BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

How IBM Watson Is Poised To Transform Customer Service And Hospitality Via AI

This article is more than 8 years old.

Artificial intelligence (AI) from IBM's Watson (of Jeopardy-winning fame) is about to transform customer service, the customer experience, and hospitality.

Customers today–especially but not exclusively Millennials–are embracing, even demanding, what I call a “Jetsonian” (for The Jetsons) customer experience, a customer experience in which those tasks that don’t require human intervention are powered by automation and algorithms, while those tasks that do benefit from human beings get true human warmth, from capable human employees, applied to them.

Along these lines, meet "Ivy" from Go Moment, an artificial intelligence-driven system capable of handling, in real time, some 90% of the requests that come in from guests on property, without, according to its inventors, requiring a lick of human intervention.  Ivy is powered by IBM’s Watson, the world-famous AI engine.  Ivy uses this Watson engine to decipher and fulfill a wide variety of requests from hotel guests (it's being first introduced in hospitality environments) in real time.

Ivy is the brainchild of Raj Singh, President and CEO of its parent company, Go Moment. Singh is both a technologist and a kid-emeritus who grew up working hands-on in his family’s hotels. Coming from this hospitality background, Singh tells me, made him aware of the detailed knowledge that would be required to ever automate the challenges that come up in hospitality, and the nuanced approach that would be required to do so in a manner that would draw guests closer to a hospitality brand rather than putting them off with a stilted, tone-deaf set of responses.

To pull this off, Singh knew he would need some powerful help. So when he heard that IBM had allocated funding for Watson-related startups–“$100 million for venture investments to support startups and businesses committed to leading a specific industry with cognitive applications powered by Watson,” according to Lauri Saft, Vice President, IBM Watson Ecosystem & Partner Programs—Singh jumped at the opportunity.

The resulting Watson-Ivy collaboration plays out as follows, starting at check-in, where the hotel’s (human) front desk attendant invites the guest to provide their cellphone number for the program, if it’s not already in their reservation profile. If you opt in, says Singh, “you get a little welcome message when you get to your room that says, ‘Micah, welcome to the W Los Angeles. How do you like your room so far?’ You can reply, ‘five stars, I love the view. Can I get a bottle of red wine? And I could use a few more hand towels.’ Each request will be interpreted and dispatched to the correct department automatically using Watson categorization.”

You can also–and the following colorful examples are obviously mine, not Singh’s–text ‘Wassup, People?! I asked for a king and got to my room and there’s two queen beds instead’ or ‘If I had a burning desire to have my ear as close as possible to the ice machine, I would have indicated that when I checked in–can I please have a quieter room?’ and Ivy will reply ‘I’m very sorry to hear that; a bellman will soon be at your door to take you to the correct room.’ (Note: If you say ‘I freaking hate this place; beam me out of here,’ or otherwise express any kind of upset, Ivy bows out and a human is immediately dispatched, as I’ll discuss below.)

Ivy is frequently called on to deliver instant answers to knowledge-based questions. For example, if a guest texts “what is the WiFi password?,” it will deliver an answer instantly, with no humans needing to be involved. And when the “reply” needed is a task rather than information, that’s in Ivy’s skill set as well. In the example of asking for extra hand towels, Watson is able to route that request directly to the laundry, for faster and better results. “That's what the Ivy platform does: it streamlines operations and automates most decisions that previously required distracting the front desk from the guest waiting to check in.” This way, says Singh, “the guest is always the center of attention and always has all of the hotel’s resources at your disposal. Now your request goes directly to the right department in a fraction of the time it would take the front desk to manually receive, log, and relay your request.”

Watson, you may remember, is essentially a “question-answering engine,” initially developed by IBM to be able to decode the many ways a question can be asked on Jeopardy in essentially zero time–and get to that buzzer first! The answers, in other words, are not really the trick to Watson; any encyclopedia can do that part of the trick; it’s the questions that matter.  So while it might not seem like all of Watson’s smarts are required to answer something like “what’s the Wifi passcode?” the power of Watson is really coming in to play here to decode the many, many ways that a guest might make such a request: fifteen different people might ask for the same thing in fifteen different ways: “What’s the wireless passcode?”  “How do I get on the internet? And so forth. Watson’s involvement here is in detecting the intention behind the words. This keeps things effortless for the guest with Watson allowing the Ivy text responses to adapt to the guest, rather than the other way around.

