Skift Take
Hotels can now create their own brand-specific digital assistants with IBM's new Watson Assistant because they can white label the artificial intelligence platform. Hotel operators also keep their own data, versus giving it away to a company like Amazon, which IBM says provides a new level of brand intimacy and trust.
This sponsored content was created in collaboration with a Skift partner.
IBM launched Watson Assistant this month at its annual IBM Think conference in Las Vegas, designed to bring enterprise-grade artificial intelligence to its partners, with an early emphasis focusing on the hospitality and automotive industries.
The rise of voice and chat platforms, including Amazon’s Alexa, Apple’s Siri, Microsoft’s Cortana, and Google Assistant, are reshaping how we interact with machines in our personal residences, workplaces, and everywhere else in the world. These AI-powered digital assistants are also driving the future of conversational commerce to expand how brands connect with consumers on a more personalized, individual level.
Watson Assistant has now joined their ranks, so what does that mean for the hotel sector and the guest user experience?
IBM is proposing five differentiators separating Watson Assistant from others, especially Alexa. In the last year, a growing range of hotel operators have been introducing Amazon’s platform into their properties to answer the demand for smarter hotel rooms.
First, Watson Assistant is enterprise-level artificial intelligence versus consumer. It’s not designed for your fireplace mantel or kitchen counter or office desk. It was developed to integrate with highly complex back-end systems, both at the property level and brand-wide.
Second, brands can white label Watson Assistant. That means hotel guests won’t be verbally talking or texting in chat to engage Watson specifically. They’ll be speaking directly to Hilton’s “Connie” robot, or The Cosmopolitan Las Vegas’ “Rose” chatbot, or whatever identity any brand comes up with for its consumer-facing digital assistant. Whereas, when guests today at The Westin Buffalo speak to an Amazon Echo, they’re speaking to Alexa (and Amazon) — not Westin.
Third, brands keep their data. With Watson Assistant, hotels are not sharing their first-party data with IBM, which is a major concern for all brands as the recent Facebook scandal shows. When hotels install Amazon Echo devices in their rooms, the data derived from guest engagement is captured by Amazon.
Next, due to IBM’s business model ensuring data privacy, hotels can more easily partner with companies throughout their location. That could ostensibly position big box brands more as destination portals to provide local travel experiences aligned with guest preferences.
Lastly, hotels can either integrate Watson Assistant voice functionality into their existing chatbots and apps, or they can build new voice and chat platforms on top of the Watson platform. Using The Cosmopolitan’s Rose bot as an example again, the hotel could work with Watson Assistant so guests can actually speak to the cheeky female digital concierge.
Here’s the IBM launch video announcing Watson Assistant.
[youtube https://www.youtube.com/watch?v=UkZJHVzVW-U?rel=0&w=560&h=315]
Skift Q&A: IBM Watson Assistant
We spoke with Bret Greenstein, VP, IBM Watson IoT, to dive deep into how Watson Assistant could potentially revolutionize the hotel guest experience.
Skift: What’s the big innovation here that differentiates Watson Assistant from other platforms, especially Amazon’s Alexa?
Bret Greenstein: I think it’s great that the market itself has kind of embraced the concept of digital assistants, but most of them have been aimed at what I would consider to be general knowledge, daily life, and home use cases. That’s made everyone comfortable with the idea of talking to their systems, to their machines, to their things in their homes.
We noticed very clearly the need for our clients to go deep, to be highly specific about what they integrate into their back-end systems. They need to have in-depth knowledge of the business process when they expose their customers to an assistant to elevate their customer service. There was also the need to integrate assistants with their own brand, and to do it with absolute data privacy.
We watched some of our own customers dabbling in some of the consumer AI, and trying to figure out what it would do for them. What they found is it didn’t really adapt to their business model or their brand experience. But, it did get people to begin asking some basic questions.
So we spent the last year and a half doing private pilots, such as the work at Thomas Jefferson Hospital [Philadelphia], for example, to put Watson in patients’ rooms to help them control the room settings and access help more conveniently. We did work with Local Motors in their self-driving car so they can provide an interface for the broader public. That helped us learn quite a bit around engagement and making user design fun, but also personalization, so we could relate to different passengers.
