Definitely Doug 7/10/20: Voice Assistants

Alexa, Should I Buy a Voice Assistant for my Hotel?

Over the past few years, there has been much media hype around the concept of a voice-controlled hotel room. It’s not hard to see why: voice assistant devices such as Amazon Echo and Google Home achieved remarkable penetration in the consumer market in just a few years. Statista reports that about 157 million smart speakers were installed in U.S. households as of December 2019, an astonishing 1.22 devices per household. I haven’t found hard numbers on penetration in hotels, but based on the companies in the market and what I know of their size and success, it’s still very low, probably still under 1% of US hotel rooms. Is that about to change? Should it?

COVID-19 has created a rush toward contactless technology for hotels, so it seems like a good time to take a deeper dive into voice assistants and voice control technologies. What can hotels expect if they start to explore this option? What are the considerations? Will voice become the “must-have” for hotels that it has already become in most homes?

There aren’t a lot of commercial solutions in the field – Volara, Angie and Intelity are the three that have been most visible. I wouldn’t describe any of these as truly mature voice technologies. Certainly, they can do some useful things, but they also fail (or provide a poor guest experience) at some things you’d really like them to be able to do. But each has had some success, and they represent different approaches worth considering. There are also at least a few custom solutions that have been built on toolsets provided by Amazon, Google, and others. I will not review any of these products in detail here today; rather, I will highlight some of the things I think are important in evaluating a voice-assistant solution.

I will start with a blanket prediction, that at some point in the not-too-distant future (my bet is 5-10 years), voice control will be ubiquitous – a must-have for virtually every business. It will be smart enough to pass the Turing Test most of the time – meaning that you won’t be able to tell if you’re talking to a human or to a machine. The underlying technology is not there yet, but it is improving significantly every year. To be sure, there will be consumers who will choose not to use it, whether out of privacy concerns, habit, or other reasons. But if you look at generational data, it’s clear that usage will increase as the population ages.

This isn’t to say that any of the existing devices or vendors available in hotels today will necessarily be relevant at that future date. There’s a lot of innovation going on at market-leading companies like Amazon, Google, Microsoft, and IBM, both internally and via acquisition. This will set the baseline for devices and foundational toolsets, particularly around natural language processing. Device form-factors will change over time in ways that may be hard to foresee. Dedicated devices may become less necessary for some applications, as capabilities are increasingly built into mobile devices, wearables, and apps. On the other hand, specialized devices have some significant advantages for hotels, such as the ability to replace the guest phone, radio, and even (for those that have an LED display and/or screen) the alarm clock, and the printed guest-room compendium.

It’s hard to argue with the success of voice assistants in the consumer market. Few of the statistics I found published online stood up to close scrutiny, but the 200 million Echo devices that Amazon sold through the end of 2019 is hard to dispute. Reading between the lines of several sources, it does seem like something in the range of 25% to 40% of searches are probably now being done by voice. Whatever the true number, it’s significant and it’s growing.

So, with all this traction, why have voice assistants been so slow to catch on in hotels? First, most of the voice tools still have only a limited ability to understand natural language, and even less ability to understand context. In the home, you may be willing to spend some time to learn Alexa-speak or Googleese in order to simplify some common tasks. For a brief hotel stay, your willingness to do so will likely be much more limited, and you may often be challenged to find the words that the voice assistant understands.

Of course, these systems are programmed to recognize the most common requests, and they can learn from their mistakes over time. But do you wonder why every product demo seem to include a request for towels? There’s a good reason: it’s pretty safe to assume that any sentence that includes the word “towel” is a request that can be fulfilled by delivering towels; it’s about as close as you can get to unambiguous. Setting the temperature and getting the weather forecast are also good demo options, because there aren’t that many ways to phrase the request, and context isn’t usually needed to interpret it. If I say “set the temperature to 72,” it’s obvious that I want to interact with the thermostat, and also that I’m using the Fahrenheit scale rather than Celsius.

Beyond that, you may struggle to guess which words will work, and the system may fail to understand your meaning and/or context. “Turn on the left-hand bedside lamp” is a reasonable request, but which lamp is “to the left” may depend on whether you are standing in front of the bed or lying on top of it. At home, you know how you trained your voice assistant (or how it trained you); that’s not the case in a hotel.

