HDI’s SPOCcast is your single point of contact podcast for service management and support insights. For Episode 13, I interviewed Chris Chagnon to discuss analytics, artificial intelligence (AI), self-help, and much more. What follows here is excerpted; for the full impact, I encourage you to listen to the entire podcast.
RA: Chris, you recently wrote an article about demystifying the terminology behind analytics. Why is it important for service and support people to get familiar with terms like session duration that we see in web analytics, and how does it help the service desk?
CC: The reason that I started doing that article was because a lot of people are using analytics tooling. Whether that’s different popular suites, they’re all pretty much the same when it comes to the metrics they’re gathering. But what I found is when we give out reports for those analytics, people can get very confused as to what those things mean. So the article was my attempt at helping people figure out why those are important.
So one of the things you mentioned there was session duration. If we’re looking at the number of people on our website, that harkens back to...the ‘90s when we had hit counters on every single website, and we were very, very…gung-ho about letting people know about exactly how many visitors we’d had. But quantity and quality are very different things. So things like session duration kind of help us to figure out exactly where people are spending their time. If we see that someone’s spending three seconds on our home page but five minutes on a support article, maybe that’s a good thing. People are looking through, they’re finding the article that they need. But let’s say we’re looking at those same analytics and we see we have a brand-new support article posted and they’re spending 25 minutes on the article. For a five-step article, we probably don’t want to see that type of thing happening.
The other really handy thing that analytics show us is the paths that people are taking. So, we can start to see exactly how somebody got from point A to point B, C, D, all the way through Z. So, we can see if the intended flow is working, or if people are getting really lost. Does somebody click on an article and then work their way through five more articles before they finally submit a support ticket? Or, what if no one ever submits a support ticket? We can probably assume they’ve done their own self-service there.
RA: Right—that makes perfect sense. In the modern systems, do you see the ability to track people from, let’s say, article to article, if they’re article-hopping to try to get more information on a particular topic, or if they’re puzzled by what they’re reading?
CC: With modern systems, luckily you can track all of those clicks. You can see exactly where somebody started, whether that was a referral through Google, or if they were doing some sort of search in your knowledge base, or—one of the things we see with students, being in higher ed—is they’ll actually share things on social media, which can make you feel pretty good, because it shows that people are using it and say, “Oh – here’s the answer you’re looking for.” But luckily because of that type of thing you can see how they were referred there and you can see exactly when people are hopping. You can see how many people end their sessions on one page, and how many people continue looking for more knowledge.
RA: Speaking of which, self-help is becoming increasingly important. Everybody’s trying to shift left, get as much off the plate of the service desk at Level 1 as possible; and we’ve just done some research that indicates that end users are asking for self-help. And yet some organizations are still having trouble when it comes to adoption. They’re not getting good adoption rates. What in your view, needs to happen to make it work?
CC: One of the biggest shortcomings you see with a lot of self-help portals or websites is that there’s a lack of letting people know that it’s there. We tend to try and think of how we can help our users—which is exactly what we’re supposed to be doing. But an example of where we kind of go wrong is, if somebody submits a support ticket to our service desk, we then help them, but we’ll copy and paste the answer into the email, ‘cause we want to be very helpful. This is very much a “you can lead a horse to water” type thing. But you can actually start to say, “Well, let’s send them a link to the exact article they need.” We can say, “Here’s the article that I’m pretty sure will help you.” Then we get those analytics we were talking about earlier, but then we also start to expose them to the site.
One of the shortcomings we had with the launch of our website was it was timed with the marketing department’s release of their brand-new website. So we were playing second fiddle to that. They did a really great overhaul of the entire university website, but in order to not detract from that, we didn’t really have a launch date. It just kind of went live and we said, “OK, it’s there.” So advertising it and letting people know about it, whether that’s through support tickets, posters, advertising, different things really can help change how people see it.
RA: I like the idea of sending people links, because we do get used to what we click on, right? We get used to going to “Oh, that was helpful. Maybe I’ll even bookmark this, because I can see things that will help me out in the future” instead of copying and pasting everything into a response. Why do you think that is, that more people don’t take that approach and try to encourage their users directly to get into the site so they can see what’s available?
CC: I think a lot of that comes down to conditioning. Over time, a lot of service desks and helpdesks have wanted to be very helpful to people and they want to go out there and they say, “I need to send them the exact answer they want.” That creates knowledge-siloing issues, and you don’t necessarily get feedback from the users about that, and you might end up with users that know something and users that don’t know the answer to that because of it. But by sending out those links, we start to see a lot of benefits there, beyond just the fact that we can track it with analytics. One of the things that’s a really big benefit for us in higher education is accessibility. We start to see we can control the format of the content and we can control how accessible it is. You don’t necessarily get that when you’re sending out an email that’s formatted by that technician at the time; but also, sometimes people really like creating PDFs or other documents. One of the most accessible versions is to create those knowledge articles out there in the web.
