Getty Images/iStockphoto


Best practices for initiating chatbot-to-human handoff

Contact centers perform a balancing act between AI and human agents. Support teams must train bots on chatbot-to-human handoffs and design AI aware of customer frustration.

Chatbots can't always understand what customers need, so contact center teams must program AI to recognize when it should perform a chatbot-to-human handoff.

Despite advances in AI technology, building chatbots that can communicate and understand information like a human can is complex. As contact center teams establish best practices for a chatbot-to-human handoff, they must first focus on the overall customer engagement required and customer needs.

Additionally, chatbot providers can configure AI to better understand when an issue is too complicated for a chatbot and determine a point early in the dialogue for a chatbot-to-human handoff. AI could present the option with menu dialogue choices or deduce the right handoff point if the chat lasts a long time without resolution or a customer repeatedly reformats the request, like using the same words in different order.

Potential handoff scenarios

Contact center teams may want chatbot-to-human handoffs to occur based on user preference, issue urgency, issue complexity or scenarios where chatbots commonly fail.

  • Preference. Chatbots should be aware of user context. When dialogues occur within a user session, the chatbot should consider the user's predefined preferences. For example, the user may prefer to interact with a human agent over a chatbot. If user preference is unknown beforehand, the user should see an option to interact with a human agent early on.
  • Urgency. Chatbots relating to travel -- such as airline industries -- must determine urgency when dialogue begins. If the user is logged into the airline system, the chatbot may already have user context and understand the individual's travel itinerary. If the user is not logged in, questions about travel dates and times can determine urgency. Urgency implies transitioning to a human agent early in the chat -- sometimes without offering a choice.
  • Complexity. While chatbots can accommodate many user scenarios, it's nearly impossible to accommodate all. The contact center team has choices. For example, it could develop chatbot dialogues for customers with multiple accounts, typically more complex, so a handoff to a human agent happens immediately.
  • Known fails. Chatbots can fail to resolve user needs like humans can fail to understand another person's intent. If the chatbot fails with the same users or questions, the contact center team should review it. It should also review successful dialogue to see what works. Ongoing reviews can reinforce success logic, adjust for failures and create alternative paths. Agents can report successes and failures through the chatbot KPI to observe trends.
Chatbots can fail to resolve user needs like humans can fail to understand another person's intent.

Steps for chatbot-to-human handoff

Chatbot-to-human handoff should occur quickly and not disrupt the flow of the dialogue. These steps generally abide by the following order:

  • Customer awareness. Chatbots let customers know when a handoff begins, the expected wait time and the human agent's name.
  • Information transfer. The agent begins introductions, summarizes the customer's needs based on information transferred from the chat session and confirms if the request is correct.
  • Resolution. The agent continues the chat, addresses the need or resolves the issue, and asks if the customer would participate in a survey to improve customer service.
  • Closeout. If the customer agrees to the survey, the agent hands the dialogue back to a chatbot or survey mechanism to conduct it and close out the call.
  • Follow-up. The agent annotates the call notes with any areas the chatbot could have done better. The contact center team collects and reviews these notes.

Best practices for chatbots

Organizations that use chatbots in their customer service strategies should ensure the AI seems human -- or, at least, human enough for customers to feel comfortable interacting with it. The following best practices can help contact center teams enable positive CX with chatbots.

  • Don't make customers repeat themselves. Chatbots should capture all the information about customer interactions and transfer it to the agent alongside the customer. Customers do not want to repeat information after a handoff.
  • Humanize CX. Chatbots should create a positive CX and incorporate elements of any good customer support interaction. The chatbot can say, "Please," "Thank you" and "How may I help?" Have the chatbot appear human.
  • Continually improve CX. Contact center teams should map out potential chatbot and customer interaction use cases at the beginning of the chatbot initiative and continue to monitor any AI advancements over time. Chatbot usage can enable new use cases, improve outcomes of existing use cases and identify edge cases. Edge cases are complex interactions where chatbot development can improve CX or where development effort outweighs the value.
  • Call back or leave a message. Offer the customer the option to call back or leave a message to avoid continued waiting.

Next Steps

Cryptocurrency broker uses Ada AI platform for better CX

Dig Deeper on Customer service and contact center

Content Management
Unified Communications
Data Management
Enterprise AI