Cutting Heads or Buying AI Won’t Fix What’s Actually Broken

Campbell’s just cut 13% of its workforce trying to turn its operations around. In the same week, a new industry report found that 98% of manufacturers are exploring AI, but only 20%, or fewer, feel genuinely ready to use it at scale.

Different decisions. Same mistake.

I worked with a company that had been a world leader in its space for decades, unmatched technically, until the market caught up. The moment leadership felt competitors closing in, they reached for the lever they believed would fix delivery and cost: an IT resource was assigned to build a dashboard for production and maintenance visibility.

The dashboard didn’t fix anything. It just revealed what was already true. Scheduling wasn’t a system, it lived in one retiree’s head. Production control ran on anecdote. Every supervisor had their own version of FIFO. The system wasn’t broken. It had never existed.

That’s the part technology can’t fix by itself. A dashboard gave that company visibility into where a part was in the process. It gave them nothing to plan capacity with, nothing to predict tomorrow with. Visibility isn’t the same as the ability to act on what you see.

The clearest sign of that gap wasn’t a chart. It was a literal sign, posted by a supervisor whose team wasn’t receiving paperwork from the department before them: “No paperwork, no polish.” Rather than walk over and ask for it, he stopped the line and waited. That’s not a technology problem. That’s an ownership problem, and no software fixes it.

Fixing it meant going back to fundamentals: reinforcing that waiting and hoping wasn’t how the team operated, that ownership meant closing the loop yourself, not waiting for someone else to notice. Not everyone bought in immediately. Real change rarely lands cleanly or all at once, and leading through that honestly matters more than pretending it did.

If Campbell’s called tomorrow, before any more headcount came out, the first thing worth checking isn’t the org chart. It’s whether the numbers underneath it are even accurate: is the capacity definition correct, is the labor efficiency figure correct, and how many active improvement projects does each person actually own. That last question matters more than it sounds. Once someone is carrying more than two or three real projects at once, completion rates fall off sharply, a principle straight out of the 4 Disciplines of Execution. One organization we worked with had 100 improvement projects listed on its dashboard. None were close to finished.

That same principle explains the AI statistic. The 20% of manufacturers who feel ready to use AI at scale aren’t the 20% with the best tools. They’re the ones who already built the operating discipline, the ownership, the cadence, the accountability, that any tool needs in order to actually work. The other 80% are buying technology to replace a discipline that was never built in the first place.

Cutting headcount and investing in AI are the same decision wearing different clothes. Both can be the right move. Neither one fixes anything if the operating system underneath is still missing. Without it, you’re not solving the problem. You’re just making it move faster.