By Puneet Mehta, Netomi
Today, providing outstanding customer experience is more important than ever. Netomi’s State of Customer Service in 2021 report discovered that 82% of individuals have stopped doing business with at least one company or have abandoned a planned purchase based on poor customer service. Moreover, expectations for customer support are on the rise, with 63% of customers having higher expectations than they did three to five years ago. What is clear from these findings is that retailers need to adapt to rapidly changing customer needs and preferences, or risk losing customer loyalty and revenue.
Enter: AI-powered chatbots. Currently, chatbots are being used by 1.4 billion people, and voice and chat technologies are predicted to bring in $290 billion by 2025. Handling many of the repetitive and mundane tasks associated with customer support, chatbots allow your agents to concentrate on those tasks that are more complex and fulfilling. With the adoption of AI-powered chatbots on the rise for retailers of all sizes, let’s take a look at five best practices to adopt when using chatbots for customer service.
- Focus on automating the right issues
Not every customer issue is right for automation, especially those that are critical or high-risk (such as returning a high-ticket item, or an interaction with a customer who is noticeably stressed or angry). Begin with training your chatbot to respond to the highly repeatable queries that have a lot of historical data to train with, which improves the accuracy of the chatbot, ultimately providing the best experience to users. Examples include questions surrounding refund policies, requests for order modifications, or order status.
- Select the right channel
When thinking about adopting AI for your business, look for channels that have a high volume of tickets. Where are your customers spending the majority of their time, and where are the conversations and dialogue occurring? As AIs learn from each interaction, deploying your chatbot on a high-volume channel will help your AI provide more value, and also be more accurate.
While live chat is typically the channel that first comes to mind for many retailers, email should also be considered if it’s an existing support channel for your team. Why is this? For many, email is the communication channel of choice - Netomi found that this is the case for 47% of customers, who prefer customer service contact via email, as opposed to phone (23%), webchat (23%), and social media platforms (2%). Email is convenient because the entire transaction is on the customer’s terms, and they need not wait on hold to interact with a live agent in real-time. Additionally, there is a record of the conversation, which can be easily referenced back to, if needed.
- Have a contingency plan
Always offer a well-defined escalation path to human agents, based on topic, user sentiment, customer profile, or when a customer explicitly asks to interact with one. If a customer is growing increasingly exasperated, for instance, a human agent may need to intervene.
- Regularly optimize
Similar to how fostering a culture of continuous learning is essential for workforces of today, this is also the case when it comes to the world of AI. That is, a chatbot’s performance on Day 1 should not be the same on Day 30 of the journey, nor should it be the same on Day 60. Chatbot development is not a ‘set it and forget it’ type of project, as AI algorithms need to be continuously tweaked over time, to account for new data that is added to the system, as well as user priorities and expectations that evolve.
A relevant concept here is that of reinforcement learning, the notion that optimal behaviors or actions are reinforced by a positive reward. For a toddler learning how to walk, for instance, they may experiment with taking smaller steps, if taking broad steps made them fall to the ground. Based on the consequences of those first few actions (i.e. falling), they will decide whether or not to repeat this. It is a similar case with a chatbot, whereby its behavior may be reinforced when it responds correctly or penalized upon misinterpreting a user’s intent. Perhaps a chatbot didn't understand when a certain person asked for their order, in which case, this new utterance should be mapped to the trained topic so it can accurately respond in the future. Throughout this process, the chatbot grows more and more efficient.
- Measure performance and iterate
As with any aspect of a business, having firm metrics in place is essential to gain a clear overall picture, and see how one’s efforts are faring. Establish and track the right success metrics, such as conversation length, bounce rate, and customer satisfaction. Keeping an eye on how the chatbot is used and analyzing its interactions with users allows you to answer key questions such as: what could be improved? Are you seeing successful resolution of customers’ queries? Is the chatbot’s performance hindering the overall user experience, in any way? Being perceptive to these finer details will allow you to improve the chatbot over time and provide the best user experience.
These are several chatbot best practices to add to your radar for 2022. To learn more about how customer service is evolving in the eCommerce arena, including the latest industry facts and figures, view our Customer Service Benchmark Report.
About The Author
Puneet Mehta is the Founder/CEO of Netomi, the world's leading customer experience AI platform that autonomously resolves customer service issues. He spent much of his career as a tech entrepreneur as well as on Wall Street building trading and capital markets AI. He has been recognized as a member of Advertising Age's Creativity 50 list, and Business Insider's Silicon Alley 100 and 35 Up-And-Coming Entrepreneurs You Need To Meet.