I have a fair amount of experience in a related field, which is messaging-based service and came to the article is spot-on. Some companies simply wanted to map

I have a fair amount of experience in a related field, which is messaging-based service and came to the article is spot-on. Some companies simply wanted to map their phone support flows to messaging (yuck!) and others wanted to do more pre-chat type of activities, such as getting basic information from the customer. For these latter use cases, some level of "AI" was useful to let a customer communicate with more natural language rather than having to enter a specific number or keyword, but you still essentially ended up with an analogue of a phone system. In my case, a goal was better agent efficiency (i.e. fewer agents handle more customers) AND happier customers. A sweet spot for these chatbots was for them to handle specific interactions but not the whole conversation. For example, handling some basic pre-chat information gathering, or in the case of an interaction around rebooking a hotel stay, handling the date and room selection when using a medium that supports rich messaging, such as FB Messenger or iMessage. Until we have fully conversational chatbots, it's going to be really hard to have a solution that a) eliminates agents from the equation AND b) results in higher customer satisfaction.