And then, it was us really just listening to clients in automotive, hospitality, and everybody who’s trying to figure out how they can extend their brands with a conversational interface, without giving away their brands. Meaning, how do they learn about their users so they can deliver better service, without giving away all those incredibly valuable insights about their users?
Skift: Can you explain in your words the difference between consumer AI and what we’re talking about here, enterprise AI?
Greenstein: Well, the methods and technologies don’t differ all that much. We’re talking about machine learning in AI, which are well established technologies. The difference is how you apply them. It’s the types of data and integrations you build into it, and the data privacy, and the enterprise characteristics.
For example, in a hotel, there’s a substantial challenge with having different people going through hotel rooms day after day. The account and all the information in the assistant is tied to you for this one day you’re in that hotel. But the next day, someone else is in that hotel. So, it’s very, very important that we can ensure the privacy and deliver the personalization element for the individual user, and then cache out and get rid of anything that shouldn’t be there for the next guest. Very, very important.
The other thing is, devices need to be managed devices. The thing you buy and stick in your living room is a device that just you manage. But when you put a thousand of them in a hotel, you’ve got to manage firmware updates and security, and make sure that they’re not moved, and make sure they’re all working well. So, there’s device management characteristics also that an enterprise or a hotel would expect that a home user might not.
Skift: You mentioned Local Motors in Phoenix, who we’re very familiar with. They partnered a couple years ago with IBM to integrate Watson into their autonomous Olli electric shuttle. That was one of the first use cases for a customized Watson assistant in the mobility space. What have you learned from that integration in the last year or so?
Greenstein: A couple things. One is that the interface and the conversational design has to be engaging and fun, and very much relevant to the kinds of things people are doing in that setting. So, we developed very specific content use cases for Watson in the shuttle that people want to know when they ride in public vehicles. We used a lot of crowdsourcing to figure out the kinds of questions people ask, and it became things like, “Am I going to be late?” and “Why are we slowing down?” or “Can we stop here?” And we found lots of ways people say those things with lots of different accents and ways of talking.
We also did a deep integration into the shuttle itself. Watson understands the data coming off the vehicle, and how, say, it’s slowing down because the roads are slippery. That happens in real time, and in context. I think what we got from that was, if you’re going to go into a self-driving car, you have to establish trust with the customer, because there’s no driver to yell at. So you have to trust the vehicle, which means you have to trust it understands you. By demonstrating that Watson understands you, and answers in context, it established a level of trust you might not find elsewhere, and that makes people much more comfortable.
Skift: So now you’re expanding on that learning and moving deeper specifically into hospitality and automotive. For hotel operators, how is Watson Assistant going to improve the guest experience?
Greenstein: It’s about the deep integration with the hotel brand and the hotel. So, when you go to a hotel, you should be able to walk into your room and ask, “Where is a great place to eat,” or “Where’s the gym and what hours is it open?” Just whatever matters to you when you’re there. The digital assistant should know the data coming off the hotel management system and the websites, and anything about the hotel, and all documentation. And that stuff should be presented back to you in as natural a way as possible, and you should have real answers, not canned. If the hours change, or if somebody is out to lunch and is not at the front desk, you should know that.
Skift: So, you’re saying, it’s all about programming the Watson Assistant platform specific to the hotel brand and whatever location it’s in, by feeding it all the data it needs from both the brand and the property?
Greenstein: We have a broad set of capabilities that all clients have, and then we have industry-specific content for hotels. First, Watson can be personalized specifically for a chain, whether it’s Hilton or Marriott or another brand, to integrate all of the brand-level content and data into the system.
For a specific hotel, Watson is deeply integrated with the room control systems, the hotel management systems, and the service request systems. We’ve built those integrations and we can integrate many of them now into a particular building and anything around it. We can go as deep as the hotel wants to go.
