Build an Assistant That Your Users Will Actually Want to Use
Not Just Another Chatbot
When one of your customers has a support problem or a sales inquiry that they want to resolve with your business, they [obviously] want to find the most expedient path to a successful outcome. And if their issue is especially nuanced or complicated, they also want to feel like they’ve been heard. But for most midsize and enterprise companies, the sales and customer service landscape involves a number of unpredictable channels, slow resolution times, and the feeling that no one really cares about me.
Chatbots emerged in 2016 as a potential solution to this problem — a quick way to clarify and resolve common sales or support inquiries in a familiar, conversational interface. Their hype grew alongside the rise of various messaging channels and vast improvements in natural language understanding technology. They were positioned as app killers—the user interface of the future. Unfortunately, most chatbots manifested as yet another disconnected, frustrating, and inflexible channel for users to get help:
Hey I’m having trouble making a payment to a friend
Sure! Let me help you pay your bill…
No that’s not what I wanted
I’m sorry, I didn’t understand
AGENT AGENT AGENT AAAAAHHH
Hmm… I can’t help you with that, but you can submit an email to our support team if you’d like!
Look familiar? While a message-based user experience works well for humans to interact with one another, it often breaks down in the interaction between a human and a machine due to…
- The sheer number of topics and permutations a chatbot designer has to program to make the bot useful
- Rigidity in the experience that the user often cannot escape from
- The inability to browse or navigate the system without knowing what it is able to handle
- A lack of structure or focus that a graphical user interface provides
- No continuous and flow context from machine to a human agent when things go south
Needless to say, chatbots haven’t had the best reputation these days. However, we still believe in the power of machines understanding natural language — especially at the beginning of an interaction. And we also believe in the initial spirit of the chatbot hype back in 2016: a consistent and centralized system in which to solve problems. Chatbots are by no means dead, they’re just in need of a reboot! That’s where Watson Assistant can help.
The 4 Cs
Working with thousands of customers over the last few years, we’ve learned that to be useful, a support or sales assistant needs to meet a few key criteria (the 4 Cs, if you will):
1 | Centralized
2 | Connected
3 | Contextual
4 | Compelling
Centralized
Your customers shouldn’t have to decide between multiple phone numbers, live chat, email, FAQs, a self-service portal, and your knowledge base to figure out how to get their job done. Users want a consistent and predictable experience on the channels that are most convenient (web, mobile, social, SMS, email, you name it) in which to solve problems. They aren’t necessarily trying to have a conversation either — they’re just trying to get a job done efficiently.
Connected
Your assistant has got to be connected to existing sources of self-service content, existing service desk tool(s), and existing systems to lookup or post information on behalf of the user. Otherwise, it’s honestly just another lonely, disconnected, and generic FAQ system that will break down when a user asks something complicated.
Contextual
How many times have your users had to explain their problem in one of your channels only to realize that they’ve hit a dead end (and have to repeat themselves somewhere else)? Once you’ve gotten a user to pose their problem to your centralized assistant, the system should be able to understand as much context about the problem and about the user as possible, ask the user (only once) about what it doesn’t already know, and then route the issue to the best place for resolution.
Compelling
Finally, your Watson Assistant needs to be compelling to use, meaning it should be seen as a tool or partner that a user wants to return to use in the future. And, again, it doesn’t necessarily have to follow a free-form conversational flow to be useful. We find that the best assistants are hybrids — they allow the user to express their problem in a natural way, but then focus the user once the problem is understood. It also has to be flexible enough to digress to a new topic when the user deviates or make mistakes in understanding (which it will) without frustrating your users.
We’re Making it Easier
And we want to make it even easier for you to launch your own centralized, connected, contextual, and compelling assistant. That’s why we’ve just launched the following new features within Watson Assistant:
- Search Skill (Beta)* that lets you centralize your assistant by crawling any of your existing, public facing web content and searching against it when your assistant doesn’t have a concrete response from the Dialog Skill
- Skill Versions that let you manage updates to your assistant’s training in one centralized place without having to deal with separate copies
- Service desk integrations with Intercom and Zendesk to allow you to connect human agent, with the full context of what happened prior to the escalation
- Intent Recommendations (Beta)* that scans your existing chat or call logs to help you decide which topics your assistant should handle explicitly (making it more compelling to use, of course)
- Disambiguation that provides a more flexible (and, yes, compelling) assistant experience by automatically clarifying the nature of a vague question from an end-user before sending him/her down a path
All of these features are available now or coming soon in the Watson Assistant Plus Plan! If you’re not a plus user yet, our sales team can set you up to try out these new features for free — just sign up for Watson Assistant here then shoot them a message requesting access.
*Note: Beta features are available in a separate beta environment, which is also free to try! To check it out, request access at the bottom of the “skills” page within Watson Assistant.
