Social media has undoubtedly changed the face of marketing. Where we used to rely on product placement, focus groups, and household polls to determine the who, what, and where of sales, we now rely on machines programmed to learn the exact same information in a much shorter amount of time. With machine learning (ML), algorithms programmed to locate and target your product’s best audiences make light work of what used to take a team of people many days to hack.

It’s a marketing dream. But, that doesn’t mean it doesn’t have its pitfalls or problems. Like any good tool, it requires proper handling to be most effectively used. And, also like any good tool, improper handling can lead to disaster–or at the very least, accumulation of rust and grime.

The misguided allure of ML is that you can simply set your course and hit snooze. On the surface, it would seem that’s the case–the algorithms are programmed to do everything from selecting the best creative to use, determining the right and wrong times to push ads, and locating the best communities to sell products to.

And they do, in fact, do that.

However, the hidden magic of a successful ML marketing campaign is in the utilization of all the data the machine has learned. If ML is the tool, then interpreting the data and strategizing creative solutions based on that data is the proper use of said tool.

To look at it another way, let’s say that there’s a ship called ML. Without a captain and the appropriate crew, a ship out at sea ends up wherever the wind takes it. The marketing machine is no different: on any given day, the tides can change. The trends can shift the winds unexpectedly, and the currents of approval can begin to form new and unprecedented eddies. Without a skilled captain and crew, there’s no way to get an ML marketing campaign back in friendly waters once the tides have changed (or to keep it from entering unfriendly waters in the first place!).

As much as it may seem like it’s possible to simply set the ship’s course and let it coast, the human consumer is fickle (perhaps even more so than ocean currents, which at least have a discernable pattern!) and as such, every bit of information learned by the social media advertising algorithms can and should be analyzed by experts versed in the data created by the human consumer.

Despite how it may appear on the surface, ML is not intended to be left alone: it is a tool intended for use by skilled people. It was created to inform us of the daily tweaks, adjustments, stratagems, and creative solutions necessary to ensure the success of any given marketing campaign.

Put more simply, machine learning is a marketing dream when and if there’s a qualified team of media buyers manning the ship.