MIND & METRIC MEDIA

FINDING 001

The Chimpanzee With a Machine Gun

Every ad platform now sells automation as strategy. The pitch is partly true. That is exactly what makes it dangerous.

Illustrative render of a chimpanzee seated in a server room holding a rifle, surrounded by data screens — visual metaphor for AI without judgment

Hand a chimpanzee a machine gun and you have not created a soldier. You have created a faster way for things to go wrong in every direction at once. The gun performs exactly as engineered. That is the problem.

Paid media is having its machine-gun moment. Every platform now sells automation as strategy: Smart Bidding, Performance Max, Advantage+, campaigns that write their own ads and pick their own audiences. The pitch is always the same. Hand over the keys, the machine drives better than you do. And the pitch is partly true, which is what makes it dangerous.

What the machine actually optimises

An algorithm optimises the objective it is given, using the data it can see. Nothing more. A target-ROAS bidder does not know that branded clicks were coming anyway; it only knows they convert cheaply, so it buys more of them. A conversion maximiser cannot tell a new customer from one who was already halfway through checkout. And the system that reports the results is the same system that spends the money. The machine grades its own homework.

None of this is malfunction. Each system does precisely what it was built to do. The failure happens one level up, where nobody decided what the machine should want.

The quiet failure mode

The visible disasters are rare. The expensive failures are quiet: budget drifting toward remarketing and brand terms because they score best; broad match wandering into cheap, irrelevant intent; a dashboard reporting a record quarter while a switch-off test would show revenue barely moving. The account looks optimised. The business is buying customers it already had.

The judgment layer

The operator's job has not disappeared. It has moved up a level. Someone still has to choose objectives that map to business outcomes rather than platform metrics. Someone has to structure accounts so the signal stays clean, decide what automation may not touch, and run the tests no algorithm will ever volunteer for. No machine has ever suggested switching itself off to see what happens.

This is not an argument against AI. Used with judgment, automation is leverage: it executes faster and at a scale no human team can match. The point is narrower and harder to dodge. The tool amplifies the judgment behind it, and where there is no judgment, it amplifies noise with perfect efficiency.

The chimpanzee was never the villain. The villain is whoever handed over the gun and left the room. Before trusting any system that spends money on your behalf, human or machine, there is one question worth asking — the same question this agency starts every engagement with. What would happen if we didn't spend it?

Wondering what your own machine is actually buying?

The first conversation is diagnostic, not commercial.