Trends, Tools, and Human-Centered AI at KMWorld with Stin Mattu
Following this year’s KMWorld, we took a dive into what everybody on the show floor was talking about and more with our own Stin Mattu. In this podcast episode, he breaks down the big topic on everyone’s mind at the event—artificial intelligence (AI). Now while AI can do so many things, Stin assures us there will likely always be a human in the loop and highlights how that comes to light with well-organized knowledge and a great search tool.
Transcript
Pete Wright:
Hello everybody, and welcome to Connected Knowledge from Upland Software on TruStory FM. I’m Pete Wright, and today we’ve got a recap of the 27th annual KMWorld event, from The Economist, veteran knowledge management trade publication. Our own Stin Mattu was there, and now he’s here to tell us all about the event, the trends, and everything we need to know about the metasearch.
Stin Mattu, welcome to the show. Welcome back. It’s good to hear from you again.
Stin Mattu:
Thanks, Pete. Thanks for having me back. It’s always a pleasure speaking to you.
Pete Wright:
Let’s kick it off with a setup on the KMWorld event this year, 27th annual KMWorld event. What was the overall theme this year? What did you get out of it? And what are they trying to set us up for for the years to come?
Stin Mattu:
Well, I think that the hot topic of the year, from the beginning of this year, has been those two little letters, AI. At KMWorld live this year, it was again the hot topic. I don’t think that there was any vendor or any organization that was there that didn’t either have questions about AI or were basically there to sell their wares on what they’ve got coming up in their roadmaps with AI as well. So, I think AI, it was no surprise going into KMWorld this year that that was going to be the hot topic, and it very much was so across the whole of the conference itself.
Pete Wright:
Well, I think that’s a great setup, mostly because… I know we have some other things we want to talk about, but I’m so curious, from your perspective, how reliable are the messages that we’re getting from across the vendor landscape? How confident should we be in what we’re seeing as this initial product run is hitting the market in making use of AI, particularly in the field of knowledge management?
Stin Mattu:
Well, this is the thing, and it’s something that everyone is talking about. And everyone’s talking about it in a different way that relates to their products. So, depending on what the use case is that are being delivered by a certain product, obviously the messaging kind of changes.
From our perspective here at Upland, we’re taking a very assured approach with AI. One of the things that I said in my keynote was that unless we can realize the benefits to the user, any good feature is just a line of code on a server somewhere. So, we’ve really got to understand and dig deep into what the users actually need when we’re developing new technologies. From the work that we’re doing, what we started this year, very strong with an AI Knowledge Assistant, which is there to speed up the process of creating knowledge and improving knowledge for knowledge workers and authors. So, there is a benefit there to them.
From the agent’s perspective, in our last release we’ve just come out with two key enhancements to the project. One is looking at retrieval augmented generation, so RAG. That was a very common subject across the whole of KMWorld. And that is to generate answers from a knowledge base where you’ve got the sources. So, the knowledge is there, but instead of giving you the articles, it’s actually curating a response based on the information that is available to that particular user through the knowledge sources that they have, giving them citations. So, if they do want to go down and have a deep dive into that information to go on to find something that is deep within that article, if originally the answer doesn’t serve the purpose, then they’ve got the option to do that. And if they want to have follow-on questions, giving them the options of follow-on questions that would be related to that. That is something that, again, will have a massive benefit, especially for call center agents. Because the cognitive load for a contact center agent is to sit there and know what that customer needs in that moment. And obviously the way that we’re finding knowledge at the moment, with articles, it still relies on them to go and read the articles. That’s why it’s so important, with the work that we do with customers, is to make sure that the way that content is created is readable and is in a format which they would be easily scannable for them, and they’d be able to get the information.
And then the other side of what we’ve just done in this last release for RightAnswers, is we’ve brought about a hybrid search. So, taking the traditional keyword search that you would use and coupling that with some natural language understanding and semantic search. So, the content that you’re going to get and the results that you’re going to get are going to be more tailored to the intent of why you need that information.
