Every firm chasing "AI for productivity" is optimising port turnaround in 1955 - measuring the wrong thing while the system around them gets rewritten.
Most AI ROI conversations sound like cost-saving spreadsheets: cheaper drafts, faster summaries, fewer headcount. That's the productivity frame.
The productivity frame predicts the wrong outcomes - because AI's real lever is the same as the container's. It's not how fast each component goes. It's how the whole system coordinates.
Automation makes one part of a system faster. A 20% faster crane gets you 20% more throughput, no more. The economics are linear and capped.
Coordination makes the whole system align differently. When containers standardised global shipping, the gain wasn't faster cranes - it was that factories could be on a different continent than the customer, supply chains could stretch across the world, and entire ports either thrived or collapsed based on the new logic. Same ships. Different coordination.
AI is the first technology in decades that can coordinate across fragmented systems without requiring them to standardise first. It reads unstructured inputs, builds shared representations on the fly, drives aligned action across actors who never sat in the same room. Read it as automation and you'll measure productivity. Read it as coordination and you'll see the system reshuffling.
It was a new logic of retail. Most chains missed it.
In the 1970s, both Walmart and Kmart adopted barcodes. Both saw faster checkout, fewer errors, lower labour costs. The first-order win was identical for both.
Walmart treated the barcode as a coordination layer. Every scan fed a satellite network that connected stores, warehouses, and suppliers into one system. Walmart could renegotiate supplier terms in real time, restock proactively, standardise SKUs across thousands of stores. Kmart treated the barcode as a checkout speedup. Same technology. Different architecture.
A decade later, Walmart had restructured the entire retail industry - and Kmart was on its way to irrelevance. The technology didn't make the difference. The choice of frame did.
Reading AI as automation is the modern equivalent of measuring crane speed in 1955. You're not wrong about the number. You're wrong about which number matters.
Used barcode data as a live feedback loop across the entire supply chain. Built private satellite systems. Forced suppliers to integrate. Restructured how retail negotiated with manufacturing.
Outcome: became the largest retailer in history. Captured the value the coordination layer unlocked.
Used barcodes to scan faster. Kept store managers making independent decisions. Kept the supplier-led negotiation model. The technology was adopted; the architecture wasn't.
Outcome: long, slow decline into irrelevance. Same technology. Wrong frame.
What does your firm's AI roadmap measure success on - and would Kmart have used the same metrics?
The first-order win is real but small. The second and third order is where the system reshuffles. Frame matters more than tooling.
Don't measure crane speed. Find the coordination layer your industry is about to run on - and own it.
The advisory engagement that takes AI seriously as a coordination layer - mapping your industry's coordination failures, designing the agentic coordination model, and claiming a position you can credibly own as the layer the ecosystem routes through.