How to decide when self-service bots should call human assistants for help

In the enterprise support field, there’s a trend to have self-service bots replace human assistants.

In fact, by 2020, Gartner estimates that customers will manage 85 percent of their brand relationships without talking to a human, and brands are answering this call for more self-service. However, there will occasionally be a point at which self-service tools alone will not satisfy. Sometimes circumstances call for a live, human interaction.

Virtual assistant chatbots are usually the first point of contact. A customer might be looking for tech support for a cell phone plan or inquiring about a new bank account. Many questions can be automated for the chat session.

While chat engagements with human agents can be more effective and have a higher first-time resolution rate than self-service, it can get expensive because the brand has to pay a human being to handle each and every interaction.

The right balance

An ideal intelligent bot enables consumers to talk to it using natural language, and knows when it’s time to get a human involved. There has to be a balance between the initial dialogue and inquiries, and a point when a human has to take over.

The first step is for the virtual assistant to be able to understand the intent of a question. The bot can ask clarifying questions, if needed, and reference the context from the entire conversation, thereby personalizing the interaction.

Just as important, it should have no problem with slang or typos. For example:

VA: Hi, I’m a virtual assistant. How can I help you today?
You: I’m having problems signing in to my account.
VA: Are you having problems with the username or the password?
You: With my psswrd
VA: You can reset your password on this site. A new one will be sent to your email.
You: I’ll try that. Thanks.

In that example, the chatbot was able to understand the customer meant “password” in the chat.

The smart VA’s own language should sound natural, utilizing scripts from common live chat interactions. What were the most frequent questions, and how did agents most successfully answer them? Brands can learn from full-service experiences that assisted customers effectively and incorporate them into their automated conversations. In essence, the live chat agent becomes a robot trainer and, over time, can make the virtual agent better at answering questions and understanding content.

Smooth transition

Of course, there are times when chatbot interactions and human assistance overlap. What happens with a customer whose circumstance doesn’t match any of the pre-formulated conversations? It’s always annoying to have to restart a conversation with the bot.

A smart A.I.-powered chatbot should be able to instantly connect to an actual person within the same customer interface, preserving the context of the conversation. The human assistant should be able to pick up where the conversation left off.

Even more importantly, the human should take over at the optimal moment (before things get frustrating for the consumer). The person chatting with the bot or human might not even know which is which.

When are these optimal moments?  A virtual assistant should be able to tell that a consumer needs to be transitioned to a live chat agent when:

• The VA is not able to understand the user (sometimes after several ways of asking the same question)
• The user messages or clicks that the answer provided did not answer their question
• The user asks to speak with a live human being or a contact other than the bot
• The user expresses frustration / dissatisfaction with the VA

Here’s an example of how this might work:

You: I’m still having problems signing in to my account.
VA: What is the email provided in your account?
You: me@yourbrand.com
VA: I can’t find this email in our database. Can I connect you with a live agent?
You: Yes, please…

You are transferred to a live agent

Jessica: Hi, I can see that you have issues logging in with the email address me@yourbrand.com.

This escalation to a chat agent would happen inside the same engagement window.

When the automated-to-human transition can be done naturally and seamlessly, the consumer will embrace the artificial intelligence involved in the customer experience, rather than be deterred by it.

Adjusting the scales

In order to keep the balance between technology and the human touch, brands need to constantly evaluate the chat experience and improve the conversations. Analytics from self-service tools, live chat, and A.I. conversations should be combined to optimize interactions that lead to increased revenues, lower operating costs, and higher customer satisfaction.