So, when we’re looking at these kinds of technologies and we’re looking to bring them into the products that we’ve got, we totally understand what the benefits should be for the human being. Again, that is something that is massive in the debate of AI at the moment. A lot of people think that this is there to replace human beings. I personally don’t think so. I have an 80/20 rule. 80% of the hard work that needs to be done, yeah, let’s get AI to do that. Let’s get AI to do the heavy lifting. But then leave the 20% of the work that’s left to do, so for the agent or the author, they can use 100% of their focus on the remaining 20% of the task.
What do we get there? We get a workforce which is more supported. The likelihood of churn of agents, especially in the contact center, would be less because they’ve got tools. They’ve got a genie that’s there that they can just bring out of the bottle whenever they need to and get the answers that they need. And again, it’s up to their intelligence. They’re still working their own brains and their own thought processes. They know their customers better than any machine ever will. The sentiment that’s coming across in a call, it could change in a second. So that’s where it’s really, really important when we’re talking about AI, and we’re moving forward with AI, is to always keep the human in the loop. I’m really glad that the majority of the people that I spoke to personally at KMWorld, and I’ve heard from at KMWorld, we all strongly believe that. It is a case of, this technology is there, and if we use it correctly, it’s going to be something that is enhancing the work that humans do, rather than replacing the work that humans do.
Pete Wright:
Well, I really appreciate what you have said here around the way implementation is going to continue to build trust back in the human agents. This human-centered AI concept is really important if we’re truly going to be able to adapt and change our jobs and let AI, trust AI to do the 80%. You have talked about using AI to address the pain points in knowledge management with the agent. I wonder how you think about that in terms of the cycle of knowledge when it comes to what AI can do to augment and support agents in a call action, and how humans then get information back into the AI (into the trusted system) to leverage it for future knowledge. How are you thinking about both the inflow and outflow of information?
Stin Mattu:
The human touch is definitely needed. Again, where I spoke about the AI knowledge assistant that we’ve got. It’s not there to go and create articles out of the blue for you. Because one thing that we all know, and it’s well-documented, is the biases that you get with large language models. They’re taking content, which is available to them, and how they’re being trained. So the nuances that are there sometimes aren’t correct. It all depends on the knowledge that goes in. That’s why it’s very, very important, even with the content creation process, is that anything that is being created and assisted by a machine should definitely be checked over by a human to make sure that it is fully accurate before that is then being able to be served to agents, and hopefully, at some point, to customers as well.
So, it’s making sure that you’ve got that governance in place when you’re creating the content that you need, and not just relying on a machine to think, well, this is going to be quick for us, let’s just get it done. The machine can do it in… It can probably create you an article in less than 20 seconds. But in that 20 seconds, if the content that is there, even if there is one mistake in 100 things that it’s done right, it’s that one mistake that is going to be the brand damage to an organization. Because as soon as that information is given either by an agent or by a chatbot, or any other kind of form of AI that is basically speaking to customers and delivering knowledge for customers… The minute that that does that, it’s the one instance that everyone knows about. It’s common knowledge, isn’t it, Pete?
Pete Wright:
Yeah.
Stin Mattu:
That, okay, if things go well, people don’t talk about it. But the moment something goes wrong, then that’s where the swarm will come on top of it and then just really, really go to town on it. So, it is 100% important from the content creation period that you’ve got to get that right. And even the sources that you’re going to use. Obviously with large language models, we can point to different repositories and different sources to go and find the information. It’s very important that each one of those sources that you do want to connect are maintained, that the knowledge that is in those sources is always checked and made sure that it’s correct. Because obviously unmaintained knowledge will be the root cause of the wrong information going out. So, there is a lot to think about when we talk about AI and how we apply that in knowledge management, especially where it comes to creating content.
Pete Wright:
I think there are some easy assumptions to be made around what corpus you’re training your AI models on, and that it should seem so natural that if we’re training our bot on our internal knowledge base, then surely, it’s completely safe, right? Surely, it’s going to be free of those sorts of biases that you mentioned, right? I’m curious if you see anything specifically wrong with those easy assumptions.
