Conversational UX for ecommerce: how human is too human?

With the tightened financial times we’re currently living in, businesses are looking to find ways to work more efficiently with what they have and get more value from their existing customers. This has often led to ratcheting up automation across industries. Companies are developing AI solutions to make their online stores more efficient, from automating customer service to pushing for improved conversational user experience (UX).

Interest in these technologies is growing. For instance, Statista predicts that the chatbot market will reach $1.25 billion (€1.1 billion) in the next five years, up from $190 million (€168 million) in 2016. Customer trends reflect this; a recent report by Zendesk has found that ticket reporting is rising across all channels, with three-quarters of customers expecting instantaneous responses.

Building a chatbot, however, is a daunting prospect. Regardless of size, you probably don’t have the resources to build chat functionality from scratch. But if you’re using a platform with premade content and an easy-to-use builder, like Certainly, your job will become 100x easier.

What is Conversational UX?

Conversational UX (not to be confused with conversational UI) encapsulates the user’s experience as they interact with your website. The ultimate goal of good conversational UX is to enable your customer to complete their interaction without resorting to any means other than the conversation and with as little friction as possible. A Certainly chatbot, for example, can inform the user of a product, help them find the right fit or color, add it to the cart, and support their checkout, all from the chat window.

Beyond this, the more human-like your bot is, the better your conversational UX. This is all very well, but people have been working for over 50 years to make human-like chatbots, so let’s look at some ways you can develop this element of Conversational UX without building a sentient AI.

Several things will make your customer relate to your chatbot more early on. One that a lot of our customers do is name their chatbot. A great example of this is Feastables’ FeastyBot; a brand mascot that speaks in the brand’s voice. This creates a connection between customer and chatbot, making them feel as if they are speaking to an actual person, even if it first says, “I’m a chatbot” (more on that later).

MrBeast's Feastables FeastyBot chat widget showing the "Secret Door", a fun section of the chatbot with jokes and social media.

Similarly, you don’t have to make your chatbot speak like a robot; write its script as if it was a human agent! Have it speak in the first person, use a bit of slang, or give them emotive responses. For instance, instead of saying, “This belt will match these jeans,” have the script say, “Oh, do you know what? I think this belt will go GREAT with those jeans!”. Same meaning, but much more personal.

AI-powered chatbots

These linguistic tricks are just the start. Firstly, creating a “contextual chatbot” is one of the best ways to humanize engagement. This is a bot that is aware of what has already been said in the chat and uses this information to produce a smoother user experience.

Suppose the chatbot cannot remember a customer request from a few messages ago, let alone the last time they visited your webshop. In that case, the interaction will frustrate the customer, and they might even abandon their purchase. The chatbot will also be unable to cross-sell, given that it won’t be able to connect what the customer might want with what they’ve already purchased.

Secondly, a must-have is a chatbot built with NLU/NLP (Natural-Language Understanding/Processing). NLU is the ability of your bot to understand and respond to natural (human) language using context, pre-built dictionaries, and learned responses.

the introduction flow of Quadbot, the chatbot of Quadlock.

Instead of guessing which specific words your users might use and populating individual responses to each word or phrase, a chatbot with NLU capabilities can respond to groups of terms based on tone or theme. This will save you time and resources and reduce the risk of dead ends in the conversation or the chatbot misunderstanding the user.

This may seem a bit complicated, but some of the more user-friendly builders have these functions out of the box, trained on ecommerce with pre-built intents.

Things to avoid for Conversational UX

This is not to say you’re trying to pass the Turing Test. In fact, there are plenty of things you should avoid doing when making your customer service chatbot.

For instance, it’s crucial to inform the user early on that they’re talking to a chatbot. It establishes the conversation’s parameters and helps build trust with the customer. If they think they’re speaking to a human operator and, suddenly, the chatbot can’t deal with their request and offers a handover, the customer will potentially lose trust or get frustrated.

Another thing to avoid is trying to make the scope of your chatbot too broad. For example, suppose your bot is intended as a customer support rep. In that case, it only needs to be able to respond to customer service-related queries. It shouldn’t be making small talk or dealing with irrelevant questions. The main reason is that it streamlines the process as much as possible, both development and user experience.

Having some non-utilitarian options for the customer is fine, of course. For example, Siksilk’s chatbot, Melo, will tell a fun story about how it used to be a voice haunting their Scarborough office until they hired it. The crucial part is to give the customer what they ask for.

SikSilk's chatbot Melo telling the user its life story.

In the same way that not being able to grasp the context of the conversation will lead to friction in the interaction, as will the chatbot interrupting the flow to tell a joke or make small talk.

A delicate balance

Conversational UX is a tricky balance of human-like habits and automated responses. Ultimately, you want to create a friendly chatbot that users can engage with and which creates value by helping users navigate product selection or helping businesses decrease cart abandonment.

Fergus Doyle wrote this article with visuals by Vital Sinkevich.

Related blog posts