Chasing J.A.R.V.I.S Part II: Hotel Concierge
A VUI case study on a truly conversational hospitality experience.
Speaking of jobs being taken by machines, one of them is the hotel concierge.
Well not really. For now at least we still need a human to run the higher level stuff. But the mundane, like getting towels, asking basic questions about amenities, and taking room service orders are all things that a virtual assistant can handle.
The Project
What if Marriot wants to put an Alexa device in every hotel room?
For this exercise, we wanted to consider what the system persona would have to be like to have a truly conversational feel to it.
We will assume that when the guest checks in, the Alexa will be signed into their Marriott account, which has their credit card on file.
We also do not need an invocation name. It is possible to create a custom app that automatically jumps into the Marriott app without saying “Alexa, launch Marriott” Whenever the guest interacts with the hotels Alexa, it will assume you are interfacing with the Marriott skill.
Use Cases
We wanted to rapidly prototype for two use cases. When the guest of the hotel wants to:
- Order some towels
- Order room service
We went out and did some guerrilla user research, talking to people about these features.
What we did was quickly role play with each other and people on the street using these prompts:
“You are in a hotel and you would like to order (more towels/room service) there is an Alexa in your room, what would you say?”
We would then tell them to talk to us like we are Alexa. The following is some of the quotes from our research.
“I would say, yo Alexa, bring me some towels”
“I will always tell the concierge what room I am in when I call them”
“Usually I order a pack and play when we go to a hotel, I rarely order room service because it is so expensive”
“We usually order new towels when my daughters dirty them up. They go through them quick!”
“We almost ordered room service, but we just decided to go downstairs to the restaurant because it’s the same food but take more time and is more expensive”
“I usually want to get a time range of when something is going to come to my room”
The last phrase got me thinking about a step by step notification like what Dominos does with their pizza delivery. “Your pizza is in the oven/the delivery is on it’s way/toppings are being put on!”
Towels seem like a common order to be fulfilled. But room service is more of a high ticket experience. Peoples main concerns with the service was the amount of time it took to get to you and if the price was worth the same food they could get faster and cheaper if they went a floor down.
Potentially, if we allowed for frequent updates so the guest doesn’t feel like they are forgotten might help improve the experience and is something a voice assistant is in a great position to add value.
The System Persona
After talking to users, we thought that the most natural adaptation to a virtual assistant would be a persona that was a personal concierge to the guest.
Interactions with a human concierge are formal, but relaxed, transactional, but casual. Meaning that the interaction has an underlying understanding that the concierge will give you the information you want, but the stakes are not super high so the conversation shouldn’t need well prepared questions.
Usually, the guest will ask about basic services and amenities and the concierge will answer in the most hospitable way possible.
The conversational aspect requires the system persona to sound like they are a trained concierge that is fresh and well rested. They are helpful and make the guest feel heard and at ease.
Scripting
The first iteration of scripts looked like this:
First iteration was more to the point. We let them know what they would be getting into and how long it would take to get there.
But after considering some variables. We thought about how there was a certain kink in the conversation nature of the VUI. We didn’t consider how many responses we could handle in the room service example. Towel requests are simpler in theory, but the user ordering room service needs more reassurance that they are in good hands ordering from a robot.
Iteration two went like this for the towel:
We wanted to make the room service experience better than a normal human concierge. By adding time updates we can hopefully offset the cost of getting the food themselves.
Prototyping
After nailing down a solid iteration, we placed the scripts into a Sayspring prototype.
Next Steps and Challenges
We currently haven’t had the chance to test the skill, but there are a couple challenges we face in order to get the tests right.
The room service prototype need to have a menu that compliments the supported responses. After doing that, we need to program in multiple supported answers for each menu item before the prototype is ready to test.
The second challenge we had as a group is being able to explain a voice design concept to people we are interviewing. We haven’t yet developed a natural way to role play with people. How do we go about mentally prepping them that they can talk to you like you are Alexa? I’m still trying to figure out how to explain what we are trying to do in that front so it isn’t awkward and the interviewee can act more naturally.
To be honest, when there is a bunch of potential supported answers, I think a conversational VUI must refer to in room references like the physical menu for things like room service. If there is a failure to support an answer, the voice assistant should suggest that the user call the front desk.
I thing a failed to mention earlier is that an interviewee brought up the idea that you should replace the phones with Alexa. I think this would be a good trigger moment to encourage guests to use her to make calls and call for room service.
