Ford spent months letting an automated system catch design flaws its senior engineers used to catch. It couldn’t. So Ford rehired 350 of them. IBM automated 94 percent of its HR requests and is now tripling entry-level hiring anyway. I watched both of those stories break this year, and neither one surprised me.
I recently spoke with TechSlassh about the one call I’ve never handed off to a machine, even after years of running my companies with AI built into almost every function.
Here’s the pattern behind Ford, IBM, and Commonwealth Bank of Australia’s reversals: each one automated the easy 94 percent and assumed the hard 6 percent would follow. It didn’t. That 6 percent was never a data problem. It was judgment, and judgment doesn’t scale the way a workflow does.
I run AliveDevOps on infrastructure that has to work correctly around the clock, and I still won’t let a system decide whether a client relationship, a hire, or a strategic bet is the right one. AI won’t replace good judgment. It’ll amplify it. That distinction sounds small until you’re the one issuing a public correction like Ford did.
I also hold a genuinely aggressive view on AI everywhere else. I’ve said I expect to see the first solo founder run an entire company at real scale with AI handling nearly everything, an idea I dug into further in a post on the solo founder model. Those two positions only look contradictory until you separate execution from judgment. AI can run the pipeline. It can’t yet weigh a tradeoff that depends on knowing a person well enough to read between the lines.
I trace my comfort holding that line back to competitive golf, where you can’t blame a bad decision on faulty equipment, something I’ve written about before when it comes to how athletic discipline shapes decision-making. The tools work now. That just means responsibility for the call shifts entirely onto the person making it.
Read the full piece for the Ford, IBM, and CBA numbers in full, plus how this line shows up in the AI-powered workflows we build at Pabs Marketing.
The founders who name the decision they’ll never delegate now are the ones who won’t be issuing a correction later.