RA: Speaking of higher education, you are a doctoral candidate, and can you tell us a bit about what you’re working on and how it’s related to service and support?
CC: I have bene very lucky to be in the position that I am in. I started working at my current university with the intent of just working there, but because of their generous benefits I was able to get my master’s degree there, and then continue on. Once I met my current advisor, I knew that she was the one for me. You know it was “love at first homework assignment” I guess you could say. She was really great at inspiring me to want to learn more. So, through the PhD program that I applied for, I am able to start working on user experience and service experience. And with my background in IT, I’ve been able to apply that as an end-to-end…. How do we keep IT moving forward? What sort of things can we do to say how is new technology going to impact IT.
A huge area of interest for me has been in helping those Tier 1 workers. We did some preliminary studies where we found out that if a student worker worked for our service desk from the moment they got to the campus as a freshman all the way through their graduation four years later, they would have the equivalent of about six months of full-time experience. That makes it really difficult to train those people and really difficult to have consistency across the board. So, helping those people do their job better really works with that shift-left mentality of saying, “Self-help and Tier 1 are the way that we should be focusing, and we won’t need to escalate it.”
So, I’ve been looking at technologies like machine learning and artificial intelligence in order to say, “What are the things that we can do here in order to make their jobs easier?” Specifically, we’ve been taking all of that historic data that we typically don’t use within our organizations and using machine learning practices on it to turn that learning into suggestions for those Tier 1 workers. So a new ticket comes in and we can say, “Based on history, this is where this ticket should go.” We then take it a step further and use some optimization techniques from some mathematics in order to say, “Here’s the exact person that should be working on this ticket,” based on their skill set and the skills required to fulfill it. Preliminary studies have shown really good results on that. We’ve been running an experiment since the fall with the service desk, implementing this program where we run it every single morning, and the ticket time to close has been much shorter, the quality of the responses—because people know what they’re talking about—has been much better, and we see a satisfaction among those people who are doing those tickets to be much higher as well as from the users who are having their requests fulfilled.
RA: That’s all awesome, and that’s the kind of thing that I think we all need to be looking at these days. Obviously, the technology isn’t going to solve all the problems. But it certainly can assist us, and one of the ways it can assist us is in accessing a lot of information rapidly and analyzing it and putting things where they belong, as it were. As you said, pointing it right to the right analyst, or right to the right group or people to work on a specific issue…
Everywhere we turn, we see articles and studies and conference sessions and etc., about artificial intelligence. We know a good deal of it is hype, one way or the other; either we’re all going to lose our jobs, or you won’t have to work anymore. So what are some of the areas where you see AI being a real help and where not?
CC: Well, one of the things that I want to address there is replacing all of our jobs…. Machines are here to help us, and one of the early areas of research that I was looking into is the concept of automation versus augmentation. Automation is where we use that same data I was talking about, and because we achieved a certain confidence level, we can say, “OK, replace all the Tier 1 techs with this program.” Augmentation is where you say, “Well, no, I can’t replace all Tier 1 techs because there’s going to be somewhere where the machine goes wrong, or I can’t get to a confidence level that I’m comfortable with.” Through augmentation we provide that data to the techs. And that’s been really beneficial because we don’t change workflows. We augment them with new things that help people do their job quicker, better, faster, all of those things. But we don’t really need to worry about the accuracy of the program or have somebody who’s checking in on the program every so often, because we have those prompts and checks and balances, and it gets better over time.
Machines are here to help us.
The other thing I think we’re going to start seeing a lot more is this Big Data analysis. AI and machine learning are the new “it” word, but they’re really just an extension of all that talk about Big Data that we had from a few years back. Where I don’t think we’re going to see it is in fully replacing people or replacing their jobs. We’re going to see it start to make things a lot better for people and we’re going to start to see a lot better algorithms that do sound a little more human-like. I’m sure we’ve all had some of those customer service interactions with various companies where it’s just, you know, a call and answer type of bot where you type one thing and it answers with the same message every time. Those are some of the old ways of doing it, and I think we’re going to see a lot more convincing things coming out of that.
I think the other place we’re going to start seeing it is not necessarily in chatbots like that, but in visual information sent out to our teams. We’re going to start seeing our algorithms learning how to help IT and augment their jobs rather than just to our customers directly. So, and example of that is, if we have a server and the hard drive is getting full all the time because it keeps getting these Windows update things that are going on, or something like that, where an update folder just keeps getting more full and more full, we can start to use machine learning to say, “Well, every time someone has intervened for this, they’ve gone through and deleted that update folder, and that solved this issue.” Machine learning can then start to…reach out to people and say, “Hey! Do you want me to do that same thing this time?” They can get more intelligent over time, and they can start making those suggestions. So the people are more approving than thinking and solving every problem from the beginning.
