AI Doesn't Fix Your Problems. It Speeds Up How Fast You Find Them.
Follow the S-Curve
Two pieces ago, I argued that innovations arrive on emotion and only survive on fixed-cost reduction. Last piece, I broke down why fixed cost doesn't fix itself — a bottleneck is usually 90% process conflict and 10% human problem-solving, and the two delays compound instead of canceling out.
Here's where that leaves AI, specifically.
AI does not fix your problems. Without a focus on fixed-cost reduction, AI makes your fixed-cost problem happen quicker.
Speed Doesn't Discriminate
AI is, at its core, a speed and scale multiplier. It doesn't know the difference between a well-designed process and a broken one. Point it at a genuine fixed-cost reduction — a real bottleneck, correctly diagnosed, with the human side already resolved — and you get real gains, faster than you'd have gotten them any other way.
Point it at a process that's still fighting itself, or a decision that a person has been quietly avoiding for two years, and AI doesn't resolve either one. It just runs the broken process faster, at more volume, with more confidence, in front of more people. The bottleneck was always going to cost you. AI just moves up the invoice date.
This is the part most of the current AI discourse skips. The framing is almost always: deploy this, and the friction goes away. The more honest framing is: deploy this, and whatever friction you didn't already resolve gets exposed faster and harder than it would have on its own.
The Lesson Was Always Going to Arrive. Now It Arrives Sooner.
Here's the part that actually makes this useful instead of just cautionary.
In a slower system, an unresolved bottleneck can hide for a long time. The process conflict grinds along, the human issue stays unspoken, and the cost accumulates quietly enough that nobody's forced to act. Organizations can run on this kind of low-grade dysfunction for years.
AI collapses that timeline. Because it executes faster and touches more of the process at once, the failure that used to take a fiscal year to become undeniable now takes a quarter. Sometimes a sprint. The lesson that the organization was eventually going to have to learn — that this process conflicts with that one, that this person has been the actual constraint the whole time — gets forced into the open on a much shorter clock.
That's not a flaw in the technology. That's the technology doing exactly what speed does: it takes lessons that used to arrive the hard way, over years, and makes them arrive the hard way, over weeks.
And lessons learned the hard way flow up. They don't stay contained at the level where the failure happened. A process conflict that used to be absorbable at the team level, quietly, over enough time that nobody upstream noticed, now surfaces fast enough and visibly enough that it has to be escalated. Someone above the team has to see it, name it, and decide whether to fix the process or keep tolerating the delay. AI doesn't make that decision easier. It just makes it impossible to keep postponing quietly.
The Actual Choice in Front of You
This is why "adopt AI" was never the real strategic question. The real question is the one from the last piece: have you already done the fixed-cost work — resolved the process conflict, forced the human decision that's been sitting there — or are you about to find out, faster than you expected, that you hadn't?
Organizations that clear both hurdles before they scale AI into a process get the compounding gain the tools were promised to deliver. Organizations that don't get the same lesson everyone eventually gets — they just get it on a shorter clock, in front of more people, with less time to absorb it quietly.
The S-curve was never going to let anyone skip that step. AI just changed how fast you find out you didn't.
Herbert Roberts, P.E. is a licensed professional engineer with 30+ years in aviation research and development across two companies, and has spent eight years analyzing accidents for attorneys under his P.E. license.