And then we also use location and weather, because we bought The Weather Company, to go very, very deep into weather insights. We can also integrate location and weather data with crowdsourced information from things like Yelp and TripAdvisor to provide more real-time, location-specific information to the guest. So, if you want to do something outside the hotel, we know what’s going on. We know where it’s busy, how to get somewhere, if it’s going to rain, where you’ll like to eat, and all the stuff you might need to get things done when you’re in a city.
Skift: Okay, let’s say I’m at the Hilton in Austin. It sounds like the people at that Austin hotel don’t have to update information themselves into the Watson Assistant platform because it’s integrated into the back-end systems. So, is it the same situation for brand-wide updates, as well, because Watson is connected to the chain’s marketing and distribution ecosystem?
Greenstein: So, the hospitality industry in general is quite complex. There are some chains where a central team can push out a standard that goes across all hotels. There are others where the operators have much more influence and control over what gets done locally. So, it’s a combination. I can’t say it’s the same every time. We work with some casinos, for example, that are very much focused on just that location, that building. And then there are chains that have a lot of standardization across their inventory. We do both models.
We also do quite a bit of work with partners that’s deeply integrated into retrofits and new construction. We’re setting the specifications and designs to be the standard for televisions, displays, mirrors, and things like that in hotels, so we can build Watson Assistant into those across the portfolio. We can build it, set it up, customize everything, and handle maintenance and operations. We can do the whole thing, along with clever training, for example, where we preload all of the staff names.
But we also work with partners, like any global or regional system integrator, or hotel. There are a lot of companies that do services and design around hotels, and any of them are able to integrate Watson Assistant. The hotels themselves can, as well, if they have an IT staff who wants to do the customization.
Skift: Is there a machine learning capability here where after, say, six months at a specific hotel, the Watson Assistant has learned to deliver more nuanced location-centric information in more sophisticated ways?
Greenstein: Yes, but think of it as assisted learning. All of these systems are best where the machine learning provides some feedback and input to people who can then tune and tailor the system based on their own judgment as well. So, the system does learn and it does capture, for example, when people have to ask three or four times for something, or if they ask the same thing every time.
There are multiple stages of learning. One of those stages of learning is personalization. So, because you always ask for Greek salad when you go somewhere, we know that your preference is Greek salad, and so the hotel knows that. Not IBM, but the hotel knows that. And then if you go to Watson and say you’re hungry, it could reply, “Would you like your usual?” That’s a degree of learning Watson can do at a personal level for the hotel chains, like learning your profile and your preferences better and better.
We also do some of the learning, or we’re constantly improving, for example, the recognition. We can do that across clients, and we can do that through crowdsource data. We will keep training the AI to be better at recognition and understanding and [hotel] domain knowledge. Then there’s also feedback. When people seem to be unhappy or frustrated with answers, there’s a whole dashboard analytics platform around that going back to whoever did the system integration, meaning the hotel or system integrator or IBM.
So, then we say, “Well, look, these are the kinds of things that aren’t working well. Maybe we need to improve the documentation behind it so it feeds better answers,” or things like that. That’s why it’s kind of a hybrid. There’s things that the system learns just through collecting a lot of data, and then there’s also assisted things where we’re getting feedback and then tuning the system to make it a better experience.
Skift: Okay, say I stayed at Hilton Austin and I engaged Watson Assistant. And then a month later I’m at a Hilton-operated property, of any flag, in Berlin. This might be an obvious question because everything is cloud-based, but the Watson Assistant at the Berlin property recognizes me, right?
Greenstein: Yes, we do this behind our client’s brand. I’m not saying Hilton is committed to do this, so we’re just using them as an example here. But Hilton has the ability to tie this to their profile and user account for you as a Hilton HHonors member, for example. They could create that seamless customer experience across all Hiltons if they choose to.
This is where some chains have more control than others over how things are used, but anyone who has any sort of loyalty program has a really strong opportunity to do the learning across their entire chain. Those same chains also have the ability to also link to their partners. They can choose to suggest to you, for example, to connect your Hertz rental car account to your hotel user account. So then, when your car is coming up to the hotel, the hotel knows it’s arriving. Watson can tell you where it is. Your favorite music is also playing in the car.
