It can be a mighty task to create a conversational user experience (UX). To be sure, there is plenty of hardware and software on the market to support the foundation of conversational UX, such as voice and chat box technology. But, the very definition of conversational UX requires that this experience has to mimic, as closely as possible, an actual human conversation with all its fits and starts and inferences to context.
“Conversational experience goes beyond simply knowing order history or addressing a consumer by their first name in an email, it adds a layer of human connection between brand and consumer,” said Lauren Pietersen, director of products & partnerships at HelloWorld.
Starting With the Basics
Getting there is not easy — and there is a case to be made that the ideal conversational UX does not exist yet — and it is all but impossible unless the requisite tools are at hand. More often than not, those tools have to be supplemented with personalization and customization products and deliver something that resembles conversational UX. Pietersen gave the example of texting, a channel that many consumers prefer over voice. “Fortunately for businesses, companies like ZipWhip have made it possible for businesses to text consumers using their existing landline, VoIP number, or toll free number,” she said. “This type of simple enablement, combined with emerging technology like Rich Communication Service (a new online protocol meant to replace the current texting standard SMS with messaging that has more multimedia capabilities), makes convenience, immediacy and personalization a reality.”
After all, Pietersen continued, why call your gym to reschedule a class or open a bespoke app when you can simply send a text message instead and get one-on-one service?
Related Article: What Is a Conversational Experience?
Controlling the Conversational Flow
When building a solid conversational UX, Brian Bagdasarian, head of conversational growth strategy at HubSpot, likes to say ‘craft the narrative first, build the bot second.’ He bases this on having been involved with the development of literally thousands of chatbots. “There are multiple platforms out there that allow you to build the technical side of a chatbot, or conversational UX,” he said. “The art of the process comes not from that, but rather from being a great non-linear writer.”
Why is that? Because a great conversational UX creates the illusion of free choice, while in reality it is scripted in a way that controls the flow, or path, of the narrative and its related experience, to move the human user forward. “Taking the time to plan the conversational narrative, identify where certain questions will be asked, and how they can be answered, and planning ahead to handle the natural human errors (like typos) that can occur are all core to creating a strong conversational UX.”
To do this, creating a flowchart of the conversational narrative is essential, Bagdasarian said. “Many tools will do this by their inherent design — you design the flowchart as you go. That said, tools like BotSociety.io allow you create powerful, interactive mockups of your conversational UX, and share them with a team for feedback and revision.”
It’s also important to know where answers to questions that may be asked reside, and will be pulled from, and how that knowledge is going to be organized, he said. For example, will the bot use an API, or will answers be hard coded? Will it need to be able to transfer to a live person, or trigger an email to be sent? “You can start with nothing more than a pen and paper, but the better the plan, the more effective the end result,” Bagdasarian said.
Having the answers in disparate sources can complicate the process — although that is how many businesses have their data structured — so at the very least a company should have a centralized place for user data, such as a CRM system, according to Connor Cirillo, conversational marketing manager at HubSpot. “This will remove friction, like not having to ask the same question more than once,” he said. But the most underrated way to create a great conversational UX is to make it feel natural, he added. “Roleplaying the flows you've written with a teammate will identify blind spots. Forcing yourself to stick to the script will surface updates you need to make.”
Related Article: Creating Great Conversational Experiences Has Its Challenges
Adding Machine Learning/AI
Consumer UX does not necessarily have to include artificial intelligence (AI) and machine learning — Pietersen’s example about rebooking a gym class via text illustrates that — but it is very helpful. Indeed, it is highly doubtful that even the most advanced personalization and customization tools can create the natural, free-flowing back-and-forth between company and customer expected in a conversational UX. However AI can.
For that and other reasons, AI is now being deployed in chat channels such as web chat, SMS automation and Messenger, as well as branded chatbots. “When AI is machine learning-based and can expand its capabilities and understanding over time, consumers will have a more natural, conversational and satisfying experience,” Pietersen said. The goal with AI interactions should always be to reduce the friction of interacting with your brand, and to narrow the funnel of interactions that actually need human intervention, she said. “By training an AI bot on your brand’s particular nuances and integrating it via web services to enable instant access to critical data such as shipping information, product recommendations, deal notifications, loyalty program information, account balances, hours of operation, FAQs, return processing, order placement etc., you will effectively serve most consumer’s needs.”
Don’t Set and Forget
Conversational UX is not a static application and therefore requires constant monitoring. Hence, performance analysis once the program goes live is critical, said Mike Smart, R&D engineer at Jadu. “The nature of natural language means that we can't simply ship a service and forget about it; the service will need to adapt to multiple factors that are always moving,” he said. “Whether it's the context that the service is deployed into, or the target audience it is designed to serve, we need to be iterating and monitoring its effectiveness regularly.”