Todd Zabelle is back. Last time we talked flow. The time before that, implementation. This one, we put AI on the table and watched it fail at the thing everyone keeps promising it will fix.
The planning stage isn't broken. It's missing. Engineers got pulled out of the how. Tradespeople got pulled out of the what. The CM sits in the middle making bars. Three groups, three languages, nobody speaking the same one. AI doesn't fix that. AI walks in, asks for a price, and gets handed a schedule made by someone who never built anything.
Todd makes the case that AI is a subordinate tool. Useful for equipment data, dossier work, requirements management, traffic routing. Not useful for designing the production system. That still takes a human who's been on the tools and knows operations science. One plus one doesn't equal two. Capacity utilization isn't linear. The sum of cycle times isn't the cycle time. None of that is in the model.
Then we got into the data center build. Switchgear delays. Labor shortfalls. A town that kicked out its own council to stop a data center. GPU supply tightening. Banks getting jumpy on syndication. Todd's read: it's not 1999.com. It's 2012 fracking. Everybody at the feeding trough. Three years from a thinning.
The Empire State Building got built in less than a year with sketches. We can't build a tilt-up on time with 3D models and an MRP system. Todd has a theory about why. So do I.
This one's for the executives who keep buying AI as the answer without asking what the question was.