If you want to link your preferences between any preferred partners, Watson can do that. The user might decide to connect those two accounts together, because then both companies can give greater insights and greater service. So, there’s a lot of potential here for our clients to manage their own data to use across their own chain, as well as in their partner ecosystem, as long as the users opt in for that.
Skift: Let’s talk about data ownership because that’s such a hot topic today. It seems that one of the key differentiators for Watson Assistant is that brands can white label the AI. And at the same time, the brand keeps its own data, rather than IBM getting it. Can you expand on that?
Greenstein: Exactly. Yes, almost every client we have wants to work with us because they can extend their brand using Watson without their customers speaking specifically to Watson. If you’re talking to your BMW assistant in your car, it’s powered by Watson, and it’s the same exact Watson AI that could be powering something else. But in this case, it’s BMW’s artificial intelligence platform, and so you’re talking to BMW. You’re engaging directly with them, and all the data belongs to BMW.
The reason that’s so important is they’re establishing trust and credibility between themselves and you as a car owner or passenger. So what they use your data for is their choice, and they have complete control over it. And this is key, they have to respect the wishes and needs of their users to establish and maintain the trust in the brand.
Our clients have told us very clearly what they want, and it’s also our business model — we don’t own their data. We help them to deliver more value and insight so they can provide an elevated customer experience to their end users through their data.
Skift: Ah, so when Wynn Las Vegas and other hotels are installing Amazon Echo units in their rooms so guests can access Alexa, that means Amazon is keeping all that data, correct?
Greenstein: Correct. Correct. Amazon has access to the data because it all goes through Amazon’s services. That’s a pretty big exposure for most brands. But, most companies aren’t engaging their users this way yet. Most companies have a mobile app that barely anybody uses. But if you can get people to engage directly and interface with your business through voice and conversation, you can learn a lot about your users. That can help you to derive significant insights about a person likes or dislikes. Those insights are the lifeblood of a company.
When we started on this path with Watson Assistant and IoT, it was a time when most companies sold you stuff and they never heard from you again. That is, until there’s a warranty problem or a complaint or something. But they never learned about all of your daily needs. Now, in the case of hospitality, for the two weeks before you go to a hotel, there are tons of opportunities to engage with you. The hotel can remind you of the weather, or let you know about great things happening, like local events, or things to remember to bring. There are lots of moments of engagement that could occur through the mobile app, for example, with Watson Assistant giving you useful insights.
And then when you get to the hotel, Watson remembers everything you talked about in the weeks leading up, and then it reminds you, “Don’t forget to go to this great concert. It’s just around the corner, and tonight the hotel has a special deal on it.” There are lots of ways to engage you, and all of that data belongs to that hotel. They’re the ones creating the experience, so they should be able to control the data. Turning that over to Amazon or anyone else is just a huge exposure for them.
Skift: So, the big picture here is that this type of AI-enhanced customer service can increase loyalty between a brand and a customer, right? Because Watson plugs into the hotel brand’s loyalty program like you’ve mentioned. So, is this the next generation of building brand loyalty?
Greenstein: It is, although I think it’s also about creating value and service. Each company has a slightly different goal when they talk to us. A lot of it is about engagement, extending the brand, building loyalty, but also providing additional services. In the case of hospitality, today, you go to a hotel and very often there are flyers on your desk reminding you to go to the pizza place around the corner. Obviously, somebody has an agreement with that pizza place so they can promote it in that hotel. Imagine as a brand, you can start to encourage and suggest and refer people to all kinds of places. It’s also an opportunity to create new revenue streams by referring people to other businesses, based on the relationship and any special deals you may have, which are tied with the guest’s personal preferences.
