At XTRF, we’re a small company. We’re so small, that when a client asks how big we are, we have to spell it out for them. Instead of sixty, what is typically heard is sixteen, so we have to say six-zero.
Anyway, we’re small. And, like in every small company, if you can’t work harder, you have to work smarter. That’s why automation runs in our blood. It can be seen in our product, but also in our day-to-day work.
Marketing and sales tasks can be very demanding. Not only are they time-consuming, but they have the potential to take even more time than necessary. Lead research and lead qualification are fine examples of that. That’s why we have been looking into automation in this area and decided to have a chatbot on our website.
Unlike a simple live-chat app that connects you to a real person straight away, a chatbot tries to keep the conversation going by itself. It follows scenarios that we have configured based on real conversations with our clients. It asks a question and presents the website visitor with a list of answers. Based on the answer chosen, the conversation continues until the end of a certain pre-defined path.
Each path can be redirected to or intercepted by a real person. The decision on whether or not to step in is based on our human judgment. What’s more, every scenario or conversation path has a score that we have assigned to it during configuration. The scores range from ‘not a lead’ to an ‘ideal target prospect’.
Sometimes, when we see how our chatbot fails, we tweak the scenarios. After all, these are our scenarios, so it’s our failure. There’s no AI behind it, so it’s really as if we were asking those preliminary questions. Apparently, it’s only us seeing it this way.
At certain times of the day, no human being is available and the bot is there all alone. It’s when both Europe and the Americas are asleep, or at least should be, that some of the strangest human-machine conversations take place. Website visitors insist on talking to a real person, but when they find out nobody is available, something changes. They take a short pause as if they were looking around to make sure nobody is watching… and then let their worst impulses get the better of them.
If you have seen the science fiction television series Westworld, you probably know what’s coming. The series is set in Westworld, a fictional, technologically advanced Wild-West-themed amusement park populated by android ‘hosts‘. The visiting humans come to experience a Wild West adventure, but instead, let the darkest side of human nature take over. Just like our website visitors.
I want to say that we’ve seen it all, but that’s not the case, unfortunately. Imagine having your morning coffee, opening your laptop to check the bot conversations from last night only to see a new combination of insults. Some just write, some use CAPS and exclamation marks, others prefer to use emoji. And it doesn’t stop after one line! When the scenario is complete, the bot will no longer answer, so they go on and on.
I’m sure you remember the Microsoft project “Tay” — an AI chatbot that was shut down only 16 hours from its launch. But that was different — Tay was let loose on Twitter, while this is a website of a professional service provider visited only by other professionals. Turns out it’s not that different.
That’s why I think we’re not ready for that kind of automation. Not until we learn how to talk to a non-human being and treat it with respect.