AI Simulation Is the New Standard

Inside the AI Transformation of Parts, People & Processes

Hi, I’m Lily. I live in the world of distribution, where for years, the bet was on hardware: more AMRs, more conveyor systems, more robotic arms.

What's the point of having more robots if they can't work together? This week's stories point to exactly this kind of question emerging on the warehouse floor. One that's less about adding machines and more about making the ones you already have think in sync. Less about deploying technology and more about simulating the entire operation before a single robot moves. The physical world is hard to automate; it always has been. But something is shifting. The warehouse is becoming a software problem.

If you only read one thing this week, it is this:

Otto Group built a Coordinated Autonomy Layer (CAL) to make robots from different vendors operate as a single system across its warehouses, using an NVIDIA Omniverse digital twin to simulate and optimize before touching the floor. One layout change alone cut robotic stop-and-go events by 20%, already implemented live.

What’s Working in the Field

UNFI's AI Inventory Tool Goes Network-Wide

UNFI is deploying an AI-powered inventory planning platform across its full network of 49 distribution centers, with roughly a dozen going live this month and full rollout targeted by August 1. The implementation is already delivering improvements in customer fill rates, inventory management, and free cash flow, according to CEO Sandy Douglas. The system complements UNFI's decentralized procurement model, enabling local DC teams to read demand signals and adjust purchasing decisions faster. The company views this as an early-stage transformation with significant value still ahead.

Körber and NVIDIA Build AI Warehouse Twins

Körber and NVIDIA are partnering to build physics-accurate digital twins of logistics facilities, using NVIDIA Omniverse to simulate real-world operations before any hardware is installed. For logistics operators, the practical payoff is significant: complex scenarios like peak demand surges and system breakdowns can be stress-tested virtually, compressing the path from concept to go-live. Körber CEO Stephan Seifert framed the twins as enabling "faster and more confident decision-making" and "measurable performance gains" for customers worldwide. The partnership signals a broader industry shift: the warehouse of the future is being designed in simulation first.

WSJ: Bezos Bets Big on Physical AI

Jeff Bezos is in early talks to raise $100 billion for a fund designed to acquire manufacturing companies in sectors like chipmaking, defense, and aerospace, and transform them using AI-driven automation. The vehicle is tied to Project Prometheus. The fund would rival the scale of SoftBank's Vision Fund, with sovereign wealth funds and JPMorgan Chase among those in early discussions. While operational deployments are still ahead, the fundraising effort signals a decisive shift: serious institutional capital is now betting that physical AI is the next frontier for manufacturing transformation.

Otto Group AI Makes Robots Collaborate

Otto Group is solving one of warehousing's most persistent problems: robots from different vendors that can't coordinate with each other. Using Boston Dynamics' Spot to scan a pilot warehouse and NVIDIA Omniverse to build a precise 3D digital twin, Otto created a Coordinated Autonomy Layer (CAL) that sits between its WMS and its robot fleets, managing task assignment and routing across the entire floor. Simulation-driven layout changes have already delivered a 20% reduction in robotic stop-and-go events, and that result has been implemented in the live warehouse. The next step is an AI "artificial brain" that dynamically adjusts how robots are assigned and routed in real time. Otto's approach reframes the warehouse automation question entirely: it's not about how many robots you have, it's about whether they can think in sync.

Zalando Deploys 50 AI Robots Across Europe

Zalando is rolling out up to 50 AI-powered Nomagic robots across its European fulfillment network, scaling from a pilot that delivered 100,000 picks per day to deployments across six countries. The robots, nicknamed "Richard," use computer vision to recognize product shapes and materials. The expansion also introduces a world-first: the Shoebox Picker, which solves one of fashion logistics' most stubborn automation problems by combining AI vision with precision grippers that keep shoebox lids intact during handling. VP Logistics Network Marcus Daute framed the rollout as smart automation that works alongside people, not instead of them, freeing employees for higher-value tasks.

What’s In My Ears

The State of AI in Warehousing and Omnichannel with Matthias Winkenbach, Eva Ponce, and Alejandro González

In this episode of MIT Supply Chain Frontiers, Matthias Winkenbach and Eva Ponce of MIT and Alejandro González of Mecalux unpack findings from two major 2026 reports on AI in warehousing and omnichannel supply chains. The verdict: AI has moved from experimentation to operational necessity, with only 1% of warehouse operators not yet using or piloting it. Winkenbach spotlights GENESIS, an AI-based inventory simulator that cuts a three-day optimization task down to three minutes. A sharp, data-grounded listen for anyone tracking where AI is actually delivering in warehouse and fulfillment operations.

Lily’s Quick Take

Three things stood out this week. First, simulation is becoming the new standard. Both Otto Group and Körber are designing and optimizing operations in digital twins before touching the floor. Second, the coordination problem is finally getting addressed: having robots from different vendors that can't talk to each other is one of the industry's oldest headaches, and Otto Group's CAL is a serious attempt at solving it. Third, UNFI is a reminder that the most durable AI deployments are built into a broader operational model. The technology follows the process, not the other way around.

Until next week—keep your systems learning!

— Lily @ InstaLILY AI