Customer Service: How Amelia AI Improves Efficiency

Michael Krigsman: I’m Michael Krigsman, an industry analyst and the host of CxOTalk. I’m in New York City at the IPsoft Digital Workforce Summit, and I’m speaking with Edwin Van Bommel, who is the Chief Cognitive Officer at IPsoft. Hey, Edwin! How are you?

Edwin Van Bommel: I’m good! Nice to meet you!

Michael Krigsman: What does a Chief Cognitive Officer do?

Edwin Van Bommel: I build really industry-specific solutions on top of the base platform. And I’m responsible for all of the implementations and operations support that we do for our clients.

Michael Krigsman: So, Amelia is all about customer experience. And, you are using data and connecting data to customer experience, and to conversations.

Edwin Van Bommel: Yes. You just make Amelia the front and center of your experience. And you just ask a question, Amelia starts solving your problem, which is in my mind, cutting out a lot of unnecessary irritation. It really becomes a great experience.

Michael Krigsman: How do these relationships work?

Edwin Van Bommel: Well, there are actually three things which are important. There is the data for Amelia, first of all, to understand what you need, and what you want for what your intent is. Then the second data element is all the data that she needs to solve the problem because Amelia solves problems really end-to-end. So, I need data from you, for instance, to open up your bank account. I need your name, I need your address, and these types of elements. And the third element, which is really new for V3 is we also use data and analytics to really enhance the experience to make Amelia even smarter.

Michael Krigsman: The goal of customer experience, in this case, is to make Amelia seem indistinguishable from a human.

Edwin Van Bommel: In general. We have a passion for really beating the Turing Test, and really making it possible that when you talk to Amelia, you actually wouldn’t know whether you are talking with a human or with a machine.

Michael Krigsman: So, that’s your reference point? The Turing Test.

Edwin Van Bommel: Yes. If you have a conversation and I ask you, “Guess what? Was it a human or a machine?”, you would actually not be able to tell. So, probably you would tell the human, because you would not expect a machine to be at that level.

Michael Krigsman: What are some of the key use-cases that your customers are applying Amelia to? What are the problems that they are solving?

Edwin Van Bommel: Let me first discuss the use-cases, and then the problems. So, there are three different types of use-cases. The first one is really interactions with their customers, for both services and sales. The second one, some clients just want to support their people with assisted intelligence. So, for instance, we have a case where there are unlicensed agents, and they want to make sure that unlicensed agents really stay within the rights they have, and what they need to do, so really a compliance case. And those unlicensed agents are talking with Amelia and saying, "Look. This client wants to move and change an address," and Amelia will ask a couple of questions which the agent then checks with the client. And depending on the case, she will say, "Okay, you can do the address change," or "this is a specific service." And, truly, I learn a lot but its' truly different by state. So, Amelia understands by state what is allowed and not allowed.

And then the third one is there is, within companies, a lot of processes which employees need to run, which are, honestly, just a waste of their time, like passwords where Amelia becomes more like an internal service agent and make them more productive. And when I talk with my clients, what I want to resolve is they want to, first of all, improve their customer experience much more, overall. But secondly, they want to free up time for their expensive people so that they can really spend more on value-added task. So, one of the speakers here, one of our banking clients, actually made a point earlier today that he said, "We really have not more time in the call center to deal with really very human and complex problems that our clients have." And their employees are actually super happy that they now have more time to help their clients with that.

Michael Krigsman: So, Amelia frees up the agent from rote tasks, routine tasks, so they can spend their time on actually solving …

Edwin Van Bommel: Absolutely. I have always … When we free up the capacity, then it’s the decision of the client how to use that freed up capacity. Some of them use it exactly for this. Others might use it for cost reduction. That depends on the economics of the company.

Michael Krigsman: What are the kinds of metrics and ROI measurements that your customers are looking at?

Edwin Van Bommel: For all of our implementation, we look at three measures. First of all, coverage. And coverage is easiest to explain as if you have a channel. It's just a market share in the channel of Amelia. Can she handle 30% of the conversations or 60% of the conversations? Now the good news is to handle 60% of the conversations, that might be only 30% of all the content which is required in that channel because there's obviously a couple of questions where the most popular ones are the most obvious ones. But, increasing the coverage is important because you want to make Amelia relevant. Amelia often has to say, “No. I’m not trained for this,” it’s not a great user experience. So, the first one is coverage.

Then the next one is accuracy. If she helps you, she needs to do the correct thing, and if she can’t help you, she needs to still escalate and not give you a wrong answer. So, that’s all inaccuracy. It’s about intent recognition, it’s about then doing the right process completion, and we always measure these things. So, when you look at what our clients show today, some will show coverage 40, 50, 60%, and they will see coverages typically ranging 85% it’s a complex case, very unbounded questions with a lot of variations, up to 95%+ for very structured cases.

And then there’s the third KPI, which is end-customer or end-user satisfaction, because we want to have returning visitors, which, in great accuracy, I still might have created the most horrible experience you can imagine with a lot of “yes/no” type of questions. Very unpleasant, but very safe. But then, customers will not come back. So that’s why we have the third one to make sure that we, in some cases, take a bit more risk, open-ended questions, and that the technology just needs to be as flexible as possible to still understand what the end-user wants.

Michael Krigsman: So, in effect, many of the metrics are their traditional contact center metrics.

Edwin Van Bommel: Absolutely. She’s like a human agent, you know? That’s absolutely true.

Michael Krigsman: What advice do you have for companies that are looking at this type of technology to deploy?

Edwin Van Bommel: I would say try to be bold. If you make it very small, you don’t see the impact. So, I always would start with very relevant cases, because if you start, and you believe you still have to convince colleagues, it can only convince colleagues when you show something which is relevant, when they say, “Oh, wow.”

The second one is I would scale fast within a certain domain, because then, actually, you see the exponential impact of it. You’ll then also learn that customers start really to value it and see it almost as an additional channel and for some of them, might even become really their favorite channel.

And then the third one, what I would do, is always keep the end-customer in mind. How, though, do humans really want to communicate? If it's all about being extremely safe and very compliance-driven, or just copying your rep flow, that's not a great experience. We have, for one of our insurance clients, we actually started interviewing agents. Their own agents. And what we discovered was that they first had a personal conversation before they even started talking about "what's your address," and stuff. And that really enhances very much the user experience, and you have major impact. It's more difficult to implement, but it's not impossible at all. It was a fix of a couple of weeks. But now you see that the dialogue is significantly more pleasant.

Michael Krigsman: Okay! Edwin Von Bommel, thank you so much!

Edwin Van Bommel: Was a pleasure, Michael. Thank you!