AI Reveals the Integration Gap

Inside the AI Transformation of Parts, People & Processes

Hi, I’m Lily. I live in the world of distribution, where automation has become the scapegoat for every disruption it was supposed to prevent. When supply chains falter, the diagnosis usually lands the same way: the algorithm failed, the system overreacted, the tools moved faster than the people. That story is convenient… and wrong. Automation didn’t create these vulnerabilities. Poor integration did, while automation simply stopped hiding them. This week’s stories live at that dividing line: what it costs when integration is missing, and what it delivers when it’s done right.

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

Jabil, operating more than 100 manufacturing and supply chain facilities across 25 countries, deployed an AI sourcing optimization platform and generated $25 million in logistics savings and cost avoidance. Sourcing cycles shortened by roughly a month; analysis that once consumed weeks now resolves in hours. When AI connects to the right decision workflows, it closes the loop.

What’s Working in the Field

AI Investments Fall Short on Integration

A PwC survey covered in Supply Chain Dive found that 92% of warehouse technology leaders say their investments haven’t fully delivered, and 47% trace the gap directly to integration complexity. Three named deployments anchor that finding: Amazon is building a natural-language communication layer between warehouse robots and floor workers; Walmart has deployed AI-powered autonomous forklifts through Fox Robotics; and GXO Logistics is running Dexterity’s AI-equipped arms for depalletizing, labeling, and repalletizing at scale. Of those surveyed, 57% had already integrated AI in some form and still couldn’t demonstrate meaningful results. To illuminate the main point: the tools were in place; the connections weren’t.

Jabil Closes the AI Sourcing Gap

Jabil, running 100+ manufacturing and supply chain facilities across 25 countries, deployed an AI sourcing optimization platform and realized $25 million in logistics savings and cost avoidance. Sourcing cycles dropped by roughly one month; the analysis phase that once consumed weeks now resolves in hours. These results came from a horizontal procurement platform, not purpose-built for any specific industry. When a general-purpose AI tool moves this much value by connecting procurement decisions to supply chain data, the ceiling for vertically designed agents, built for how distribution and manufacturing actually operate, simply ceases to exist.

AI Agents Enter the Control Tower

ABI Research surveyed 490 supply chain professionals this year and found that 65% rank AI and generative AI as important/very important to their technology decisions. Control tower platforms (Blue Yonder and FourKites among the leaders) are embedding agentic capabilities that can execute corrective actions without waiting for escalation. Among manufacturers, 77% are already considering, piloting, or deploying autonomous mobile robots and AGVs. One finding stands out: executive appetite for AI autonomy runs strong at the tactical level and drops sharply at the strategic one. Autonomous exception management is IN. Autonomous strategy is still earning its mandate.

Supply Chain Leaders Code with AI

Researchers at Ohio State and Grand Valley State argue that agentic coding tools are handing supply chain professionals a meaningful new capability: building functional prototypes and decision-support tools without a dedicated engineering team. Domain expertise, knowing where handoffs break, what planners actually need, how freight data moves, is now a direct input into software creation. A supply chain leader who grasps the process can test a working solution in days rather than queuing for months in an IT backlog.

Stord Bets $250M on Physical AI

Stord raised $250 million in a Series F round at a $3 billion valuation and launched Stord Labs, an Atlanta facility for testing physical AI and robotics in live warehouse conditions before deploying across its network. The platform serves 1,000+ customers across roughly 100 fulfillment locations, handles over $15 billion in annual gross merchandise value, and reached nearly 20% of U.S. homes in 2025. For mid-market fulfillment operators, the signal is clear: physical AI working alongside active robotics is moving from proof-of-concept to infrastructure expectation.

What’s In My Ears

Rethinking Global Trade Infrastructure with Rathna Sharad

In this episode of the Supply Chain Podcast, Rathna Sharad, CEO and founder of FlavorCloud, walks through building an AI-native stack for cross-border logistics. Her background (UPS, Menlo Worldwide, Microsoft) gives operational grounding to what usually reads like a software pitch. Worth a listen for: how FlavorCloud automates customs classification and compliance at scale, what it means to serve as importer of record for hundreds of brands, and why the gap between digital commerce speed and physical logistics infrastructure is still the field’s most underrated constraint.

Lily’s Quick Take

The organizations making AI work aren’t the ones with the most tools. They’re the ones with the fewest gaps between tools. Amazon, Walmart, and GXO are running sophisticated robotics deployments, and 92% of warehouse technology leaders surveyed still say their investments haven’t delivered. Jabil connected AI to actual decision workflows and recovered $25 million. The difference wasn’t capability. It was integration.

What’s shifting now is where that integration work is happening. Control towers are starting to close the loop autonomously. Supply chain leaders are building decision-support tools themselves, without waiting on IT. Work that once sat in a years-long backlog is moving in days. That’s not a technology story, but rather an organizational one.

Automation scales what you build. Integration determines what’s worth scaling.

Until next week—keep your systems learning!

— Lily @ InstaLILY AI