future roadmap. The organizations that move first won't just have faster models — they'll have more confident decisions, more explored options, and more prepared playbooks for when conditions change.
The modeler bottleneck is broken. The model sits at the center. The question is whether you're ready to use it.
The modeler used to be the bottleneck. Your team had questions — dozens of them, hundreds maybe — and one person standing between the question and the answer. Greg Mueller, former Cargill and Best Buy supply chain leader, describes it plainly: "The modeler became the bottleneck because there were many, many more questions that you needed to answer than there were modelers."
That bottleneck is gone now. Not because the questions disappeared, but because the work did.
Agentic AI is shifting how supply chain design gets done. The organizations paying attention now will set the pace for the next decade.
Agentic AI refers to specialized AI systems that perform complex, multi-step tasks autonomously while keeping humans in control of the decisions that matter. Chatbots answer questions. Dashboards visualize data. Agentic AI does the work — classifying your data, building your models, running your scenarios, and generating your reports.
The distinction matters. Plain AI gives you speed. Agentic AI brings the domain intelligence to know what questions to ask, what opportunities to surface, what data gaps to fill, and what risks to flag.
Vik Srinivasan, Vice President of Solution Architecture at Optilogic, frames it this way: "Agentic AI has really helped us take a lot of the burden out of our user to automate areas that can be automated very easily, but still provide the human-in-the-loop experience where the user is still in full control, can steer the direction of the agentic AI workflows, can validate, override assumptions, and can continue to iterate and tweak as the agents build these models."
That last part is critical. You're not handing over the keys. You're getting a team of tireless analysts who do the tedious work and bring you the decisions that require judgment.
Most AI promises sound like this: "Leverage advanced machine learning to optimize your supply chain operations." Vague. Generic. The kind of sentence that could appear on any vendor's website.
What actually happens with agentic AI looks different. The prompt is three words: "Build a Model."
Neeru Bhopal, Director of Product Management at Optilogic, demonstrated exactly this: "The Agent Chat, which is the same interface for how this model is actually built, used the Modeler Agent, and it started life as a three-word prompt — Build a Model — to transform messy raw data into a fully functional baseline model."
Think about what that replaces. The weeks of schema mapping. The manual data cleansing. The consultant calls to figure out why your shipment file won't load. Gone.
The system recognizes what you bring it. Vik explains the shift: "Once you connect it and bring it to our systems, we're able to understand first what type of data is it. Is it a shipment file? Is it an order file? In the past, you had to do exact mappings to specific schemas. But really with the technology we have and the ontologies we've built, you can bring your supply chain data into our platform and we will be able to recognize and provide the right data dictionary around it."
The AI handles the classification. The AI handles the cleansing. The AI builds the model. Then it asks you — the human — what question you want to answer.
Let's be specific about the compression. Vik Srinivasan puts a number on it: "What would take several weeks has gone down to a few hours or days."
That's a category shift, and the implications ripple outward.
Your team stops rationing questions. When a model takes three months to build, you ask one question and hope it's the right one. When a model takes hours, you ask twenty. You explore. You test the assumptions that kept you up at night but never seemed worth the project time.
Competitive advantages built on analyst headcount shrink fast. The company with a larger analytics team used to win by volume — more models, more scenarios, more confidence. That advantage compresses when any team can produce a model in an afternoon. What matters now is how fast you can move from data to decision.
Your strategic planning horizon expands. Three-year network design projects made sense when the work took a year. But supply chains don't move on three-year cycles anymore. Tariffs change overnight. Suppliers fail without warning. Markets open and close in months. When you can model the response in hours, your planning can finally match the pace of change.
The skepticism is understandable. Every AI vendor promises to "automate" something. Usually that means automating the parts that were already easy and leaving you with the hard stuff plus a new integration problem.
Agentic AI in supply chain design works differently. The 80% it handles is genuinely the tedious part — the data prep, the model construction, the scenario generation, the output formatting. These tasks require precision and patience, not judgment.
What remains is the work that matters: the strategic decisions, the tradeoff calls, the judgment about which scenario reflects your market reality.
Greg Mueller describes the shift this way: "It's really now shifting the way that we can start to interact with these models through apps, through AI, natural language to be able to put the model at the center instead of the modeler."
The model sits at the center. Not the modeler. Not the AI. The model — your representation of your supply chain, continuously updated, always ready to answer the next question.