Many of the things hotels are trying to do with voice are much harder than towels or temperature. To be sure, some of the commercial products have made progress, and I urge you to test out some of these scenarios with the different products because their abilities vary significant. But talking about some of the hard things, I’ve seen a couple of demos of room service ordering by voice. One was much better than the other, but both still had a long way to go. The system simply can’t anticipate all the phrases someone might use in the process of ordering a meal. Consider, for example, someone who has questions about allergens and needs help choosing safe options. And even in the simple cases, it can take lot longer to get through the overly structured ordering process than it would talking to a human or using an app. If I were buying a voice assistant for a hotel, I think I would focus on doing a few things well, and not try to do hard things like room service ordering or spa appointments. When one of the commercial products can make ordering room service almost as easy as talking to a human, then for sure I’ll be looking at it, but that hasn’t happened yet.

While some of the vendors have been pushing the limit of what’s possible, there is one compelling but often overlooked reason to consider a voice assistant in a hotel room, and that’s the all the “easy” things it can do. Even a consumer device like the Echo Dot with no hotel integration can provide high quality music, an alarm clock, VoIP calling, and generic information services. It can replace the clock radio with something of higher quality, and with a little bit of customization, it can replace the hotel telephone as well. Many of these consumer devices cost less than the typical hotel telephone, too. I might still put a human on the other end of ‘Alexa, order room service’ request, but I wouldn’t underestimate the benefits of the basic consumer device for the things it does well.

As you’re considering your options, another issue is the speaking voice, which is usually some variation of mono-robotic. We need to remember that we are in the hospitality business, and we expect our human staff to be helpful and friendly, warm and empathic. To be sure, I’ve talked to voice bots that have a bit of (often quirky) personality, and some of them are fun to interact with. But I never get the warm-and-fuzzies I can get from talking to a human. A key question in evaluating any voice assistant is whether the actual voice conveys a brand image that’s appropriate for your hotel. It’s one thing to put a basic device into a hotel room so guests can make standard Alexa-like requests and get an Alexa-like response. It’s quite another to have Alexa’s voice describing your hotel’s five-star room-service offerings in robotic monotone.

Another challenge to voice response in hotels is integration. It’s certainly possible to implement voice control of lighting, TV, drapery controls, temperature, the telephone, and the do-not-disturb and make-up-room lights. For a single hotel, it’s usually not too difficult (although in some cases it could be expensive).

On the other hand, it is extremely hard to do it consistently across a large brand. Most of the systems you want connected are chosen by hotel owners rather than brands. Across their global portfolios, large brands may have dozens of different providers (or more) in each category. Many legacy devices may not be programmatically controllable at all, meaning voice control will require replacing or adding devices, at a cost to the hotel owner. Other devices may be controllable but require an interface that doesn’t yet exist or that needs to be purchased (another cost). Yet for a brand, it makes little sense to implement voice response unless it can work consistently across the entire portfolio. Without this, guests’ expectations will be set at hotel where everything works, and then disappointed at the next one. Even for a single hotel, the costs of adapting existing systems can often be larger than the cost of the voice system itself.

We can’t ignore the issue of privacy, either; it’s reportedly why many of the major brands have hesitated to put voice assistants in their guest rooms. How do consumers know the device isn’t listening when it’s not supposed to? Are requests being recorded and saved for analysis? Can requests be tied to personally identifiable information? Are the systems protected from hackers? These are legitimate concerns, and the vendors in the space seem to have taken them seriously. Some of the latest designs make it much easier to see when the device is listening for its trigger word, actively listening to a request, in standby mode, or totally off; done right, I think this is a good idea. The truly paranoid guest can still unplug the device – that’s something I have done for years with the alarm clock; it was easier than figuring out how to make sure it wasn’t set for 4am from the last guest.

Another key question is, what’s the right device? Many consumers are getting comfortable with using Siri or Google voice commands on their mobile devices, so would they not be happy using this to make voice requests to a hotel? Some of the chatbot providers are looking at taking the text transcription of a request made through Siri or Google and passing it off to their text recognition engine; this capability could be embedded into a hotel app as well. This will be something to watch in coming months – all of the necessary technologies exist, but they haven’t been brought together yet in a commercial product. In the post-COVID world, the device least likely to cause a problem to the hotel is the guest’s own mobile phone: the hotel never has to clean it, isn’t responsible if it gets contaminated, and to the hotel, it’s free. Furthermore, guests who may be concerned about Alexa-type devices eavesdropping or recording, seem to be blissfully unconcerned that their mobile devices are able to do the same things. And if they do have their privacy violated, they will blame their device rather than the hotel.

For the past five years I have said that voice recognition technology was coming, but that it wasn’t yet ready for the hotel industry. To date, the market has borne that prediction out. But between consumer trends and COVID-19, I believe we may now be at an inflection point, and this could be a very good time to take a serious look.

‍