Vibrio Is the Result, Not the Cause: Why AI Won’t Fix a Broken Process

A farmer calls, panicked. His shrimp are dying by the thousands. He’s already had the water tested and he has his answer: Vibrio. A bacterial infection. Everyone in the industry knows the drill — you have Vibrio, so you kill the Vibrio, and the problem goes away.

So he treats it. And the shrimp keep dying. He treats it harder. They keep dying. Because here is the thing thirty years of farming taught me the expensive way: Vibrio is not the cause. It is the result.That bacteria is everywhere, all the time. It only takes hold in a system that’s already stressed — bad water, crowded ponds, weak husbandry. The dying shrimp weren’t sick because of Vibrio. Vibrio was just the loudest symptom of a pond that was already failing upstream.

“We need a chatbot” is you killing the Vibrio

I hear the business version of this call constantly. “Our support is drowning — we need an AI chatbot.” “Sales is slow — we need an AI tool.” The owner has already run the test and reached the diagnosis, and the diagnosis is always the same shiny bacteria: buy the AI, kill the problem, move on.

But the drowning support queue is Vibrio. It’s the result. The cause is usually upstream and boring: a product that confuses people, a returns process that generates the tickets, a knowledge base nobody maintains. Bolt a chatbot onto that and you’ve automated the confusion — faster wrong answers, at scale. You killed the bacteria. The pond is still failing.

Diagnose before you prescribe

The operator’s move is unglamorous: before I recommend a single tool, I go find the actual cause. Where does the work really pile up, and why? What generates the tickets in the first place? Is the goal even clear enough to automate against? Nine times out of ten the highest-return fix isn’t AI at all — it’s the upstream thing everyone routed around because the symptom was louder than the source.

This is the same lesson I learned spending over a million dollars fighting the wrong problem on a shrimp farm, and it’s why finding your first AI use case starts with the process, never the tool. Get the diagnosis right and the treatment is often cheap. Get it wrong and no amount of AI budget saves you.

The tool is the last question, not the first

AI is real, and used right it moves numbers that matter. But it is a treatment, not a diagnosis. The companies that waste the most money are the ones so sure they know what they have that they never ask what’s actually killing the shrimp. Before you spend a dollar on the cure, be honest about the cause. That’s the whole difference between AI that pays for itself and AI that just makes an already-broken process fail faster.