We ran a live agentic AI demo on real shipping data. The modeler approved every move.

Ask any senior supply chain modeler where their time actually goes and you’ll get the same answer. Eighty percent of it disappears into data prep. The strategic work, the scenarios, the recommendations to the business, all of it gets squeezed into whatever time is left.

We applied agentic AI to address that eighty percent.

Last week we ran a live demo in front of an audience. Our head of solutions Bo Wang opened a raw shipment data file pulled straight from a transactional system. 95 customers, 34 manufacturing sites, one DC. The kind of file that typically eats weeks or months of a modeler’s life before any real work starts.

Ten minutes later, we had a working baseline at $30 million total network cost and two scenarios run on top of it. One tested a 10% demand increase. The other pulled three manufacturing sites out of the network. Both were triggered by typing plain English into a chat box. Neither ran until Bo reviewed wha t the agent proposed and approved it.

That last part is the part worth pausing on.

A lot of vendors are racing to make AI more autonomous. We went the other way on purpose. The agent handles the mechanical work that swallows the modeler’s day, mapping data, building tables, configuring scenarios. The modeler keeps control of every decision that changes the model.

What we’re not claiming.

The dataset was mid-sized. We have not pressure-tested the agents on a multi-billion-dollar global network with messy, multi-source data yet. The agents are in alpha, with broader platform support coming. The output agent is still in development.

We’re showing the work in progress because the workflow is the point.

We don’t think the goal of AI in supply chain design is to remove the modeler. The goal is to give experienced modelers their week back, and to give teams without a deep modeling bench a way in.

Watch the full webinar recording

If your modeling projects are still losing months to data prep, the recording is worth 30 minutes of your time.