Hey everyone,
We recently changed the way we provide support to our users. We provide ticket support to Free Subscription users and Chat Support to our Paid users.
What we did is that we created two workflows for each type of users but we wanted to keep Fin AI as a first instance of support because we have around 30 % resolution rate.
So we have two similar workflows with the only difference that in one we are converting the chats to tickets automatically. They look like this.
However, now we are facing an issue where a lot of users are abusing the bot by typing “Speak to human”, “Talk to agent” or similar which leads to a scenario where the bot is bypassing the workflows and it’s directly transferring this to a Customer Support agent which is not intended. In the workflows we have different paths where we use tagging and this bypass of the workflows is breaking the data.
Here are some examples that I can show:
- User is typing “Agent please”, the AI is completely ignoring the workflow, asks for more info and transferring to a human being. Instead, this should be converted to a ticket because it’s a free subscription user. Because of bypassing the workflow, we also don’t tag and collect data on the type of the request;
- Here we have the same scenario:
I was advised by the Customer Support team at Intercom that this is intended behavior, AI is sometimes unpredictable(which is a shame in my opinion) and that there is no workaround.
However, because of this we are unable to build a proper Support strategy, conception and configuration if we are unsure how it works and if it's too unpredictable.
I would love to hear your thoughts, if any has any ideas or a solution.