One of the key issues plaguing chatbot and conversational AI developer communities is the quality of experience it delivers to users. The answer to the problem could come from UX. In this article, let us explore a popular tool in UX circles called the UX Honeycomb and apply it to enrich chatbots and conversational AI experiences.
Peter Morville’s UX Honeycomb is a very useful tool to qualitatively measure the kind of experience your product or service offers users. Originally designed to improve website experiences, it identifies various factors that need to considered in designing a product or service in order to provide a great user experience. These factors are shown in a hexogonal pattern and hence the name — honeycomb.
The elements that make UX great are as follows:
- Findable — Is your product easily findable? Are the product’s capabilities findable?
- Accessible — Is your product easily accessible in the face of environmental constraints in which it may be used?
- Useful — Is your product actually useful to the user?Does it actually solve any pain points of the user?
- Usable — Many products are badly designed that although useful, they are not usable. Is your product usable?
- Credible — Is your product providing credible information? Is what it says true? Is it trust worthy?
- Valuable — Is your product solving user pain points that are actually valuable? Does it save time or considerable amount of money?
- Desirable — Is your product giving the user an experience that the user desires? Would she come back for more? Would she tell her friends about it?
Well then, how can UX Honeycomb be used to enrich CONX? Let’s look at each of the above elements and explore how they can be used to provide a great conversational experience.
Is your chatbot easy to find? While it is easy to introduce a phone based or web based service, it is not the same with chatbots. If you create a new website and provide the URL to your customer, he would know straight away that he need to type that into a browser to get to your website. Chatbots are tricky as they can be embedded into many channels — websites, mobile apps, messenging services, telephony, SMS, etc. So it needs to clear to the user, where and how the chatbot can be found. It makes sense to embed your chatbot in channels that your customers already use so that they can easily find it — dialling your phone number could connect the customer to a voicebot as first point of contact or livechat could be enabled to house your chatbot as a 24×7 default service.
Another findability concern would be that of customers finding about your chatbot’s capability. How would they know what the chatbot is capable of?What can it do? What are its affordances? This needs to be clear so as to avoid confusion about what the chatbot is actually for. Many chatbots start with a greeting message followed by a generic prompt — ‘How may I help you?’. This is a recipe for disaster. For starters, chatbots are still a novel concept. Many customers would not have had any experience using chatbots. Since they appear to be on conversational channels, the default assumption would be that they are capable of all kinds of conversations. So its a good idea to give customers a hint of what the chatbot is good at and how they can access a human agent if necessary. In short, make your chatbot’s capabilities readily discoverable.
Is your chatbot accessible? This is where one must think of scenarios in which and devices through which your chatbot could get used? This does relate to user goals but most importantly we must also consider their environmental constraints. For instance, if you are building a recipe chatbot to help someone cook, then it is obvious that the conversation is going to happen in a kitchen. The user is going to be doing the cooking and his hands are going to be engaged. So if the chatbot requires the user to type, then it is not accessible. It probably needs to be voice enabled. If the conversation is going to happen in a crowded space, then the chatbot needs to allow for text inputs rather than voice inputs. Conversation can happen over different modalities — face-2-face, voice, free text, buttons, or a combination of these.
In addition to mainstream users, think about extreme users as well. How can people with impairments access your service? Would a visually impaired user be able to interact with the chatbot using voice? Would a hearing impaired user be able to interact using text? Make sure you got the right choice that suits the user’s capabilities, environmental constraints, and goals.
How can you make your chatbot usable? One of popular complaints that you will receive after you make your chatbot live is that it is not usable. There may be many reasons for that — not understanding what the user is saying, not knowing how to respond, not knowing what to do with the user’s request, not knowing about its own limitations, and even if it does, not know how to communicate it. Conversation is a complicated affair. We don’t know it because the complexity of conversational dynamics is largely invisible to us. And when a machine talks to us, we tend to think that it knows how to talk as well as we do.
The tech is not there yet. We just broken two barriers — recognising what the user says and understanding them to a limited extent. There are many such barriers to cross — what to do with the layers of information the user just (knowingly and unknowingly) provided us and how best to respond. The happy path is the easiest to design. But the unhappy paths are a challenge. Creating a chatbot that is usable hinges on good design at the moment rather than technology. Focus on user emotions and try to not leave your users in negative emotional states — i.e. sad, frustrated, or confused. Long story short, make it easy for users to recover from errors.
How useful is your chatbot? Is it solving any of user’s pain points? Chatbots need to have clear goals and provide clear value to the user. E.g. book tickets, submit an insurance claim, money advice and so on. You could measure its usefulness by measuring the number of conversations closed successfully. While many chatbots do focus on features like chit-chat, socialising, making jokes, etc, they are to be seen as additional features to being useful to users. A chatbot that makes chit-chat can look lively and interesting, but users will get bored after a while as it is not being useful. So having clear goals and ways to measure them can increase the system’s usefulness.
In addition to being useful and usable, your chatbot needs also to be credible. Credibility builds trust and trust is an important element of UX in the long run. Is the information provided by the chatbot accurate? If not, does the chatbot makes the user aware of that? If there other channels like a website dashboard or a mobile app that provides the same info as the chatbot, make sure that they all sync up. Users could express a lot more their thoughts and feelings to a chatbot as its a conversational interface. If the information provided by the chatbot is different from that of other channels, it should be prepared to have conversations about the discrepancies too. Make sure the information it provides is not outdated or at least say things like “last time I checked..” to express low confidence if the information might have changed. Providing consistent and accurate information to the user improves credibility and therefore make the service trustable.
Another important factor that improves trust is to keep promises and manage expectations. For instance, if you promise to keep user’s data safe and private, make sure you do. If you promise to get back in a few minutes with answers, make sure you follow up. Manage expectations by telling users clearly what your chatbot can and cannot do. Open prompts like ‘How can I help you?’ are misleading. So try ‘I can help you with blah blah.. What would you like me to do?’ Developing trust may be the hardest thing to do, but doing so improves the relationship your chatbot will have your customers.
Is the chatbot providing value to the user? This is related very much to usability and usefulness. These factors could result in the user saving time or money or both. It could make chat service available 24×7, and thereby improve your company’s brand image as being customer centric. Your chatbot conversations could result in behaviour change in users which could add value to users, their families, and to society at large. While the chatbot achieves its goals, it also provides measurable value to users, your business and society at large.
And finally, your chatbot and the experience it provides need to desirable. It means, the experience need to be in such a way that it makes your customers come back for more. It needs to delight them in such a way that they will want to tell their friends about it. To create such an experience, you will need to go beyond usefulness, usability, credibility and other elements discussed above. This one is really the icing on the cake.
To deliver a delightful conversational experience you could try a number of things. Firstly, you could use visually appealing user interface and graphical elements that complement conversational elements. Secondly, provide your customers a personalised experience. And finally, create a persona for your chatbot with a likeable personality that aligns with the brand identity of your business. These elements should be designed in such a way that would together create an emotional connection between your customers and your chatbot.
All these elements work together as a pattern and create a holistic experience for users. It is therefore not always easy to get things right the first time. But we need to iterate to identify the right combination by experimenting and observation.
I hope I have encouraged you to think about how to create a great conversational experience using the UX honeycomb. Please do share your thoughts and experience in applying UX tools for chatbots and other conversational AI solutions. Best wishes designing great conversational experiences :)!