The key to a successful conversation is managing expectations which lie beyond the happy path. Especially when we cannot help our users, we need to make sure to fail gracefully. In this article we talk about how to avoid customer frustration when the virtual agent is not able to help. This escalation strategy is relevant for any virtual agent.
Not being understood by the other party is not only unpleasant in a human-led conversation, but also in a conversation with the virtual agent. First, we have to face that we cannot prevent this situation from happening, because even in a human-led conversation we sometimes have to repeat questions or don't get the other persons intention, fail to answer properly and with that cause frustration. But we can implement some methods for that and alleviate the frustration by counting how often the user goes through the default branch of the flow and add different questioning and helping strategies.
One possible way to deal with this is by asking whether the user can provide more details. On the second try, the virtual agent gets more specific and says that it did not understand the user and they should rephrase their question. As a last resort, the agent offers different topics as well as a possibility to talk to a human agent.
However, this process can be changed to the needs of your target group or adapted to the already existing escalation paths in your contact center. We'll now take a look at how the example above works in Cognigy.AI.
In the If-Condition we check whether the user reached the amount of the frustration level and if so escalate on Then.
When the escalation takes place on Then we need to remove the frustration context, otherwise it would continue escalating.
In the Else Path we add an Add to Context Node where we trigger a counter for context.frustration.
With this pattern we make sure to not only manage user expectations but also to avoid frustration when intentions weren’t recognized.
You'll find all of the showcased Flows in the attached package.
Feel free to look at other conversation design patterns to improve your virtual agent’s conversation design capabilities.