Managing Chatbot Handover to Human Agent

15th Nov 2018
One of the most common questions a client asks when thinking about implementing a chatbot service is usually, “but how will the bot know when to transfer over a client to a representative? How will it do it in a way that will seem seamless and natural?”

The answer to this question depends on the type of chatbot that is installed within the company.

If you refer back to one of our articles titled “Rule-Based Vs AI Chatbots,” you will find that two of the most popular chatbots to implement are rule-based and artificial intelligence bots.

When it comes to rule-based bots, the transfer from bot to agent is easier and more accurate than with a rule-based chatbot (if a transfer is needed at all). As stated in the referenced article, rule-based chatbots allow for buttons and carousels that can offer users with a clear menu of operations that the bot can perform. If the customer wishes to do something that is not displayed within those option, there is also a button available to be transferred over to a live agent.

This straightforward process of being able to choose between possible bot operations or speak to a representative makes the handoff process easier for the user since they can control when they want the transfer to occur.
When the does handoff occurs, the bot transfers over the information it collected from the user at the start of their interaction in order to give the agent clearer information on who the user is and what they need.

On the other hand, with AI chatbots, carousels and buttons cannot be implemented. All interaction between an AI bot and a user is communicated through text. In this scenario, the handoff takes a little more intelligence from the bot. “Scenario-driven” transfers and natural language analysis are used by the AI bots in order to decide when the transfer to an agent is necessary.

In “scenario-driven” transfers, the bot collects the user’s information and analyzes their request by running it through a query of its programed capabilities. If it finds that it can handle the request, it will proceed with doing so. If the bot concludes that it cannot fulfill the needs of the user, or if the bot is confused by the request, it will connect the user to a live agent. This transfer again is accompanied with the information that was collected in the beginning of the interaction so the agent is aware of the scenario they are being handed and the handoff can appear more seamless.

Another way for AI bots to understand when an agent is needed is through the analysis of natural language and sentiment. When the AI bot is trained to understand verbal cues, it is able to better pick up if a user is angry or frustrated and needs to speak to an agent.

It is important to note that even though AI bots can solve more challenging issues presented by the user-- due to users being able to write in their needs-- there is more room for misinterpretation since the bot is taking decisions for itself on what the problem actually is and how to fix it instead of following a decision-tree solution through“user-driven menus.”

In each scenario, once the need for a live agent is recognized, the bot can inform the user that an agent is on their way to help. At this point, you can decide if you hand the user off to the first available agent, or if you want to assign the customer to specific agents based on language restrictions, geographical preferences and so on.

If you wish to learn more about which transition process works best for you, contact our experts and begin building the perfect chatbot for your company.
⬅︎ Back to blog