Why You Can’t Just Convert FAQs into a Customer Service Chatbot 1:1

avatar

As companies are exploring the use of customer service chatbots, one of the first candidates of content that come to mind to “botify” is the existing FAQs often represented on corresponding sections on a website. It is tempting to think that you could just take these and convert them into the form of a Conversational UI (CUI) – they are already in the form of answers to questions after all, right?

In reality though, it takes a bit more to build a chatbot that really does a good job of serving customers. Taking existing content and just swapping out the User Interface is a natural gut reaction whenever a new UI emerges. This was the case with early Websites and mobile apps on multi-touch smartphone screens. Early Websites were full of text and had little rich media, let alone interactivity. And it wasn’t until Angry Birds came out, a full year and a half after the launch of the app store, to truly embrace gestural multi-touch UI. And these are just of few of the many other examples that can be cited. Now we have the CUI and it will take some time for the designs to adapt themselves to this relatively new paradigm of man-machine interaction.

Best practices are being established, and first books about the topic are being published (e.g. Designing Bots: Creating Conversational Experiences by Amir Shevat, or, with more of a focus on voice: Designing Voice User Interfaces: Principles of Conversational Experiences by Cathy Pearl). Tools that promise to just absorb an FAQ section of a website and convert that into a conversational bot are out there and sound promising – yet they quickly fail with simple real-world examples.

So WHY does an FAQ not translate 1:1 into a customer service chatbot? There are several reasons.

1) The nature of a dialog: content is king, but context is queen (and she runs the household)!

This is the probably the most important reason, so I’m mentioning it upfront.

Chatbots are meant to chat with the user, not to produce one answer and then end the conversation. Humans don’t communicate in question-answer pairs. Real world communication is much messier than that. By offering a medium commonly used for conversations among friends and family, say SMS, iMessage, or Facebook Messenger, you have to assume customers expect their experience will be similar when chatting with a business. Yet many chatbots can’t even respond to “hi“, one of the most frequent messages sent to a chatbot.

As a side note: to me, this is the number one reason why IVR systems so often disappoint. We expect human, but get “press 1 or press 2” – an unfamiliar and unexpected user interface on a highly familiar channel.

It is in the nature within a dialog for people to ask follow-up questions. And when they do, they like to use pronouns to refer to previously mentioned topics or things. This follows the least-effort principle that permeates human language in so many areas. Consider the following four FAQs a bot covers that we built for an automaker earlier this year:

  • What is Concierge Service?
  • How can I contact the Concierge Service?
  • What is the price of the Concierge Service?
  • Can my partner or a co-user use the Concierge Service?

Once the context of “concierge service” is established through a question such as “tell me about concierge service“, a follow-up question might be “how much is it?”. It is thus crucial to maintain context to be able to answer questions such as “how much is it?” or “can my wife use it?”. What is it? Or, as a linguist would ask: Which antecedent does the pronoun refer to? To handle that in customer service chatbots you will want to design your Conversational Architecture.

Sometimes even the exact same question can yield different answers, depending on the flow of the conversation. Context awareness is critical for a customer service chatbot. Without it, users will read “I’m sorry, not sure what you just asked” a little too often… a mistake made frequently with the early bot implementations we saw in 2016.

2) Scope of typical FAQ content

Frequently asked questions are just that – questions that are frequently asked. For the not-so-frequently asked ones, users are left alone to figure out how to get human help. A customer service chatbot that just covers the questions that are among the FAQs and doesn’t provide an easy path to human help for those questions that aren’t, can easily get frustrating.

3) Domain of typical FAQ content

FAQs are by nature questions that have answers which are in the “public domain”, i.e. can be put on a website and apply to everyone. Oftentimes however, customers come with questions such as “where is my order“, or “I need to change my upcoming appointment“. FAQs do not answer these and instead point to a place on the website where you can log in and find an answer. If a customer service chatbot does the same rather than actually tell you where a customer’s order is or asking what day they would rather come in for their appointment, you produce friction and a break of medium, which doesn’t help the experience. Customer service chatbots need to integrate with your CRM and other customer-related systems of record to make them truly useful.

4) Nature of typical FAQ content

Websites are media-rich environments that allow for a high-fidelity display of information. There is no boundary to the amount of information you can convey, nor to the format within which it can be conveyed. However, chat is quite the opposite: chat is a medium that is based on the idea of a conversation and constrains the information throughput that can be achieved at a time. Copying & pasting content from the website into a messaging bubble is not conversational.

As part of the transition of FAQ content into the chat medium, consider your message and conversation interaction design. Neutral formulations such as “Customers can register here:” should be converted to second person singular: ”You can register here:”; longer messages – and they sometimes cannot be avoided given the subject matter – must be split up into smaller pieces to fit into the constraints of Facebook Messenger or SMS.

Finally, it is not enough to teach a bot how to understand the question if asked (mostly) verbatim as it’s represented in your FAQs. Questions in FAQ sections are designed as broad questions with all-encompassing answers, and they are worded in “written language style”. Chat, however, means conversation, gradual discovery, and colloquialisms. Where an FAQ might formulate (somewhat awkwardly) “What are the conditions for the utilization of the adapter“, a real person might ask “what do I need so I can use the adapter“, or simply “how can I use it?“…

5) Intent of an FAQ section on a website

FAQ sections on websites are meant as an additional source of help, among others. Sometimes they are meant to be an entry point into the content of the website. But they live on your website, and are thus focused on information there, often pointing the user to other places on your website. A customer service chatbot, say one that lives on Messenger or SMS, greets the user with a “blinking cursor”. It is the customer that sends the first message, and that message could be anything. There is no way for you to guide the conversation; you have to respond to what the customer is sending which might be a simple “public domain” question, a CRM-type account question, or a complaint. You can overcome this hurdle by offering human backup for all those things the chatbot wasn’t designed for. Not doing so runs the risk of creating something that worsens rather than helps with your CX.

Pulling in FAQ content into your customer service chatbot is a good idea – the more information the bot has at its disposal, the better. By bundling FAQs with the questions you are already handling on other channels, say your IVR, integrating it with your enterprise backend systems, and applying the rules of the CUI, you should be able to reap the rewards quickly. Wondering what the business case of a customer service chatbot could look like? Have a look here. And if you need ideas for how to get started designing your own bot, have a look at our resources on www.aspect.ai.

As a next step, why don’t you check out your company’s FAQs and think through what it would take to “botify” your FAQs?

avatar

Tobias Goebel

Tobias is Director of Emerging Technologies at Aspect. He has over 14 years of experience in customer care technology and the contact center industry with roles spanning engineering, consulting, pre-sales engineering, program and product management, and product marketing. As part of Aspect's product management and marketing team today, he works on defining the future of the mobile customer experience, bringing together channels such as mobile apps, messaging, voice, and social. He is a frequent speaker and blogger on topics around customer service and, more recently, the (re-)emerging chatbot, NLP, and AI technologies. Tobias holds degrees in Computational Linguistics, Phonetics, and Computer Science from the universities of Bonn, Germany and Edinburgh, UK.
avatar