Giving AI Street Smarts

“Alexa, is the store closed?”

“Alexa, does the store have clothes?”

When spoken aloud, these questions sound very similar. The difference is obvious to a human because of context, but to a voice-activated assistant, the distinction is a fine line and may result in some confusion.

On the human side of the conversation, frustration ensues — and among millennials, it can turn into “Bye, Felipe,” as the customer churns out of whatever business created the bad experience.

Which means that, as IntraNext CEO Patrick Brown previously told PYMNTS, voice-activated assistants powered by artificial intelligence (AI) won’t be replacing call center agents anytime soon. The risk is still too high: Implementing AI poorly can cost a business more than the implementation itself was worth.

However, AI may soon be working alongside human agents, Brown said. In a recent interview with Karen Webster, Brown predicted where AI will gain its first footholds and how. Together, the two envisioned what its role could look like in the more distant future.


Brown believes that the two types of companies most likely to become early adopters of AI are ones in highly specialized fields and ones with “low-hanging fruit” that would be easy to automate.

Specialized fields with their own unique jargon present, perhaps counterintuitively, a lesser challenge, if you ask Brown. Because of that jargon, there are fewer linguistic variables and fewer opportunities for confusion.

To use Webster’s example, the phrase “double shot” could confuse a more generic AI, but a specifically tailored barista bot would know that the customer was trying to order two shots of espresso — not two shots of liquor and not a basketball arcade system for their man cave.

As for the “low-hanging fruit,” Brown suggested that there are scenarios where existing AI is good enough to solve problems for customers without creating further frustration. In high-volume situations, it could vastly improve efficiency by taking rote tasks off employees’ plates.

For instance, if a customer calls tech support, Brown said it would not be difficult to create a bot that could walk that customer through a checklist of options: Is the computer plugged in? Is it turned on? Only callers who made it through the checklist and still needed help would be handed off to a live agent.

Brown explained that this would free up agents to address more complicated issues to deliver that last-mile, personalized tech support in situations where it’s truly needed.


It’s one thing to automate frequently asked questions; but what about when things get more personal and a user must provide identifying information to authenticate the session?

What about, wondered Webster, conversations that end with a purchase — say, if a customer calls his insurance company to change his policy and needs to pay the difference? How low does the tokenization and credit card data storage fruit hang?

Brown said it’s not out of reach. But before picking it, organizations must be confident in the technology that will keep that data secure. Only then can they be confident in handing over the task to AI, he said.

And, of course, the technology will have to get past the “closed/clothes” roadblock before consumers and organizations will trust it with anything more sensitive than “Is your computer plugged in? Yes/No.”


To use America’s favorite eCommerce giant as an example, imagine having Amazon’s Alexa call an airline for you to book a flight, or calling your insurance company to adjust your policy, or calling your cable company to negotiate a better monthly rate — a dreaded call that most Americans make multiple times per year.

This isn’t an instance of asking Alexa to dial and then conducting the call yourself. Imagine Alexa handling all the nitty-gritty parts — in particular, authentication and payments. Imagine that she’s the one haggling with the cable company, then delivering the lower payment once negotiations are settled.

Why not? More Americans have an Amazon account than not. That account contains credentials that the user can tap for faster checkout when shopping online. If voice commerce is the next frontier, then it seems almost inevitable that Alexa will one day learn to do this. Her usefulness won’t stop at ordering a pizza or a new pair of shoes by voice.

Brown said the key is going to be baby steps. Introducing palatable amounts of AI technology across channels will help average consumers get their feet wet to the idea without feeling they’ve been unceremoniously thrown into the deep end.

For example, said Brown, “A baby step might be where Alexa attempts to place an order, but when the order can’t be fulfilled, Alexa is able to navigate the interaction and, through intelligent call delivery, connect the customer to the appropriate agent to discuss next steps.”

His advice? Not every company needs to be at the “bleeding edge” of technology. Because the thing about the bleeding edge is that it bleeds customers. Ultimately, said Brown, businesses that are worried about churn and about keeping happy customers will want to ensure they invest in the right partnerships with the right entities rather than chasing the latest and greatest.

“You’ll need to adopt sooner rather than later,” said Brown, “but you better do it very methodically.”