Creating Ivy has required its creators to get down and dirty with the particular issues that come up at each particular client hotel. In doing so, they’ve found that there are about 1400 issues that can come up, barring an outlier event such as a fire. According to Singh, more than half of these don't require human judgment on the other side, and that half account for a significant majority of the interactions that come up.

Of course, the issues that don’t make up this “significant majority not requiring human judgment” should be of great concern to anyone considering deploying such a system.  Because here’s the rub: If an organization blows it on the items that do require true service recovery, it just doesn’t matter how many mundane issues were correctly handled; that latter, emotionally-charged failure is what customers are going to remember.

So a concern I had when exploring Ivy–and that you should have before putting laying down your money for such a system–was whether Ivy’s ability to answer the somewhat mundane majority of what comes up would cause its developers to develop hubris and inappropriately deploy Ivy to try to act as if she/he were human, with potentially guest-alienating results. So I wanted to find out how, and whether, “she” knows to escalate the text received that require or would benefit from a human response.

Here, I’ve been so far reassured by the logic behind Ivy. Ivy is programmed to escalate any expressions of dissatisfaction to humans, according to Singh: “The first question that Ivy always asks the guest is ‘How's everything going?’  If any negative words are detected in that response or if the person says, one star, my sheets are dirty, my toilet's overflowing, and all of those kind of common issues–we’ve got hundreds of them cataloged for every single hotel that is on our portfolio–we escalate any negative experiences from the customer instantly to the front desk and then track how quickly the desk agent follows up with them. So let's say you express that you’re having a bad experience: let’s say you type ‘my sheets are dirty.’ Ivy knows what all of us in the hospitality business know: that ‘dirty’ is a really bad word. So we send that off to the front desk for personal attention. From the time the staff gets a text message and email alert, they usually have twenty minutes to resolve the issue before Ivy escalates the issue to the general manager.”

(Ivy will also dispatch a human when Watson doesn't have a high enough confidence –98% or higher–of the meaning of a question or of how to resolve it, for example if a guest says something like ‘my internet is broken.’ From this, Ivy can detect that the guest is talking about the internet, but doesn’t have a confident way of resolving this issue of the internet being broken, because Watson may not be able to understand what ‘broken’ means in that context. In such a case, Singh says, “Ivy uses an algorithm to decide whether or not it should be dispatched to a human.”)

Taking the guest’s temperature in real time (and before they get on TripAdvisor!)

One promising way to think of systems such as Ivy is as “real-time satisfaction monitoring.”  (Or, if you prefer, to think of it as “pre-TripAdvisor satisfaction monitoring.) Either way, it’s a powerful concept. You can think of it as having a sensor attached to every single guest that blinks red every time there's a bad experience–so you know right away to care of it. It’s a lot less spooky for a hotelier this way: previously, that in-room experience was pretty opaque: you know everyone's in their rooms, but you don’t know how they're feeling; you end up waiting ‘til checkout to ask for the guest's feedback– and the response rate you is in the dismal, 2-3% range. (And, by a staffer reviews this feedback, the guest is gone and may be publicly sharing their experience on social media.)

A system like Ivy, on the other hand, is designed to give hoteliers ten times as much feedback, and get it to them a whole lot earlier: a hotel will get feedback from 30% of their guests in as little as 20 minutes after check in. So if the guest encountered dirty sheets, Ivy can alert staff in real-time and give you a better chance of ensuring that every guest leaves with a smile on their face– a long time before they ever get to TripAdvisor and have the chance (and inclination) to skewer you publicly. In fact, Go Moment provided me with data (I have not verified this, so take it for what it’s worth) suggesting that Ivy has assisted one of its clients, The Declan Suites San Diego, in rising thirty-five ranks on TripAdvisor–an increase in TripAdvisor visibility similar to what you’d expect from undertaking a multi-million-dollar renovation.

The possibilities outside of hospitality are also intriguing  For example, Singh (as well, quite likely, as some other API developers with whom IBM Cognitive has been similarly collaborating) see opportunities almost everywhere.  In Go Moment’s case, according to Singh, “We've already been approached by the world's largest airlines, airports, and retail brands, who are interested in bringing the Ivy functionality to their vertical.

Micah Solomon is a customer experience consultant, customer service speaker, trainer, and bestselling author on customer service, hospitality, and the customer experience.