4 Essentials for Building a Well Mannered Customer-Service ChatbotIf your chatbot is aggravating customers, human error is to blame.
Opinions expressed by Entrepreneur contributors are their own.
In the 2009 filmHer, Joaquin Phoenix's character falls in love with a disembodied, but seemingly human, voice. Set in the near future, the film examines loneliness, the meaning of love and of how the relationship between humans and increasingly "intelligent" machines will evolve.
Skip forward to 2016, when 250,000 Americansaskedtheir Alexa wireless speaker to marry them. While we're not quite in the realms ofHer, it's evident that a large number of people are starting to form new types of relationships with AI systems. Companies keen to deliver more personalized, responsive services are rushing to leverage AI to facilitate deeper customer relationships. It's why十个里有八个公司now say they either have adopted or will by 2020 be adopting AI as a customer service solution.
As AI moves into customer-facing business processes, companies that previously relied on solely human employees to engage with customers must now focus on training AI-powered chatbots to interact with customers. There is no authoritative rule book yet on how to do that, but I have four recommendations based on my experience building and rolling out AI customer services in retail, education and travel.
Related:Top 10 Best Chatbot Platform Tools to Build Chatbots for Your Business
1. Determine when a person needs to take the call and when a chatbot could.
AI should not be the only way a customer can talk to your company. AI-powered interactions, whether voice- or text-based, augment the ability a company's human agents to deliver personalized, real-time customer interactions. Companies must decide when a human should be the first point of contact for a customer, and when an AI system can do the job.
Two elements should be considered in making that decision. First, identify the top issues that your contact centers currently handle on a day-to-day basis. This helps you prioritize where a chatbot can help offer service efficiency at scale.
Second, determine whether a chatbot could solve one of those issues more easily and quickly for customers. Always build solutions starting from the customer's point of view. Remember, chatbots are designed to provide instant value, rather than deflecting customer complaints. Your chatbot will cause far more problems than it solves if customers begin to suspect it's part of a cost-cutting, deflection measure.
Related:How to Create a Facebook Messenger Chatbot For Free Without Coding
Chatbots tend to be useful for tasks like:
Communications at scale:fielding questions about a service outage.
结构化数据共享:hours of operation; location of a business.
Simple personalized responses:order status; account balance, or even upselling.
Self-service for customers:an efficient approach for initial interaction with a high volume of free-trial customers.
Pre-handoff to a human agent:gathering data like order number or the issue to be resolved in advance of handoff to a human representative to speed time to resolution
Chatbots are not universally applicable though. You should avoid using them for situations like:
High-touch, real-time situations:troubleshooting a piece of technology.
Emotional and regulated situations:security or privacy violations; news about health or finances.
Frequently changing content (like a restaurant menu):it's difficult to update a chatbot to match the data.
2. Tell customers what a chatbot can and can't do.
Chatbots can't solve every customer problem. In fact, attempting to build a chatbot to solve too many problems will lead to a worse customer experience.
Chatbots need to make clear at the beginning of their conversation with customers what they can and can't help them with. Overpromising will ruin the customer experience. The contrasting experiences of Apple and Amazon with their AI-powered voice assistants is instructive here.
When Apple launched Siri, the company promised the world, and frustrated customers when it became clear that Siri had very real limitations. Apple is stillstruggling to make up for that mistakein the eyes of their customer base.
Amazon's Alexa was offered as a home speaker, with a very limited use case. Alexa easily handles queries such as "tell me the time," "set a timer," or "play this music." While Alexa has gotten better over time, those initial constraints helped customers understand how to engage with the system, and begin to integrate Alexa into their daily lives.
Companies should follow Amazon's lead and make clear how and when your chatbot will help customers as soon as they open that chat window or pick up the phone.
Related:Make Chats With Chatbots Work
3. Give the customer easy access to a person.
One of the most important elements of using chatbots is to build an "escape hatch," a mechanism where a chatbot can "hand off" a customer to a human when an issue requires escalation. The leading chatbots are those that can conduct this handoff while maintaining all of the context of the conversation.This handoff process can be one of the most challenging elements of a rollout of chatbot technology.
By tying chatbots to your existing customer information, or CRM systems, the handoff can be smoother. Linking to CRM systems can provide a chatbot with information on what agent is available, with what skills, at any given time. It also means that the data it collects can be automatically loaded into the CRM system.
Related:Chatbots Are the Next Big Platform. Here's How Entrepreneurs Can Stay on Top of It
4. CRM data will turbocharge yourchatbot.
AI systems learn and improve by spotting patterns in data. However, too few chatbots nowadays take advantage of, or have access to, historical data for training. As a result, chatbots can require months of collecting and manually preparing training data, and they only offer the most general advice. With access to historical CRM data for training, the amount of time needed to get chatbots up and running is significantly reduced. In addition, a chatbot with access to rich CRM data has the context to deliver a more personalized conversation for each customer.
For instance, let's say you run a food delivery service and a customer requests a chat session less than five minutes after their last order. The chatbot should know about the customer's history and that they are likely calling about the most recent order, prompting the chatbot to ask, "Are you contacting us about your order? What do you want to do with it?"
For companies looking to embed AI-powered chatbots, or voice assistants, into their customer engagement systems, these four pieces of advice will put them in a stronger position to leverage AI to deliver more personalized, useful, and immediate service -- and maybe even solicit a marriage proposal.