There are tons of ways to add value in that experience. By knowing your customer preferences, you could do things that anticipate people’s needs. We’ve built into Watson Assistant how to wait and respond for you to ask for things, but we’re also able to anticipate your needs based on patterns and triggers and moments that might occur. For example, I’m going to Chicago next week and it’s going to rain. Maybe I checked, maybe I didn’t, but the hotel in Chicago certainly should know that. They should remind me to bring my umbrella before I’ve ever even asked. Yes, it builds loyalty, but more importantly, it builds value. And some of that value actually can produce revenue, in the case of referrals to other businesses.
Skift: So, are you saying the hotel becomes a more comprehensive travel portal into the local community, because you can integrate many more businesses in the destination, and recommend them to the individual customer with much more relevance? Is Watson, then, a travel destination platform, because we’re going way beyond partnerships with Hertz and the pizza place around the corner?
Greenstein: Correct. Usually, in the early stages, it’s really more about the hotel at the center, and recommendations based on what they know about your preferences, and the hotel’s partners and ecosystem. If they don’t have a partner — let’s say my preference is vegetarian food — maybe there’s no partners that even relate to that. But, at least just offer me the Yelp recommendation for that. That’s a perfectly good way to do it. But over time, hotels have the opportunity to build those relationships, tie them to your preferences, and suggest things that you might like. You may choose to do them or not, but when you do, the hotel recommended it, so they should get some credit for having done that. If it makes you really happy, it builds into your preferences and you’re more likely to go there again.
Skift: Back to data. If the hotels are capturing the data, and they own that data, is there any kind of evolution in terms of how they can use it and access it and leverage it more productively? Or is that already inherent in what we’ve been talking about?
Greenstein: That’s a good question. It’s inherent in the way that that brand manages, for example, their loyalty program today. If you think of the privacy policies and data use policies that exist in a given hotel, they already know a lot about us. They know our TV viewing habits. They know when we check in and when we check out, and what we bought and how much alcohol we drank. They know all these things about us. Every hotel obviously is built on trust, and so they’ve established their own data use policies.
Here, what we’ve done is we’ve created an opportunity for significantly more engagement, which also means significantly more data. That also means those policies need to be looked at and reviewed carefully. So we have some best practices we do, but for our clients, they know how to establish trust with their users. We just recommend best practices and good policies that they can use, but it’s really up to each of them to establish trust with their clients.
I think for all of us in the world of IoT, all this increased volume of data and insights is certainly a risk, and there’s no question, it’s a risk. But that’s where you need companies who understand data privacy, who work with GDPR and HIPAA compliance and all of these regulatory things in every country, and understand how to deliver it.
Skift: Can hotels completely integrate their chat platforms with IBM Watson Assistant and give them the same name? For example, The Cosmopolitan Las Vegas has a cool chatbot named “Rose.” Can I now talk to Rose using Watson Assistant? In other words, if I’m chatting with her via text on mobile during the day, can I speak to her via voice in the room at night? Is that an issue at all?
Greenstein: It’s not an issue at all. We’ve have two models. One where people already have some sort of basic chat functionality and they want to extend that data to Watson Assistant. Watson Assistant, yes, it’s conversational voice, that’s the primary interface. However, it’s not really voice dependent at all. It works with text input on mobile, or we can go through Facebook Messenger or Slack or any form of conversational input. So, yes, we can take the input from any chatbot and feed it into Watson Assistant.
We also can provide, literally, the engine for that chatbot. People can just build the chatbot itself right in Watson Assistant and then tie in voice in the room. There’s a built-in integration in that case, so the conversation that comes out of the chatbot is sent into Watson Assistant, and then we have that for your profile. Our inputs are quite broad.
It makes the entry into artificial intelligence and conversational interfaces so much easier. A lot of people were experimenting with chatbots and trying stuff, so we knew we had to be able to support that. All of our input, our conversation functions — whether they’re spoken or typed, or even a button in a mobile app that triggers an input — all of these are valid inputs. We kept a very open mind on that.
The above content was produced by the branded content SkiftX team for the upcoming Skift Cities platform, defining how cities are connecting visitors and locals to co-create the future of urban UX.
Have a confidential tip for Skift? Get in touch
Tags: artificial intelligence, IBM Watson, skift cities
Photo credit: IBM Think 2018. IBM