Your team doesn't disappear in this world. They become more valuable. They spend less time wrestling with data and more time interpreting results, challenging assumptions, and making the calls that shape strategy.
The agents don't stop at building baseline models. Consider what happens with demand.
Optilogic's demand modeling agent — called Pulsar — does something manual analysis rarely achieves. Neeru Bhopal explains: "Pulsar analyzes shipment history across multiple hierarchies and models growth at different customer product levels across all possible combinations, but it also surfaces causal drivers."
That last phrase matters. Surfaces causal drivers. Not just "here's what demand looked like" but "here's what drove it." That's the difference between a report and an insight.
Manual demand analysis typically examines the dimensions you think to ask about. The agent examines all of them — every customer, every product level, every combination — and tells you which ones moved the needle.
For executives who've watched demand planning teams struggle to explain variance after the fact, this is the shift: the explanation arrives before you have to ask.
Here's a problem agentic AI solves that rarely makes the feature list: organizational access.
Traditional supply chain models are built for control, not collaboration. One team owns the model, one team runs it, and everyone else either waits for the output or quietly builds their own version with different assumptions. You end up with competing analyses pointing in different directions — and no clean way to know which one to trust.
The alternative is what Optilogic calls always-on what-if modeling: "Hundreds and hundreds of users in your company should be able to leverage the model, but in their own sandbox, play around with as many scenarios without having to change what the planning teams are planning on today."
Each user gets their own space. They test their ideas — the regional manager wondering about a new distribution center, the procurement lead modeling a supplier switch, the finance team stress-testing a capacity decision. None of them break the production model. None of them wait in queue.
Supply chain design becomes an organizational capability, not just a technical one. The bottleneck wasn't only the modeler. It was the access.
The narrowest version of supply chain design lived in one team. The analysts who knew the tools. The consultants who built the models. The specialists who could translate business questions into mathematical constraints.
Agentic AI opens the aperture. Vik Srinivasan describes the expansion: "The AI allows access to many more personas who are probably more in the tactical horizons, like planners and analysts who can get the same value of scenario planning within their capabilities."
This isn't about simplifying the analysis. It's about removing the technical barrier between good questions and good answers. The people who needed the insight were always there. They just couldn't get to it without going through someone else first.
That changes what's possible across the whole organization:
The agents don't stop at building and running models. They tell you what the results mean.
As Neeru Bhopal describes it, the output analysis agent generates an executive-level report — headline scenario outcomes, cost driver decomposition, facility hotspots, and prioritized recommendations.
That used to be weeks of work after the model finished running: interpreting outputs, building the narrative, translating optimization results into language executives could actually act on. The agent handles that structured layer immediately — the decomposition, comparison, ranking, and formatting that always had to happen before any human interpretation could begin.
Your team reviews the draft, adds the context only they know, and presents something meaningful — instead of spending those hours assembling raw materials for a report that might never get written.
The practical path from raw data to working model looks like this:
Connect your data. Optilogic supports Snowflake, Databricks, Google BigQuery, Amazon and Azure data factories, AWS S3, CSV files, and planning system extracts — about nine different connection types. Bring what you have. The system figures out what it is.
Let the agents classify and cleanse. The data agents profile your files, identify what type of data you're working with, map it to the right schema, and fill gaps where needed. No manual mapping. No consultant engagement.
Build your baseline model. Build a Model. The modeler agent constructs a working network optimization model from your data. You review it, adjust assumptions, and iterate through conversation — not through weeks of project work.
Run scenarios. You can run scenarios, ask questions, and test assumptions directly — without waiting for someone else to set up the sensitivities manually. The what-if agents handle that.
Get your answers. The insights agents summarize what matters: the headline outcomes, the cost drivers, the recommendations. You make the decision.
What took months now takes hours.
Ask yourself what your team would do with the time back.
The hours spent cleaning data. The weeks spent building models. The months spent waiting for scenarios to run. What strategic decisions would you make if you had the capacity to model them first?
Your competitors are asking the same question. Some of them are already answering it.
Agentic AI in supply chain design is available now, not somewhere on a future roadmap. The organizations that move first won't just have faster models — they'll have more confident decisions, more explored options, and more prepared playbooks for when conditions change.
The modeler bottleneck is broken. The model sits at the center. The question is whether you're ready to use it.
Fill out the form to unlock the full content
.jpg)
.jpg)
.jpg)