Chapter 6. The Elements of Starting

You get only one shot at making a good first impression.

With voicebots, such an impression is formed by users and conveyed by VUI designers with the conversation’s opening sequence.

Here are ten tips that can help you craft an effective opening prompt.

1. Be brief

Belabored, verbose opening prompts confirm the worst stereotype of the dumb, overbearing voicebot. If you force users to listen to 30 seconds of instructions, information, and disclaimers before they can take the first step towards solving their problem, you will not only have started on the wrong note, but would have given the user a whole 30 seconds to change their mind and terminate the conversation. Every single word in your opening prompt needs to be absolutely indispensable. If you can get rid of a word without losing meaning or effectiveness, do it.

2. Use an audio icon

A well crafted, distinctive, pleasant audio icon -- or, even better, an audio icon that the user is familiar with and is likely to associate with the brand that the voicebot represents -- played at the very start of an interaction is not only an effective way of signaling to users that they are interacting with a voicebot and not a human, but also a good way of communicating to users that the voicebot was designed and crafted with care. The implication is that the brand they are engaging with does care about its customers and will work hard to ensure that customers are satisfied. This starts the interaction on a positive note upon which further experiences can build for usability.

3. Drop the “Welcome to...”

Instead of using the perfunctory “Welcome to,” simply have the voicebot announce the brand’s name, preceded or followed by an audio icon, and then followed by the company’s tag line, if the company has one? Such an opening will not only set your voicebot apart from the garden variety ones, but will have shortened the length of your prompt.

Instead of:

Voicebot: Welcome to Widget Solutions, where Intelligence is at your service.


Voicebot: <Chime>. Widget Solutions. Intelligence at your service.


Voicebot: Widget Solutions. <Chime>. Intelligence at your service.

4. Never ever say, “Please listen carefully as our options have changed”

This is an awful invention and it must be banned once and for all. Upon hearing it, the user will immediately think of the old style “phone tree,” and that is not a good association.

5. Have the voicebot refer to itself in the first person

Empirical studies have shown that users like to have the voicebot refer to itself with the personal “I” rather than the impersonal “assistant” or “system” or “skill.” 1

Instead of:

Voicebot: This skill tells you what state a ZIP code belongs to.


Voicebot: I can tell you what state a ZIP code belongs to.

6. Drop “You can interrupt me at any time”

As we have argued in Chapter One, we advocate for the establishment of a new world order where the Voice User Interface designer does not feel that it is their responsibility to educate the user on the basics of engaging with a voicebot. In addition to being able to say “Repeat” to have the voicebot repeat what they just said, or “Skip” to go to the next item when the voicebot is speaking an enumerated list of items (e.g., summary of news stories), users should be able to interrupt voicebots at any time.

7. Keep the origination context in mind

If at all possible, keep in mind the context users may be starting from in their interaction with the voicebot and the assumptions they are likely to make about those interactions.

For instance, if the human is engaging the voicebot in reaction to an ad that said, “To place your order, just say, ‘Hey Google, Talk to The Mighty Green Grill,”

Instead of:

Voicebot: <Chime>. Grills are Us! Which one would you like to order: The Mega Red Grill, The Mighty Green Grill, or The Sturdy Blue Grill?


Voicebot: <Chime>. Grills are Us! Are you ready to order The Mighty Green Grill?

8. Remember the user’s preferences

If the voicebot is able to identify the user (for instance, their smart speaker device ID, or the telephone number they are calling from) and has a record of their language preference, then the voicebot should drop the language question and simply proceed with the language preference it has on record.

Another example would be asking them for their frequent flyer number or their member ID.

Of course, should the user need to access sensitive information or execute transactions, a layer of protection needs to be introduced (for instance, asking them for a PIN code).

9. Anticipate user-specific requests

Similarly, if the user has recently placed an order from the brand’s website or mobile app, then they are probably engaging the voicebot to find out about the status of that order. If safely providing the user with such information in the opening prompt does not take more than a few seconds, then have the voicebot volunteer the information before offering the standard main menu. Or, at the very least, ask them if that is why they are engaging with the voicebot.

Instead of:

Voicebot: <Chime>. Grills are Us! I see that you have placed a Mighty Green Grill order with us. Do you want to find out what the status of the order is?

Human: Yes.

Voicebot: Ok. Your order is about to be shipped to the 6505 address. Anything else?

Human: No.


Voicebot: <Chime>. Grills are Us! Just so you know, your Mighty Green Grill order is about to be shipped to the 6505 address. Anything else?

Human: No.

10. Anticipate general user-base requests

In addition to leveraging specific information about the user, the voicebot should also leverage information about the user base as a whole.

Here’s an example: Your customers are launching your voicebot to report a sudden service outage and are overwhelmingly selecting the “technical support” option. If the voicebot is experiencing such a spike on a specific menu option across users, then have the bot adjust its behavior to offer that option first before presenting users with the standard main menu – e.g., “Do you want ‘technical support’?” Again, the voicebot -- and therefore the brand -- will come across as smart and proactive, and in the case of an outage, the customer will immediately feel better knowing that the company is aware of the outage and is therefore probably working on it. All in all, your brand will be perceived as being on top of things. The voicebot should of course revert to the pre-spike menu behavior when the request spike subsides.

1 Boyce, Susan J., “Natural spoken dialogue systems for telephony applications,” Communications of the ACM, 43(9): 29-34, 2000.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.