Why Most AI Projects Are Failing — And What to Do Before It's Too Late

Tom Schmitt is CEO of Radial, North America's largest third-party fulfillment provider — and a company about to become something bigger. As part of the Bpostgroup's rebranding to Paxon, Radial is merging with its sister companies into a single global logistics brand built on decades of e-commerce infrastructure.The Watson Weekly is sponsored by Avalara. For more information on Avalara, visit - https://avalara.watsonweekly.com/Tom sat down with Rick Watson to talk about what he's actually seeing on the ground as AI enters supply chain — and the answer is more nuanced than most headlines suggest.Three things worth your time from this episode:The trust gap is real and measurable. Radial surveyed 1,000 US consumers: nearly two-thirds are uncomfortable sharing payment information with AI agents. That's not a technology problem — it's a brand problem operators need to solve first.Most AI projects are failing their ROI test right now — and that's expected. Tom's crawl, walk, run framework isn't a hedge; it's a diagnosis. Demand forecasting and pick-path optimization have measurable returns today. Fully autonomous warehouse orchestration does not — yet.Agentic commerce has a data standards problem nobody is talking about. Radial is a founding member of ONX — the Order Network Exchange — an open industry standard for how order, inventory, and fulfillment data moves between systems. Agentic commerce only works if agents read the same language. That work is already underway.Plus: the history of GSI Commerce, how Michael Rubin invented the direct-to-consumer industry, and why Fred Smith's old line about FedEx still tells you everything about where supply chain AI is headed.Subscribe to the Newsletter: Get it at https://www.watsonweekly.com/subscribe