RA: One of my recent conclusions…is that organizations really need to get their knowledge and data organized in such a way that it’s going to make it easier for them to “feed the machines” as we go forward.... What do people need to do to prepare for the coming technologies?
CC: One of the biggest things I’ve seen that helps people shift how they’re thinking about their knowledge and self-service is asking the question, “How is this related?” A lot of times this tooling or the way you write things out is feature by feature by feature. But that does us a disservice because that’s not necessarily how users are browsing our content. So, an example of that is that everything on our self-service portal is 100% interrelated. From the service catalog, you can get to knowledge articles, or what we call “actions” which are part of our request catalog but also hyperlinks; we’ve combined them into one concept. But the idea is that everything that you can do, there are no dead ends. We’ve tried to work with how the users are browsing and how they want to use the site.
Commonly what you’ll see is that someone will launch their knowledge articles, and then they go down the path to knowledge articles and that’s a dead end. You have to go back to the home page, and then you can view the service catalog. From there you can back up and view the software request catalog. From the get-go, we realized everything in IT is related.
So, somebody might want to start in a knowledge article, and then realize that they want to look at a room or a location to have a piece of software installed from that knowledge article, and then from there they want to view the software catalog to see what’s already in that room. Those are all different transactions that are…part of the same process. If we tried to silo them out too much our users might get lost on our site or we’re just going to see tickets getting created increasingly from it.
RA: That’s how we work. Things are all related to each other even if there’s a jump in between. If we’re making the jump from looking at something on a website to typing something in Microsoft Word, there’s some mental process that joins those two things together, and that’s the kind of thing you’re trying to emulate, right?
CC: Absolutely. And then as far as preparing things for AI and machine learning I’m a big fan of separating things in to separate attributes, or rows, or columns, or whatever you want to call them when you’re doing things. A lot of the times, we’ll say, “Well this knowledge article can be one big box.” The way that we started trying to the think about that is saying, “Well, what are the little boxes that make that up?” So, from our news postings we have a lot of little attributes. “What is the benefit of this change in the news post? What is the impact to me?” Little things like that whereas, we could make just one big description box, let anyone put whatever they want. But when you start to get stuff for machines, this process of labeling all of your data really makes it helpful so you can look at trends, you can look at changes, consistencies and inconsistencies, and that’s where the machines start to go really well.
RA: So, categorization and labeling—super important.
What else can you tell us about what you see coming down the road over the next few years that we’re not even looking at yet? Is there anything on your radar that the rest of us in the field should know about?
CC: We’ve started—very far down the road—the shift to cloud. But what we’re starting to see is that shift to cloud has changed our transactions with our users. While we’re going to continue offloading things into cloud environments, I do think we’re going to see a return to having some things locally, and we start to use the cloud for what it’s benefits are. That’s scalability, computing power, things like that. But if you were to try to set up a server on your own on your own organization’s campus and say, “OK, let’s try to run some really deep optimizations and AI, machine learning, different algorithms. That’s probably not the most efficient use of that. That’s why we’re going to see all of that stuff shift. But I do think we’re going to start to see certain simpler things return back home.
The world is greatly interconnected, but we’ve started to be a little spurned by the lack of control that SaaS (software as a service) has provided for us. An outage is no longer something that we can control, and our users’ tolerance for that stuff is going away.
Formerly, when we were assessing whether or not to go to the cloud we just said, “Is it cheaper to run the bare metal server, or to go to the cloud?” That was the big driver behind it. But now I think we’re starting to think a lot more about “How does this modify the user experience or the customer service impact of this?” We’re also starting to see—because we outsource so many things—we have our sysadmins who sit there, and they’ve become business roles now. Rather than sitting there and patching things or running updates or configuring settings, they’re now acting as a liaison to customer service in the cloud, which has required some of those shifting skill sets. But then we have somebody who has, you know, years of database experience, years of server management experience picking up the phone and saying, “Why is this [cloud service] down?” And so I think that is a huge impact of that.
About Chris Chagnon
Chris is an ITSM application and web developer who designs, develops, and maintains award-winning experiences for managing and carrying out the ITSM process. Chris has a Master of Science in Information Technology, and a bachelor’s degree in Visual Communications. In addition, Chris is a PhD Candidate studying Information Systems with a focus on user and service experience. As one of HDI’s Top 25 Thought Leaders, Chris speaks nationally about the future of ITSM, practical applications of artificial intelligence and machine learning, gamification, continual service improvement, and customer service/experience. Follow Chris on Twitter @Chagn0n.
Roy Atkinson is one of the top influencers in the service and support industry. His blogs, presentations, research reports, white papers, keynotes, and webinars have gained him an international reputation. In his role as senior writer/analyst, he acts as HDI's in-house subject matter expert, bringing his years of experience to the community. He holds a master’s certificate in advanced management strategy from Tulane University’s Freeman School of Business, and he is a certified HDI Support Center Manager. Follow him on Twitter @RoyAtkinson.