You’ve probably heard that AI is taking over the world. Usually, the first thing that comes to mind is Apple’s Siri, Amazon’s Alexa, Windows’ Cortana, or chatbots on customer care support, but AI is actually more extensive than that. These are the ones customizing your user experience on Netflix, sorting the emails you get, and creating custom playlists for you on Spotify.

And then, there are chatbots who talk to you on websites.

Every now and then, people working in customer support have to hear about how chatbots will someday take their jobs. Yet according to customer service experts, currently only 10 to 35 per cent of inquiries are resolved by chatbots. That can easily change in, give or take, a few years. Here are reasons customer support needs humans, not chatbots.

Chatbots are used for inquiries. In the case of customer support, plenty of queries can be complex. Bots can only answer to specific types of questions and offer limited answers. Kai Stinchcombe, CEO of TrueLink Financial, shared his experiences with Uber’s AI customer service rep. He booked an UberX instead of an Uberpool, saying he couldn’t afford extra stops. He ended up being thirty minutes late for his meeting, and $14 instead of just $4.

Stinchcombe contacted Uber, to which Uber’s AI could only give the same formatted replies. In his blog post, Stinchcombe concluded AI’s limitations: the inability to solve unexpected problems.

The ability to solve unexpected issues is the whole reason customer support even exists. People are experts, and not programmed to think in algorithms. We know how to navigate issues that come without warning.

Late last year, an AI called AlphaZero learned to play chess and eventually beat a chess grandmaster. But that level of reasoning hasn’t reached chatbots for customer service. Similar to the first reason, part of why AI can only give limited responses is because they fail to understand a customer’s concern the way another human would.

When your customer service provider is another person, you have more freedom to ask about your concerns without tailoring your question to fit a chatbot’s possible responses.

If you’ve ever asked Siri questions, you’ll find that Siri only responds to one request at a time, and does not recognize previous requests or queries. When chatbots fail to understand concerns, supporting statements may not help entirely.

Chatbots are also known to be confused by people’s requests. Going back to the Uber case, as Stinchcombe kept explaining his concerns to the chatbot, it generated the same replies without giving a concrete solution.

On the other hand, had Uber’s customer support provider been a human being, he or she would have handled the issue better, and maybe come to an arrangement that benefits both parties.

Admittedly, the shortcomings of AI can be attributed to them being in their early stages. However, when it comes to customer support, chatbots can only solve a limited number of problems (even though they’re programmed to be all-knowing), show partial retention, and are unable to understand real concerns. When it comes to customer care support, humans can get the best help from humans as well.

Report this Page