Stin Mattu:
That’s the ideal world, isn’t it? That’s the utopia. That’s where we’re trying to get to. Right? But one thing that we know, and from experience as well, in the organizations that I’ve worked for, the best intention is to make sure that you’ve got the best strategy for content and the maintenance of content. So, making sure that when content is created, it is up to a standard, whether that’s the content itself that goes into it, and also the way that the content is written. Okay, so that is really, really important. And these are knowledge management basics. This is the 101. And that’s that you can’t progress with technology and get it to do great things if the foundation isn’t solid. So, it’s really important to make sure that we’ve got those solid foundations.
This is where something like KCS can really come to the fore in the way that content is created and the way that large language models can be trained, on content which, like you said, it shouldn’t have any kind of biases. It should have legitimate information in there that is always checked and improved, so that methodology is there to do that. I would say it’s imperative to do that, to make sure that your knowledge is always used, it’s always reviewed, and it’s always kept up to date. Now, if you’re in that situation and you’ve got this solid strategy for your knowledge content, the optimization of the content that you’ve got, then the risks are lower. And I say that they’re lower. I wouldn’t say, at this moment in time, 100%, hand on heart, that we’ve got to a stage where AI can be 100% trusted with the information that’s there. We still need to take cautious steps, assured steps, towards utilizing AI. I’m not saying that we shouldn’t use it, because if we look at all of the technologies that have come out in the last 20 years, AI has been there. Maybe we haven’t understood that it’s there. Just think about it. You take your mobile phone out, you’re going to start sending a message to someone, predictive text. What is that? That is machine learning, which is a form of AI. So, we’re already there. We’re already up to that point. A couple of weeks ago in London, in an event here, a Unicom event. There’s a gentleman called Ron Young, who’s a knowledge management professional, and also a leading authority on AI and knowledge management. A question that was put to him in that conference, as part of a panel that we were on, it was, “What’s the tipping point? When is the tipping point for AI in knowledge management going to come?” And he said straight, “It’s already here.”
Pete Wright:
Yeah. We’re done.
Stin Mattu:
“We’re in the midst of it.” If you’re not using AI to a certain degree at this moment in time, whether that’s as a customer or as a vendor, then your time is numbered. This is that moment. We’re in that moment, and it’s moving forward. But at the same time, Ron also agreed with the 80/20 split, that the human in the loop, it’s the key. It’s the key. And that’s where, obviously, as we move forward with AI, sometimes we have these fantasies of, right, okay, what the technology can do, and we really want to buy into that, and we really want to say that, “Okay, this is going to be totally life-changing,” which it’ll be, in time. But at this moment in time, we’ve still got to basically take those small steps, assured steps, and make sure that it’s fit for purpose and it does what we need it to do for the end user.
Pete Wright:
Let’s circle back to your keynote. Maybe in the lens of this overarching landscape conversation that we’re having, tell us a little bit about what you were putting across on stage at KMWorld.
Stin Mattu:
Purposely, I went in and I wanted to try to stay away from talking too much about AI, because I think that, okay, you know what? After a day or two of hearing AI in every single conversation, we kind of get fatigued.
Pete Wright:
Can only give so much.
Stin Mattu:
Exactly.
Pete Wright:
All right.
Stin Mattu:
So my presentation, I wanted to concentrate on what the importance of search is for customer service organizations and contact centers, and really put it into context. Through that, I wanted to show it through the lens of how frustrating it is for customers. There was a couple of stats that I brought up onto the screen. 33% of customers in a survey with HubSpot basically turned around and said that being put on hold is the most frustrating thing that they feel. There was another stat, let’s take it from the user’s perspective—60% of agents can’t find the information that they need in order to serve their customers, and that’s their frustration.
So, when we’ve got a frustration from the customer, who’s saying that, “I’m on a call and I really just want to get this done,” or, “I’m on a chatbot,” or whatever channel that you’re coming through, “I just want to get my issue dealt with,” and they’re stuck on hold. There’s a blinking light that’s on a chatbot, or there’s some hold music that they have to listen to. It’s a frustration. And those frustrations are what kills the brand value for that customer. So, we’ve got to think about that.
