The thing is, as with any new technology, everything’s a bit up in the air. We ask ourselves, “What are best practices for user experience, when it comes to bots or chatbots”. Should we just do text? Does it need voice? Does it support localisation and multiple languages? These are hard questions to answer, since building a Natural Language User Interface (Let’s call it LUI for short), is a difficult thing to do, albeit less of a technical task than it has been historically.
On traditional digital platforms, like websites, narrative patterns are established and controlled in how they appear to a user in the form of articles, landing pages, interactive storytelling or through other relevant pathways and user journeys. A chatbot isn’t that different when you think about it, and text or speech can be mapped to an intended action, as easily as a button on a webpage can.
When you ask the chatbot “Show me todays appointments”, it’s mapped to an internal function (fetch_appointments()), that fetches data from a database; formats it and sends it back to the user. Of course there’s a bit more to it with Natural Language Processing (NLP) being used to parse sentences, and gain meaning from it, but the concept remains the same. It’s just another interaction, that happens through conversation instead of a traditional UI. This is an important thing to realise, since it helps us reshape our thinking.
A good LUI can be very open-ended, with a lot of different conversation flows and intended use-cases, but that’s a tougher nut to crack. We get more value by keeping our efforts narrow, and I promise: Developers and designers will love us for it. Take a step back, and consider the overall strategy: Do we want the bot to empower your customer support experience? Should it sell a service or guide the user to an optimal product configuration? Should it integrate with third-party systems, or ERP? It can do all of these things, but if you or your organisation is inexperienced and ill-equipped to build a language-based interface, it becomes all the more important to pick one task, and do that one thing well.
Should the approach become too scattered, we risk creating a mess while making it difficult to control the flow of conversation over time, with flows overlapping. Mapping of user intents to conversation templates becomes increasingly difficult, due to similar triggers (e.g “fetch appointments”) and flows. That lack of focus will quickly become apparent to anyone using the bot, since users won’t find, or receive the answer they’re looking for.
When we have a clear purpose, everything becomes much more obvious, and the user journey is clearly mapped to an intended conversation flow (i.e user says something, bot responds accordingly). Simplicity is key, and acknowledging our own limitations will help guide the creation of something of real value.
So we’ve decided to leverage a chatbot. Exciting stuff! Let’s say the goal of the initiative is to drive sales.
Can the bot augment an existing user journey, or should it be an entrance to generating sales on its own? Driving sales directly through auto-generated conversation flows might be too ambitious at first — we are new to this after all, and it might come off as too head strong for our user. So instead let’s look at lead generation: What’s the primary driver of leads in the organisation? You discover that potential customers interact heavily with the product comparison tool, on the company website, and the resulting report generates a big influx of engaged users. This provides a good hook for the chatbot.
We then identify the key moment of decision, and this seems to be the product overview page. That is the main drop-off point; the potential customer already browsed a few products, and she seems ready to choose the next step, or drop off completely. We want to increase the conversation rate, so let’s gently nudge them along!
We start off with a soft introduction of the bot, since it’s the first interaction with this particular user. The bot quickly follows up with a suggestion.
Should the user choose to take the bot up on its offer, it will provide the necessary comparison, and it’ll send a tailored report to the email the user provides.
Here we identified a need, and are now using the bot as a tool to drive conversions and populate potential leads. This could be an initial approach, there are surely many others, but the important thing is to find a strategy that makes sense for you, and it’s okay to start small. I would even recommend it, since it enables iteration and ideation on a smaller scale, to the point where you build something of unique value to users, while accomplishing business goals that are otherwise harder to do.
But let’s not stop here! You also want to accommodate users that want to interact directly with the chatbot, and are used to doing so on a frequent basis.
Like mentioned before, it’s of course a good idea to capture as many entrances to the chatbot as possible, but being able to cater to all user requests is near impossible. A better way is to try and guide the user by presenting options and common use-cases. That also presents us with an excellent opportunity to direct the conversation where we want it to go.
When a user opens a chatbot, it’s important to make them feel welcome.
Here we’ve presented the user with an entrance and several pathways that can lead to more information.
Even though it’s about conversation, it can be beneficial to add UI elements, like buttons, sliders, forms and galleries to help guide the conversation and provide immediate value. Pure text can quickly become bothersome if the conversation drags out, so providing small shortcuts will be appreciated. This is also a good opportunity to think creatively!
So again: Chatbots are most valuable when they provide guidance, without demanding undeserved attention.
Suggested content tend to get lost, but here it comes natural as part of a conversation
Done right, they can help drive conversions, sales and lead generation. They can offer suggestions based on user preference, without robbing them of precious context in the form of natural flowing conversation. Suggested content tends to get lost, but here it comes natural as part of a conversation — due to how we interpret intention differently visually vs. through conversation. A recommendation from a trusted source, is different than a piece of suggested content in a sidebar. Here we have the opportunity to push, based on an established report with the user. But for this to happen, it needs to be genuine.