Purpose, potential and pitfalls of customer-facing speech AI

In 2018, Google CEO Sundar Pichai said demonstrated the Google Duplex assistant at the company’s developer conference. The assistant mimicked realistic and nuanced human speech patterns (complete with “ums” and “ahhs”) as he made a haircut appointment and reserved a table at a restaurant while in fluent conversation with a real person.

Though audiences in the Twittersphere and beyond erupted into rapturous applause, observers were quick to question what they were hearing.

Some called the resemblance “scary”, and others felt like one there was cheating involved – with the human on the other end of the line totally unaware they were talking to a bot.

In the end, the entire episode wasn’t great PR for artificial intelligence or advanced speech technology. But that’s a shame, because the truth is that voice AI has enormous potential to empower consumers and deliver value to the companies deploying it — provided there’s a clear understanding of its purpose and limitations.

Voice AI in the wild

One of the best examples is ordering food.

Skyrocketing inflation has increased costs for restaurant owners, while labor shortages have made it difficult for them to meet customer demand (which decrease slowly after–lockdown). Some smaller restaurants have been ringing the phone, while some larger ones have even been forced to keep driving customers waitingwhich leads to frustration.

So they are increasingly turning to voice technology to pick up the slack.

It makes perfect sense. As long as the speech technology is advanced enough – and you might be surprised how smart it is right now — by letting voice AI take an order, employees can get on with the important work of making great food and making sure the diners have a great experience.

No one is fooled in this scenario – this kind of voice AI tends to declare its non-human status when it’s not already clear. Customers are satisfied and service professionals are supported, not undermined.

Good service, no servants

So what about this idea: instead of each of us having our own personal humanoid Jeeves (as in the Google Duplex scenario), what if different brands and companies had their own assistants that formed a broad ecosystem of voice assistants? In this way, companies could assert their own brand identity and build one-to-one relationships with their customers without an intermediary. Customers for their part could deal with a voice AI that really know the goods or services the company has to offer, rather than an Alexa-style assistant trying to work its way through.

For example, the voice assistants of restaurants become familiar with the menu. They learn favorite combinations; they can make changes and suggestions; they learn to upsell. Why couldn’t that be replicated in the rest of the hospitality industry, retail or even professional services? The answer is: it could, and it is starting to happen.

Instead of thinking about creating sentient AI servants, we should start thinking about voice assistants as functional tools that we can refocus in this way. In the “real world,” most of us don’t have servants or envoys to negotiate for us, but we do rely on knowledgeable, pleasant, and efficient front-line workers. Why not replicate systems that work instead of legacy systems?

I believe we’re going to do that and it will make the experiences of both brands and customers more vibrant and fruitful as a result.

Importantly, this is not about replacing staff with an army of voice assistants. It’s about giving employees the time and space they need to focus on critical tasks, streamline clunky ordering systems, and help businesses grow revenue. And it’s also about allowing us as customers to move away from screens and devices to order in the most natural and human way we know how: with our voice.

Zubin Irani is CRO of SoundHound.