In 2014 Mike Seay received a promotional email from office supplies company OfficeMax. The letter was addressed to "Mike Seay, Daughter Killed in Car Crash." In fact, Seay's 17-year-old daughter had died in a car crash the previous year. The marketing misstep had catastrophic brand consequences for OfficeMax. Dozens of media outlets, from Forbes to the Wall Street Journal to the Los Angeles Times to local television stations covered the mix up, which OfficeMax blamed on a data broker for merging the wrong information fields.
Intelligence, and more specifically artificial intelligence, will disrupt many industries and ways we do business. For example, old ways of doing things, such as the venerable CAPTCHA online security tool, are already close to being blown up by AI systems that learn on their own. Relatively inefficient and messy, the marketing field seems ripe for AI disruption. But technology, left to its own devices, can make a mess quickly.
The OfficeMax incident is a stark reminder of what can go wrong when humans are completely removed or unable to oversee marketing processes or any other process that pushes information into the hands of potential users or consumers. Google also learned this the hard way when it's automated image identification system started labeling African-Americans as gorillas. And Vanguard, the highly respected and decidedly apolitical brokerage, was shocked to find that AI-driven advertising networks had placed its ads on strongly conservative news site Breitbart and other highly politicized online destinations.
The end goal is understandable and even laudable. Systems will automate marketing to better target messages and better interest potential customers. Marketing-centric artificial intelligence tools aimed at marketers will drive lightning-fast decisions based on algorithm. Those decision processes are designed to be several steps removed from humans.
A lot of the time, that's just fine. Make no mistake -- AI systems can see patterns that humans cannot easily see. They can tell us what specific terms work better, and how to get a 5% increase in traffic or response through multi-variate testing and A/B tests on just about anything that you can digitize.
If you are ready to go down that route and cede full control of your brand and your message to AI, then you probably should also think through what can go wrong in order to plan for disaster. In the case of Google, before releasing an image recognition tool to the world, it probably would have made more sense to run a widescale private beta (maybe they did this - but it didn't seem to work). In the case of OfficeMax, a simple data merge caused the problem but for a not considerable sum of money the company could have run its lists by human QA checkers in Asia or set up some jobs on Amazon's Mechanical Turk to at least screen for any potential problems.
The reality that all marketers must reconcile to and understand is simple. Most AI lacks nearly all common sense, and that includes a sense of morals, societal norms and decency. So trusting AI to talk to your customers and your users unreservedly is the equivalent of allowing an amoral actor to take over your marketing department. Be somewhat afraid. But more importantly, be ready.