I built this customer service chat simulation as a quick demo in Twine. In this scenario, you are a customer service agent trying to help a frustrated customer deal with an issue. You need to gather information from the customer and provide a solution. If you can’t resolve the problem yourself and have to escalate it to your manager, you fail the scenario and have to start over.
This is a rules-based chatbot, where users pick from a few pre-programmed choices. A rules-based chatbot is basically a decision tree, which means it has a lot in common with branching scenarios. This isn’t an AI chatbot where you type responses and the AI responds to you. I built this chat simulation in about 3.5 hours, most of which was just writing the scenario text and choices.
Try the customer service chat to see if you can help an unhappy customer.
Opportunities to recover from mistakes
The branching structure for this chat simulation includes 50 passages. I provided multiple opportunities to recover from mistakes within the chat by reusing choices. For example, in the first decision point, “I’m sure it wasn’t 5 times” is a bad choice. However, you can recover from that error by choosing “Oh wait, I see it now” and then “I’ll help you get it fixed.” In real life, customer service agents can often recover from errors by redirecting and asking the right questions.
Providing those opportunities to recover from mistakes makes this a more realistic practice activity. It’s also a better learning experience; it’s frustrating if every mistake leads to an immediate failure.
