What self-checkout failures can teach us about artificial intelligence

Bob Sullivan

We all want tech to solve the world’s most vexing problems, and we’ve all spent time with tech that hasn’t lived up to its promises. The printer that lets us down as we’re rushing to a meeting. The concert ticket app that suddenly fails as we’re trying to get into a show. The helpful website customer chat tool that only knows how to say, “I’m sorry you’re having trouble.”  Tech might feel like the solution to every challenge; clearly, it’s not.  And it often creates as many problems as it solves.

And that takes me to self-checkout lines at retail stores.  We’ve all seen them work; we’ve all seen them fail.  I want people to remember this as we race headlong into the AI world.

There have been a flurry of articles recently about the failed experiment of self-checkout. How big stores are backsliding on kiosk installations. How consumers have come to realize they are doing extra work, often enduring extra frustration, and not getting anything for their trouble. How lack of human beings at store exists contribute to shoplifting.

People in the industry say these comments are overblown. But there is no denying; self-checkout is often a headache. And it gets no beta-test bye from me. We’ve been forcing people to find the code for green peppers (organic or not?) since the 1990s.  That’s a generation of shoppers’ time to work out the kinks.

I know many of you love running into a drugstore and running out with one tube of toothpaste without having to wait for a store clerk to take your money and make your change.  As usual, my criticism is not sweeping or all-inclusive.  Quite the contrary. I want people to remember that new technologies should be implemented without forcing people to abandon old ways.  The solutions tech offers are never all-inclusive.  There are compromises to be made. There are exceptions to be handled. The obvious solution for stores is to have a better mix of human clerks and self-checkout machines.

Unfortunately, that mix is hampered by the Trojan horse fighting all good technology implementations — above all, cost cutting. See, self-checkout is really about minimizing minimum wage employees, not getting the best out of a new human invention. In a capitalist society, that pressure will always exist. But even here, this approach is short-sighted and doesn’t pencil out.  There is evidence that labor cost savings from self-checkout are erased by increased shoplifting.  And we’ve all seen stores that end up hiring nearly as many self-checkout hand-holders as they had cashiers.

So when we introduce new tech, blended, human-friendly solutions should be the first impulse, not a last resort. After 9/11, there was so much talk (and snake oil) around facial recognition.   Most of those implementations failed spectacularly, and we found that talented, attentive security guards are better at keeping us safe.  Slowly, airport security tech has gotten better, and it can now stand alongside human talent to protect our flights.  But any rush to find a gadget cure-all is doomed to fail.

Here’s another example I’ve thought about often. Not long ago, many believed virtual reality would soon be ubiquitous.  Today, you barely hear the term. Creating a fully immersive experience that fools the mind turns out to be incredibly hard. But augmented reality — where repair instructions project over a technician’s genuine eyesight as they try to complete a tricky repair — has great potential.  In a similar way, self-driving cars are a pipe dream.  But driver assistance technology is here today, works incredibly well, and holds the promise of soon dramatically reducing car accidents.

So as artificial intelligence continues to creep into our offices and our consumer products, let’s make sure people are leading the way, not forced to follow it. Otherwise, we’ll end up like so many consumers, standing in line, waiting for someone to come verify our driver’s license so we can buy cooking wine.

 

Leave a Reply

Your email address will not be published. Required fields are marked *