Now, one thing that I did in my presentation, is I brought up a stat from Forrester. It’s a well-known stat within the knowledge management circles, which is Mean Time to Know. Through the work that Forrester have done, they said that 68% of the time on a call is what agents who haven’t got knowledge management would be going away to find the information that they need. So, quite simply, I wanted to make sure that we really get this message to land. I set up a timer and a clock, which went through 100 seconds. In the first nine seconds, I just basically set up that, “Yeah, this is Forrester’s Mean Time to Know, and this is the time that you would be telling the agent about what you want to get out of the call. What is your issue? What do you want resolving?” As soon as that nine second time lapsed and we’ve gone into the start of the 68%, which would be the 68 seconds of that call, I’ve stopped talking. In a similar way that I just stopped talking there for a couple of seconds. As soon as you start to do that in this day and age, and just stop, that’s where people will start to feel anxious in certain situations.
So if for 15 seconds I’m standing on stage, I’m just looking at the audience and I am not saying a word… And you could see the looks on people’s faces. They didn’t get what was going on. So, after 15 seconds when I’ve come back and I’ve just put it into context, that, “That was 15 seconds, and you’re probably thinking, what’s going on?” Now, for a customer, this isn’t a 100-second call. For a customer, they’re going to be on a call which is likely going to take them anywhere between three and a half to seven minutes. And if 68% of that time you’ve got silence, if you’ve just felt awkward for the last 15 seconds while I haven’t spoken, just imagine how your customers are thinking on the phone. How do they feel?”
So, just put little things like that into context. Because sometimes we forget about the human side of the work that we do and especially with contact center agents. In operations, people just think about, right, we need to hit AHT. Our CSAT scores need to be this. Our NPS needs to be this. Our abandonment rate needs to be as low as possible. How many people have we got uncalled? What’s the quality? So, we look at all of that and we talk about that, but we hardly ever put ourselves into the position of the customer, and we’ve all been customers, Pete.
The frustrations that we have, we seem to forget that when we start to talk in a professional context about the importance of knowledge management, and in that scenario, search. Because if you’ve got something which has got the ability to bring you the answers that you need in a very quick and easy way to consume that knowledge, then obviously you’re going to have a better experience with your customer. Your customer’s going to ultimately then be a happier customer. And again, the metrics that we use in call centers, with NPS and CSAT, they are higher. And there’s more likelihood that that customer will stay with you and recommend you because they’re getting a good quality service. And that’s the most basic thing that I think that we all need to understand, especially in the world of contact centers, it’s that delivery of world-class customer service. The tools in order to do that really playing an important part into the whole customer experience.
Pete Wright:
Well, if I may reflect, Stin, because I think that’s an incredibly powerful message. Especially through the lens of generative answers, third party message boards and forums, all of these self-service opportunities that I may have as a customer—by the time I speak to an agent, another human being, shouldn’t that be targeted as the highest value communication that the company, that the brand has with me? Shouldn’t that be what we not forget, that I may have already exhausted not just the seven or eight minutes, but an hour trying to find the answer out for myself?
Stin Mattu:
Totally.
Pete Wright:
It seems like addressing that risk, the gap of time between customer and agent, seems like an important one, to keep allegiance between customers and the brand.
