The Chatbot Revolution - AI, Artificial Intelligence

Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML) are at the cutting edge of technology today. Imagine it: machines that are able to understand what humans are saying, what they mean, what they intend, and what they want. And the AI module is actually able to converse with the human to assist them in a myriad of ways.

They are called Chatbots, but don’t let the name fool you.

The notion of “bots” is not a new concept. “Bot” is a shorthand for robots, and bots have been around a very long time, doing very simple straightforward tasks, from answering simple questions to facilitating various administrative or other functional duties. Bots are even blamed for influencing public elections, but only in the sense that simple scripts can be written that disseminate prescribed information at very large scale through fake user accounts, attempting to make the recipients/readers believe the snippets of content are coming from actual people.

None of that is a true Chatbot.

A Chatbot is a very sophisticated computer application with some of the most advanced computer science built into it, evolving and maturing every day to be able to analyze information, correlate it, learn from it, understand it, and in ever-increasing ways – “thinking” about it, and responding accordingly.

Picture, if you will:

A consumer opens an online chat with a customer service representative.

Kevin (the customer service rep on the chat): Hi there, Tom. This is, Kevin. How can we help you today?

Tom (the customer): I’m having a problem with my coffeemaker.

Kevin: What seems to be the problem with it?

Tom: Well, I’ve had it for a few years, and all the sudden, when I opened it today to put in the coffee pod, the lid got stuck in the open position and it won’t go down now.

Kevin: The lid is stuck open? Do you have a model 2.0?

Tom: Yes it is.

Kevin: I thought so. I can help you with that.

Tom: Do I need to return it for exchange?

Kevin: No, you don’t need to. You can get the lid back down very easily.

Tom: How?

Kevin: I’m going to send you a picture. Here.

(pic)

Tom: I got it. What am I looking at?

Kevin: Do you see the black plastic piece in the bottom of the lid where the person’s finger is touching?

Tom: Yes.

Kevin: Without touching any of the sharp parts underneath there, just push down on the edge of the plastic like you see in the picture. You’ll feel it unlatch, and the lid will easily close.

Tom: It’s right here, let me try that.

Tom: Wow! That worked.

Kevin: That’s all you have to do.

Tom: But why do I have to do that at all?

Kevin: Because in those older model 2.0 units, there are a couple of small plastic tabs inside there that eventually wear down and sometimes break off. They were originally there to keep the lid from opening too far and latching in the open position. But as you just saw it can be easily released with your finger, and you can continue to enjoy your coffeemaker.

Tom: Thanks! That’s what I needed.

Kevin: Anything else we can do for you today?

Tom: No, I’m good.

Kevin: Awesome. Glad to have helped. Have a great day.

This chat was a pretty standard Level-1 Customer Service call. Right? A knowledgeable customer service rep helped a customer fix his issue and help retain customer satisfaction in a scenario that basically arose due to a product design flaw.

However, what is most important to learn from this interaction is that Kevin doesn’t actually exist. He was a Chatbot.

This example is one of a Known-Problem / Known-Solution type of use case. At some point the coffeemaker company realized that a lot of people were calling in about the exact same issue, which fortunately had a pretty straightforward workaround. The problem was, if thousands of people are calling in to report the same problem, then legions of customer service reps are needed to give out the same advice over and over.

Or a single Chatbot can.

In the first sentence from Tom, the Chatbot parsed the words “problem” and “coffeemaker.” The Chatbot was immediately accessing his database of known coffeemaker problems.

Kevin then asked a qualifying question as to the nature of the problem. Tom responded with the “lid got stuck in the open position.” The Chatbot had five different versions of contextual content, one of which closely matched Tom’s description of the problem, that pointed to a known product issue. A resolution routine then kicked in for the Chatbot to share with Tom the known solution. Easy-peasy.

And yet, in all likelihood, Tom believed he was talking to “Kevin,” an actual human being.

Now if Tom had presented an unknown problem to Kevin, or a known problem without an exact set of resolution steps, Kevin could have told Tom that his problem needed to be transferred to their Level-2 support, who would be with him shortly – and then the chat would have been forwarded to a real human technician to take over from there.

But for all the known-problem and known-solution scenarios, an actual human can become rarely needed to be involved. And that can be true for many routine tasks and interactions between a customer and their customer support service center.

Now consider the mathematical implications of Kevin to a business. There can be multiple instances of Kevin at once, hundreds, even thousands. Kevin doesn’t need to be paid. Kevin never gets sick, nor does he require any health benefits or represent any employer tax burden. And best of all, Kevin can work 24-hours a day, 365 days a year. And a lot of Kevins can also speak multiple languages for international customer support.

So with all this in mind, what if a business did a productivity analysis and discovered that 72% of all their customer service trouble tickets dealt with one of several known problem / known solution type events? Might Kevin be able to take over the burden of 70% of the trouble calls in a call center? Perhaps. And if that call center was staffed by 30 people, what if 20 of them suddenly were no longer required?

For example sake, let’s say the average customer support representative earns $36,000 a year. With a 20% benefit weighting, that roughly an employee cost of $43K per rep. So, 30 of those reps represents just under $1.3 million a year in direct labor cost – and that doesn’t count the additional costs of each one’s desk, computer, utility bills, etc.

If Kevin could eliminate 70% or more of all those costs, something maybe north of $900K a year – what then is the cost of acquiring Kevin actually worth to that business?

Is Kevin worth a $100K investment? What if he only costs $50K to build, configure and implement, with an even lower annual support cost for software maintenance and updates? That would represent less than a ONE MONTH Return on Investment (ROI).

The math also dictates why there is no time to lose before each company’s direct competition implements their own Chatbots and suddenly has a dramatic cost advantage that propels them to a leadership position in their market. All businesses seek a strategic advantage that gives them the edge. Automation of all kinds has long been the catalyst of such disruptive changes. The advent of Chatbots is no different.

Now you know why there is a Chatbot Revolution going on right now, and how this technology is getting “smarter” every day. Be a part of it, lest you discover it’s too late to catch up.