Designing a good conversational experience for a machine isn’t easy. There’s a delicate balance between being too human and not human enough. A chatbot is artificial, but it can provide a positive end-user experience. Designing an exceptional conversational user experience depends on creating a fluid and natural interaction between customers and a machine (chatbot).
At Netomi, we have a playbook on how to provide the best end-user experience on chatbots, complete the task at hand (i.e. resolving customer service tickets) and protect the brand. As you look to design conversational user experiences for AI customer service chatbots, use these 7 strategies to provide the optimal experience:
1. Define the bot’s personality, tone and approach
Start the conversation with a warm, informative welcome message to provide a positive, immediate first impression. Give your bot a name and keep all copy on-brand. If you’re a hip, young brand, train the bot to speak the way your customers do. If you’re a more convervative brand, then ensure your bot’s tone does not clash with your customers’ perception of your brand.
2. Be prepared for “urgency” and eliminate friction
When a customer uses a chat channel to get in touch with a company, they want a swift resolution to an issue. They may be unhappy or frustrated, wondering where their delayed package is, requesting a refund for a broken product or trying to figure out a flight delay. Chatbots should be designed to alleviate tension and move the customers towards a resolution quickly. The bottom line? Don’t cause delay and help right away. Cut the small talk, ask only critical questions and provide a resolution as quickly as possible.
3. Don’t try to be human, but be as human as possible
While the goal to provide good conversational UX is to communicate with humans as effectively as possible, it’s fundamental to realize that bots will never fully replicate humans. When a customer begins a conversation, a bot should confirm it’s identity by saying something like “Hi! 👋 It’s Emma, your virtual assistant!” Never try to pass off a bot as a live agent.
That being said, bots can still promote human-like interactions. A few ways to do this include:
- Never show the customer the same message two times in a row; this annoys humans
- Understand emojis (👍), slang (‘sup) and short-form text (LOL!)
- Accommodate common misspelled words (where’s my ordr?)
4. Allow handoff to a live agent anytime and anticipate when a handoff is required
Always give the user a way to opt to engage with a live agent. On the most basic level, a chatbot should always comply if a customer asks explicitly “Can I talk to a human” or “Human, please.”
Certain customer queries should always be routed to a human agent. These queries are the complex, sensitive issues or ones that require empathy. It could be managing the rebooking of a gold-star rewards customer or managing the return of an item that costs more than $100. For each brand, the rules that define a sensitive query are unique.
Furthermore, chatbots should proactively hand off the conversation if it can’t understand the user’s intent, or what the customer is trying to say. Having to repeat a question to a bot is often cited as the biggest source of frustration by consumers. If a bot can’t confidently understand a person, it should ask them to rephrase their question once and then immediately escalate. Finally, hand-off to live agents should occur automatically and immediately if a chatbot picks up that a customer is irritated, frustrated or angry. Some obvious signs might be typing in ALL CAPS or using tons of punctuation!!!!!!!!!! More modern and advanced bots have sentiment analysis capabilities that can help detect and reroute upset customers to humans for faster resolutions.
5. Keep your bot’s responses short and to the point
Keeping a conversation moving forward is very important. Bombarding a customer with a lot of options, or lots of text to read can confuse them and make it harder for them to respond. When designing the experience, keep messages succinct; send a maximum of three short messages at a time. And keep options to questions or cards in a carousel to less than four. You can of course use A/B testing if you want.
6. Leverage available content and knowledge bases
Many companies have rich FAQ sections and self-help libraries. When you’re looking to design the Conversational User Experience, leverage existing content as much as possible. This could be used to initially train a chatbot. Many companies also integrate into their knowledge base as a safety net. If a bot hasn’t been trained on a specific question or is not highly confident in the intent classification of a customer’s message, a bot can pull relevant articles from a knowledge base. This could result in even higher deflection from human agents.
7. Train for inappropriate, illegal and sensitive topics
If a user starts talking inappropriately, threatening to harm themselves or others, chatbots should end the conversation immediately. Likewise, if a customer starts discussing illegal topics, the conversation should be halted. Trolls are an unfortunate reality of our digital world. Training bots to handle inappropriate and illegal conversations protects your brand from inadvertently engaging with bad actors or creating deeper problems for your organization.
Chatbot adoption is rising: 77% of executives have already implemented chatbots and 60% plan to implement conversational AI bots for after-sales and customer service. How you design the conversational user experience influences the success of your AI chatbot. Every channel requires a different approach to deliver high CSAT. If you follow the above seven tips, you’ll be providing the best possible user experience.