phonlamaiphoto -


Best practices for initiating chatbot-to-human handoff

Customers can become frustrated with chatbots, so it is important for contact centers to design AI that can recognize this and perform a timely handoff to live agents.

Chatbots aren't always capable of understanding what a customer needs, so it is important for contact centers to program AI to recognize when it's necessary to perform a chatbot-to-human handoff.

Recently, I was down to the wire for hitting a very important deadline. I needed to set up an account for a fantasy football league, and the live draft was happening at my neighbor's house in an hour.

The league's website locked me out of the account I had used for years. I tried to reset the password three times and created a new account using a new email address. After all my efforts failed, I turned to a chatbot for assistance. The bot repeatedly sent me links and information that did not solve my problem, and I'm pretty sure sentiment analysis detected my irritation. When I typed, "This is frustrating," a pop-up asked if I would like to chat with a live agent. Thankfully, after speaking to two live agents via chat, I was able to resolve the issue.

This example highlights a common challenge that companies face when offering chatbot services. Despite advances in AI technology, companies simply can't build a bot that can communicate and understand information in the same way a human can. Instead, contact centers need to configure AI to better understand when an issue is too complicated for a chatbot, then determine the best way to handle a chatbot-to-human handoff. 

Contact centers should establish guidelines and best practices for a chatbot-to-human handoff by focusing first on the overall customer experience and criticality of the customer request.

Customers should always feel in control.

Here are a few examples of these best practices:

Don't make customers wait before offering a live agent option. Customers should always feel in control. Often, they prefer the chatbot's services over human help. But if a customer gets annoyed with a chatbot, they will only get more frustrated with having to wait longer to solve the problem. Applying sentiment analysis that can recognize keywords, such as "frustrated," "stupid" or "help," will assist in making a handoff to a human agent at the right time.

Design a bot that can make complex connections. In the fantasy draft example, the bot identified the key phrase "reset password" and continually offered password reset links. Had the bot connected those keywords with another relevant expression -- in this case, that I had already reset the password "three times" -- it could have more readily initiated a chatbot-to-human handoff.

Don't make customers repeat themselves. Contact centers should ensure bots capture all the information about a customer interaction, then transfer it to the agent alongside the customer. Customers do not want to repeat information after a handoff.

Add a personal touch. Companies should design a chatbot that creates a positive customer experience by incorporating elements that any good customer support interaction should have. Tell the bot to thank the customer, say please or ask, "How may I help?" Make the chatbot sound human. A nice personal touch in my example might have been to have the bot say, "Good luck with your draft."

Would a better bot help my fantasy team? No, but it would help me as a consumer have a better overall experience. AI used correctly can deepen the human relationship and build brand loyalty, but AI that fumbles can ruin the customer experience altogether.

Dig Deeper on AI for customer service