• by Ted McConnell , Featured Contributor, January 3, 2019

One application of AI, customer service, is so compelling that it has become core to strategy for some very large enterprises.  But relegating your reputation to a  persona with an EQ of zero certainly carries new kinds of risks. 

AI has become synonymous with service in the minds of many. The idea, generally, is that AI can replace human service representatives, or better, improve on existing technology. This was also the idea for interactive voice response systems (IVR), and every other technology that inserted a computer between a corporation and its customers. 

The base-case financials are compelling. Connect an AI-driven dialogue to an app, and what you get is the opportunity to build scale behind friendly, easy-to-use service.  You would think the net result would be to vastly improve customer service — but if history is any indication of the future, maybe not. 

Today, most of us dread working with these systems, and for good reason. 

How many times have our sensibilities been assaulted by machine-delivered messages of caring? The machine says, “Thank you,” but we know the machine is not thankful. How many times have we heard “we are currently serving other customers,” directing us to yet another machine dialogue offering no service, but rather, alternative ways to wait? Or worse, getting a pitch while on hold. 

It’s hard to love a brand that uses marginal service as a smokescreen for a heapin’ helping of CRM. For example, I recently spent hundreds of dollars at Whole Foods, scanning my new Amazon/Whole Foods app, and getting only 30 cents in discounts. But Amazon got all my shopping behaviors linked to a mobile ID. Fair trade? 

We all know brands will blithely obliterate hours of our time to save themselves a few minutes. They do this while regaling us with messages touting their own moral high ground. 

Now, raise your hand if you think anyone is blind to this hypocrisy. As consumers, how do we process this?

Sadly, in today’s world, it seems like just another small insult against which we are helpless. Delta thanks me for “choosing Delta,” even though it’s obvious that less choice means higher profits for them. 

The outcome for brands is that businesses do not fail immediately, but customer anger festers. Do cable companies not see the connection between cord-cutting and the service experience? Do retailers see the connection between long lines at checkout and losing share to Amazon? 

Any company that compromises service might heed the advice in Lowell George’s great lyric: “The same dudes you misuse on your way up, you might meet up, on your way down.” 

A Barbarian at your Gate

We are about to graduate to a world where a deceptively friendly silicon construct will be representing your company to the public. 

Since AI can make each interaction unique, you might not know what your representative is saying to people, case by case. "Ex Machina" writ small: The more human it seems, the more shocked we feel when it betrays us. 

A persona will be sitting on the company boundary. Will its character reflect your brand character? Will it be chatty, or terse? More importantly, will it be able to solve the problem at hand? Will it represent your brand nicely — or will it, and your brand, become the subject of viral mockery when it says something stupid? 

Almost 20 years ago, P&G created an AI character as an experiment: Mr. Clean!  Hundreds of contingencies had to be handled. For example, when asked if he was gay, Mr. Clean did not want to alienate anyone, so he said, “Let’s just say I’m a confirmed bachelor.” Those bits of genius were creative decisions made by a mathematician with a Ph.D.   

Who in your organization is qualified to make these decisions?  How will your brand character become correctly manifest in a programmatic dialogue? 

In short, “How may I help you?” is an increasingly risky question, because expectations about the answer will be conversational, personalized, and unconstrained. 

Try this easy test. You say, “Hi, how may we help you today?” Customer screams, “Screw you! My freaking computer is broken and you assholes won’t help me!” 

Your move. How does your AI respond, and why?