AI can connect systems fast, but only if the data is right | Ep. 2

The storefront said overnight shipping. The part showed up three days later, and the dishwasher was still broken.ㅤMatt Johnson and Floyd Blaikie open Episode 2 of Data vs. Commerce by tracing that delay past the product data they covered last time, into the place most teams forget to look: the integrations between systems. Matt walks through how an e-commerce system can confidently promise overnight delivery while the ERP and OMS behind it know the item is sitting in a warehouse across the country. The storefront wasn't lying. It just never got the truth from the back end.ㅤFrom there the two get into why national retailers and distributors end up with a twisted ball of yarn: 30 acquisitions, 17 Salesforce instances, crusty middleware nobody wants to touch, and load-bearing systems you can't switch off without the business crawling to a halt. This is a solo episode, hosted by Matt Johnson and Floyd Blaikie of Pivotree. The friction this week isn't data against commerce. It's the invisible plumbing between them, and who actually owns it.ㅤ📌 What We CoverㅤWhy an e-commerce system can promise overnight shipping when the item is actually in stock across the country, in a different warehouseHow distribution center location and on-shelf inventory live in back-end systems that often never connect to the front-end customer experienceWhat growth by acquisition does to your stack: inherited ERPs, inventory systems, and disparate data that all have to talk to one storefrontThe 17 Salesforce instances problem, and why companies insist on keeping every system they acquireBrittle, hard-coded integrations and crusty middleware held together by one person in ITWho needs the macro view of the system architecture, and why nobody clearly owns itWhere AI helps: reading APIs and connecting data faster than any human developer, plus agents that flag when an integration is about to breakWhere AI backfires: order management systems that hallucinate a part because the underlying data was wrongㅤ🔗 Resources MentionedPivotree, named on the episode for its work on AI integration frameworks and processesSalesforce, referenced via the company running 17 separate instancesThe system categories at the center of the conversation: ERP, OMS, PIM, and the front-end e-commerce platform, plus the integration layer and middleware that connect them