Building a better customer support
I’m pretty sure I’m not the only one who thinks “Oh Fun, do I really need to contact customer support?!”. I, like everyone else, already know what’s waiting for me on the other side of the line: a bot that can’t understand me, an irritating song while I wait for someone to pick up my call, or an email thread that will take days for it to be resolved.
75% of customers believe it takes too long to talk to a live agent.
(Source: Harris Interactive)
But this is not a simple problem to solve for any company. A possible solution is to add more people to your call centres. However, as the company grows this approach becomes very unrealistic.
Labor can account for as much as 70% of total business costs
(source: paycor.com).
The other solution is to better filter the customers' queries before reaching a live agent. Ideally, the ones that can be easily resolved, shouldn’t reach a live agent at all. But that has been the achilles heel for most companies. So there’s another more burning question; how do we make our chatbots an enjoyable instead of a painful experience for our customers?
The chatbot evolution
Chatbots aren’t new, they have been around since the ’60s. The first chatbot ever was developed by MIT professor Joseph Weizenbaum in 1965 and it was called… ELIZA (I am not making that up! :P )! Chatbots are digital text or voice applications that help people to put their inquiries via text or voice. The old-school chatbots primarily use pattern matching and substitution methodology to simulate conversation.
The biggest evolution happened when we started replacing the rule-based, pattern matching with Artificial Intelligence (AI). Those new chatbots-on-steroids, are AI based and make use of the Natural Language Understanding (NLU), and Machine Learning (ML) technologies. We also started referring to them with a new name, conversational apps.
The foundation
The conversational AI world brings a lot of new terminologies. For instance, instead of a graphic user interface (GUI or just UI) we now have a voice user interface (VUI); instead of graphic designers, we have conversational designers and many more.
The lack of a graphic environment doesn’t mean there’s no design process. On the contrary, this is now even more important.
I remember once, I was back home from one of my business trips. I was jetlagged, exhausted… barely functioning. A total mess! I was looking for the vacuum cleaner because I broke something, but I couldn’t find it. I went to my partner, and I was trying to explain to him what I was looking for. The problem was, my half-asleep brain could only remember the Greek word. Neither the Dutch one, nor the English one. And he didn’t know Greek. I ended up telling him “I need the electric broom 🤦♀️🤦♀️🤦♀️”. And it turns out I am not the only one!
It doesn’t matter how old, or how educated you are. People can get very creative when they don’t know, or even when they just forget a word.
A more comprehensive article on why conversational design is so important can be found here.
The process
The process of creating a conversational app is not much different than any other app or product. To better understand the process, let’s assume we own a spa for our fur friends.
At PawSome Spa we found out that although we offer the best customer experience once the customers walk in, there are some complaints for our call center. For all the reasons mentioned before, we decided that the best approach for our business is to get rid of our chatbot and create a conversational app instead.
Step 1: The question
In this particular case we already know which question our conversational app should try and answer: how do we build a better customer support.
Step 2: Research
Research is mostly part of the job of a conversational designer. They will be responsible to gather all the data, and decide which actions, or better put, which user intents should our app be able to handle, and which not. Moreover, during this phase we take all of our tech stack decisions. Should it connect to a backend? Or to call center? Will it also be available for Alexa? How about Google Home? All these questions are important to be answered before we start designing & developing our MVP. The reason is simple; avoid miscommunications between the teams, and don’t include actions our infrastructure can’t support at this moment. Imagine designing the entire VUI for modifying the appointments, just to find out that your infrastructure can’t support it as is, and some work has to be done there first blocking the rest of the teams.
PawSome Spa example: From the data we gathered we found out that these questions are the most frequent our live agents receive:
1. Information about an appointment
2. Dog yoga program information
3. Book / modify an appointment
4. Information about grooming services
5. Store information (operation times, location, etc)
6. Check vaccination validity
Our conversational designer decides that intents 1–5 should be handled by our app. Intent 6 though should redirect directly to one of our live agents. To check the vaccination validity we need to know the pet’s health card ID. This information though is not something that the user will have on hand, but they’ll have to search for it. Thus, this makes it an unfitting use case for a conversational app. In conversational apps we optimize for speed, simplicity, and hands-free interactions.
Step 3: Design
Design is — to no-one’s surprise — also part of the job of the conversational designer. While it may be difficult to imagine what the design process covers exactly, design is one of the most important steps. During the design process, the designer will create our app’s persona & will create the conversation decision tree (aka flowchart) which will then get “translated” into code. This is the step where we take care of any edge cases, and conversation repairs in case it breaks.
PawSome Spa example: After selecting the intents our MVP app will cover, our conversational designer will start with creating the basic flowchart.
Our designer will continue to iterate and modify the above flowchart through the entire phase. Once the base flowchart is created, they will then start creating sample dialogs that will cover every use case (happy path, conversation breaks, successful & unsuccessful attempts to repair the conversation).
Eventually, we need a persona for our app. The characteristics of our persona depend on
I. our clients,
II. our branding, and
III. the type of business we have.
If we’re into banking for instance, we need to use a very formal way of communication to communicate earnestness. If we have a toy store, it’d be the opposite since we want to communicate we’re a fun business. Our PawSome Spa is quite exclusive, so we’d like to keep it formal. However, we know that adults turn into little kids when it comes to their precious pets. They need to know that their pets are not only taken care of, but that they have the time of their lives, so our persona should reflect some fun. At last, we are a spa after all, therefore our voice should emit relaxedness and calmness.
Step 4: Build
The building process depends on the platforms we want our app to be available at, as well as the tools we want to integrate it with. If we want our app to be available on Alexa we need to create an Alexa skill, whilst if we want it to be available on Google Assistant, we need to create an App Action (important note: Google will sunset App Actions in June 2023). Although it may seem a lot of extra work, in reality the two platforms are very similar. Moreover, since everything we build is based on the design’s flowchart, maintaining both projects is a relatively easy task if done correctly. In the case we’re interested to have our app available on our website or our call center, we need an app (or else known as agents) that can easily be exposed to such an interface. Such tools would be Google’s DialogFlow, or IBM’s Watson. In spite of our personal preferences, at the end of the day, this decision should be taken based on your current tech stack, infrastructure, and the integrations & tools you need.
Step 5–7: Test — Improve — Launch
As soon as our build phase is complete we can start testing our app and its logic. This step includes both testing out various conversation turns, and our overall system stability. Once we’re happy with our app’s performance, we’re ready to launch 🎉. Next, we need to keep an eye on our performance metrics to confirm that our app works properly, or make any necessary improvements and alterations. Eventually, we can continue on improving and adding new features to our app as our company grows or the company’s needs shift.
A state of art customer service example
Interested?
Pixplicity.com is a team of experts who are well-versed in the latest AI technology and can help you create a chatbot that is tailored to your specific needs. With Pix, you can be assured that your chatbot will be able to understand human language and respond accordingly.
Our team can help you integrate your chatbot into your existing systems, such as customer service, sales platforms or even the Metaverse. Contact us today to learn more about how we can help you build a successful conversational AI product!