Stin Mattu:
Totally. Totally. Because your agents know your customers better than anyone. And it’s from a call to call basis. I’ve been lucky over the last 20, 25 years of working in contact centers at all different levels. I’ve been lucky enough to don a headset, and I know that when a call comes through, it’s going to be different from each call to call. Every single call is different. It could be the same subject that someone needs help with, but it’s the main character that you’re working with on that call, is your customer. And they’ve got different needs. They’ve got a different requirement, even if it is the same issue that they’re having. There’s so many different nuances there. And to connect with a customer and deliver a service in that way is one of the most satisfying things that any contact center agent can do. You’ll hear it yourself, Pete. When you’ve contacted a contact center, and you maybe had an issue and you’ve been frustrated, but you’ve got that one hero on the end of the phone that has been able to sort everything out for you, the gratitude in your voice is what they resonate with. That’s energy. That’s the part of your day in a contact center when you’ve had the six calls with the customer who has been screaming at you because someone hasn’t done something right, or the product hasn’t worked in the way that they need to. You’ve had to live through those calls, and the one where you get the opportunity to really go to town and resolve that issue for that customer to the best of your ability with the knowledge that you’ve been given, and that is at your fingertips. And as soon as they basically say to you, “Thank you. I didn’t think that anyone could sort this out. You’ve been great. Please let your manager know that I’m very happy with the service that you’ve given today.” That one call will carry you through the rest of the week. And trust me, I know. I’ve been on many calls where that one has actually really just set it off.
In the work that we do, the lens that I look at everything through is, “Okay, how can I go and support that person on the phone or on that live chat that needs to basically excel in their job to make sure that they’ve just delivered world-class customer service?”
Pete Wright:
I spend much of my day, Stin, searching for great metaphor, and I find it incredibly motivating when I find a way to realign my brain in a new way around a particular issue. You just gave me one. Particularly in call centers, in the brand relationship, you, the brand, are not the protagonist in the story of this customer service engagement. The customer is the protagonist-
Stin Mattu:
That’s right.
Pete Wright:
… and that is a really powerful model that can change the way you talk to people.
Stin Mattu:
Yep. 100%. That is all it is, Pete. Remember, we’re human beings. I know my kids always talk about NPCs, non-playable characters within video games, and how we’re all the main characters of our own story. Remember, the person that you’re speaking to, there’s 110 things going on in their life at that moment in time, and we’ve got to make sure that, with our character that we’re working with in our script, which is the call center agent, they can serve the main character in that story with what they need to make that epic for them.
Pete Wright:
You have already teased what you’re working on with RightAnswers. Do you want to give us a plug? Where should we go learn more about the work that you’re doing right now in contact center productivity solutions?
Stin Mattu:
Well, head over to the website, uplandsoftware.com. Obviously, we’ve got three different knowledge management products, BA Insight, Panviva, and RightAnswers. We’ve also got InGenius, which is the CTI solution, and now Rant & Rave falls under us as well, which is a customer sentiment analysis tool. So, where it comes to contact centers, we’re definitely experts, and in knowledge management as well. One of the lines that all of my colleagues at KMWorld was hearing from me when customers were stopping by our booth, is, “We’ve got well over three centuries worth of collective experience in knowledge management. There isn’t a use case that we haven’t heard to date, and we’ve got a product that can basically help you overcome those use cases.”
So anyone who’s interested, head over to the Upland Software website. If the products I’ve spoken to you are your bag, brilliant, take a look at those. But with over 32 different products with all of the productivity goodness that we’ve got, I’m sure that you could be there browsing for days as well. But yeah, I would definitely recommend anyone to take a look and get in touch with us, because it’s the expertise, Pete, that really stands apart sometimes. Like I said, we know knowledge management, we know contact centers, and we know use cases. We can definitely help to overcome any challenges that people have.
Pete Wright:
Outstanding. Stin, as always, I love having you on the show. I love learning from you. I so appreciate your time today.
Stin Mattu:
Always a pleasure. I’ll just leave you with my line saying of, “Teamwork makes the dream work.” So as long as we can all work together to overcome problems, I think we’re all in the good place.
Pete Wright:
Swipe up in the show notes, everybody, you’ll find the links that Stin has been talking about. We’ve got a couple of other links in there too. We’d love to have you take a look at those. We sure appreciate you downloading and listening to this show. Thank you for your time and your attention. We’d love to hear what you think. Swipe up, again, in those notes, and look for the feedback link. We can get a question to any of our past guests, and we will do our best to get them answered. On behalf of Stin Mattu, I’m Pete Wright, and we’ll see you right back here next time, on Connected Knowledge.