Advanced digital capabilities in Artificial Intelligence (AI) and Natural Language Processing (NLP) will alter how people interact with data. The growing adoption of voice platforms like Amazon Alexa and Google Assistant, as well as messaging platforms like Facebook Messenger and Slack, means humans will interact increasingly more with technology through conversational interfaces (CI), rather than other people.
As the technological challenges of CI lessen, we face new, broader challenges: how can humans interact with robots in a way that doesn’t diminish our own human experience? Building on lessons learned in the past, West Monroe takes a human-centered design approach to offer a framework for the future. We’ve seen a pattern with emerging technologies in the past. In recent history mobile apps, and prior to that websites, and even earlier with physical products—there is a habit of letting technology lead the experience. But, just because you build it, does not mean they will come. Without wide customer understanding, it’s difficult to demonstrate the value that new technology will have on broad business initiatives. To do so adequately, we must start with the human experience: identify a problem or need in a customer’s interaction with the business, and then figure out how we can best leverage technology to achieve a lasting solution.
Users don’t care that they’re interacting with a conversational interface, or that you’re using the latest technologies. They care that their needs are being met—quickly, reliably, and with as little effort on their part as possible. When designing chatbots, this must always be the ultimate goal.
I’ll discuss 10 best practices to follow when approaching a conversational interface design. Guided by human-centered design methods, these best practices will help provide a seamless user experience to meet your customers’ needs:
Best Practice 1 – Scope: Bots can’t do everything, but they can do some things really well.
Look for areas where the level of technological capabilities aligns well with use case needs. Areas where bots can be especially effective include onboarding, transactions, task completion aid, quick queries, brand promotion, gamification, frequent customer inquiries, and market research. Start with a focus on one of these areas to gain the most traction from users.
Best Practice 2 – Challenges: Understanding context is key.
Something humans are very good at, but bots struggle with is understanding the bigger picture of interactions. However, if bots are to aid in solving problems effectively, they must be aware of and act appropriately for the context of use. You must design them to establish empathy and trust with users by being transparent with their capabilities, and give users a sense of security through authentication.
Best Practice 3 – Channel Choice: Meet users where they already are to be most effective, or use a channel that reinforces the company’s other business initiatives.
Determining what platform to use needs to be one of the first decisions in a chatbot design. The way we speak is very different than the way we write. And even within that, things like text messages, email, Facebook, though written, all have their own rules of communication. In order to design an effective chatbot, the idiosyncrasies of the channel’s unspoken rules must be accounted for (for instance, consider the widespread use of abbreviations over SMS, emoji in Slack and Messenger, and full sentences with salutations over email). Having a multi-channel chatbot that’s effective really requires redesigning the conversation flow and structure to adapt appropriately.
Best Practice 4 – Personality: Who is your bot? Chatbots are more than a persona, they need a fully developed personality to be effective.
People will naturally humanize bots, so an opportunity exists for businesses to be intentional and control this narrative. Determine “who” the bot is upfront to enhance overall brand strategy, engage with consumers in familiar ways, and maintain consistency. It may help to give your bot an archetype first and build from there—e.g. a helpful teacher, a knowledgeable advisor, the friend who listens. This will directly impact the bot’s conversational style and will govern the response type, messaging, and chosen platform.
Best Practice 5 – Introductions: Nice to meet you! Bots should always introduce themselves and begin with a simple question to engage users.
Hook users with short, simple messages that either ask a question or inform the user about the bot’s capabilities. Start simple and suggest capabilities through intuitive examples versus taking time to teach the users what to say in an unnatural menu-style format listing out everything.
Best Practice 6 – Dialogue Structure: Conversational Interfaces should follow natural conversational patterns, anticipate user intent, and account for user variability in speech.
A bot should mimic natural dialogue, including turn-taking and the ability to provide enough novelty and variability with responses to maintain engagement. Start by identifying the goal of each specific interaction, and then write sample conversations to map possible scenarios as if two humans were speaking.
Best Practice 7 – Message Length and Frequency: Bots don’t need to share everything at once.
As with human conversations, interactions should be brief to avoid information overload and allow users time to process. Break messages into smaller pieces and prioritize the information given to users. If written, a user should not have to scroll to read critical information. With voice, follow a natural conversation style with pauses for user responses–no one likes speaking with someone who dominates the conversation!
Best Practice 8 – Error Correction: It’s ok not to have all the answers, as long as the bot is upfront about limitations and always responds.
Bots must adapt to different modes of conversation from different users, and be truthful when unable to answer requests. Users should always be asked a confirmation on critical steps, and they should be given the opportunity to fix misunderstandings. This builds trust and allows users to feel confident with the bot’s capabilities. In areas where frustrations could run high, giving the user a way to interact with a human directly if needed may be essential.
Best Practice 9 – Don’t boil the ocean: Start small and be realistic about what you can achieve.
Look for small proof of concepts that demonstrate business value first. Use these interactions to develop reusable patterns and refine all the previous best practices mentioned to align with your users. Build on successes to take on more complex and expanded scenarios.
Best Practice 10 – Test, test, test: Bot design is an iterative process, and refinement should be an ongoing effort, even after release.
Put users in front of the chatbot, and listen to their interactions. Use transcripts to see both sides of the conversation or shadow people as they’re using the bot. Note the specifics of what people say. If repetitive user missteps occur, additional conversation inputs may be needed to make the bot functional. If there are errors, is the bot able to correct or direct the user to a human to take care of their needs? Bots built with natural language processing can be trained over time, so learn from your users and include the variations they say in the testing process.
Remember that we’re still in the early stages of chatbot adoption, and the business reward may not be immediate. However, with rapid technological advancements this change will be coming fast, so the important thing now is to be prepared. Leveraging these UX best practices during early stages will set your business apart, and keep you well positioned to spearhead new and expanding conversational interface technologies.
If you would like to learn more about conversational technologies, feel free to reach out to us for more information. We have experience working with clients to plan, design, and build conversational systems.