3 Types of Chatbot Failures

Chatbots have become prevalent. It doesn’t matter if we don’t like chatbots or prefer talking to humans. In the future, we will have to rely heavily on chatbot support. Gartner’s report predicts that 85% of customer interactions will be managed without humans by 2021. As we have been building chatbots for car dealerships, we see the following three types of chatbot failures.

1. False-positive / Missed intent

False-positive means that the bot responds to the customer with confidence, but the answer wasn’t to the question. The rule that applies to humans communication also applies to bots. It’s better to acknowledge that you don’t know than pretending to know the answer. 

For example, a customer asks how often he should do an oil change, but the bot interprets it as the customer wants to book an oil change appointment. The earlier iteration of our bot was like that, but it has since evolved.

2. Validation Hell

Forms on websites usually have validation in place so that you provide the correct email, phone, or even address. A chatbot also wants to validate the answer to a question, such as when asking for an email address, phone numbers, etc. However, what we see is humans don’t follow the rigid flow when engaging in conversations. It’s okay to prompt for validation, but the bot shouldn’t block the user from diverting.

Our experience showed that trying to validate three times is acceptable to most customers. It is also better to use different phrases each time.

Chatbot Validation Hell
Chatbot False-Positive

3. Low Intelligence

When a conversation starts, the bot should clarify its strength and set up the right expectation, which in most cases, the customers will give it more lenience. Nonetheless, it’s a failure if it appears to be dumb.

For example: When the user asks, “do you open Sundays?” The bot doesn’t understand because of the plural on Sunday.

For another instance, the bot doesn’t understand when a user asks if a 2020 Corolla is at the dealership, but it knows to show the 2020 Corolla if the user explicitly says “show me 2020 Corolla”. 

We at DealerAI.com deliver an AI chatbot solution specifically for car dealerships. The above examples were actual chatbot failures of our AI. It handles those scenarios perfectly now. But quite honestly, it still exhibits error, just as humans make mistakes. We strive to eliminate them by training the AI model with conversation data and implementing rule-based logic. In this aspect, the advantage of an AI chatbot versus a human agent is that once the AI learns, the knowledge becomes persistent and permanent. 

If you’re interested to learn more, feel free to book a demo call with us.