AI at the service desk: What actually works (Ep. 53)

AI at the service desk: What actually works (Ep. 53)

[00:07 Weston Morris]

Well, welcome to the next edition of the Digital Workplace Deep Dive podcast. I'm your host, Weston Morris.

[00:15 Weston Morris]

If you listen to episode 52, you saw that we've been continuing a theme of AI and generative AI. In fact, my guest Alan Shen and I talked at length about what's been missing in gen AI, especially to get a return on investment.

[00:30 Weston Morris]

One of the things that we realized is that what's missing in gen AI is AI, traditional AI – making use of traditional AI techniques to curate knowledge, to capture the knowledge and integrate it with these large language models to make them effective.

[00:48 Weston Morris]

In this episode, I'm going to continue that discussion. I'd like to actually dive into how AI, both generative AI and traditional AI, are being used effectively in the real world at the service desk.

[00:59 Weston Morris]

We've all been living with this failed promise of IT where we've been hearing that technology is just going to make the service desk go away. It's all going to be automated. Our problems will be solved automatically, and we're not there yet, right? That magic, that
instant resolution just isn't real yet.

[01:18 Weston Morris]

I have gathered a couple of experts here to help us figure out exactly where AI and generative AI are making sense and people are using it. And maybe even collect some lessons learned and take a little glimpse into the future. My guests have been featured
speakers at several industry events this year, ranging from the Xperience 2024 conference, the Forrester CX Summit and most recently, Enterprise Connect.

[01:44 Weston Morris]

My guests are Brett Weigl, senior vice president of product for AI at Genesys, and Aron Meyer, solution manager for Digital Workplace Solutions here at Unisys. Brett, Aron, thanks for taking some time away from your speaking engagements to talk with me today.

[01:49 Brett Weigl]

Thanks for having me.

[01:51 Aron Meyer]

Yeah, glad to be here, Weston.

[01:53 Weston Morris]

So, Brett, I think I might start with you. Let's just get right into it. What is the most obvious area that service desk and service desk technology can be making use of AI today?

[02:15 Brett Weigl]

Well, I think the obvious place to go is containment. The idea that you can provide selfservice and with newer forms of AI, fully contain those requests, which means that your employees on the service desk can spend time on higher value activities.

[02:29 Brett Weigl]

We've got a customer at Genesys who has spent a lot of time sending every customer interaction through AI first. Some still go to humans, and some are fully contained, but with that, they've increased the level of self-service by 80% since they started several years ago.

[02:47 Brett Weigl]

Really looking at top service request types and really to look at what's the effort being expended to solve those issues. And it turns out a lot of them can really be solved through self-service.

[03:06 Weston Morris]

Well, that makes a lot of sense. AI still isn't at the level of a human. Not quite yet. It makes sense to do what you just described, feed the easier problems through AI, and then leave the harder ones for your live agents.

[03:19 Weston Morris]

But that makes me wonder. If our thinking about how SLAs and KPIs are typically set up for a service desk, it's making me wonder if I make that change where my live agents are pretty much only getting the hard stuff. It's making me think that possibly, we need to make some changes to how we organize our service desk, operate it and measure it. I mean, Aron, Brett, what are you thinking there?

[03:48 Aron Meyer]

I think first off, it starts with greater attention to agent skills, right? The better the bots get, the harder the interactions are that make it through to the humans. And that's also where the greater emphasis comes in on human empathy as well.

[04:00 Aron Meyer]

Consumers expect the bots to be transactional, but their expectations of the humans continue to increase as well. It's not just good enough to get things done, but they're paying closer attention to the experience they're getting while we're providing these services for them. So a big consequence of that is increase in handle times, just as you were alluding to there. Organizations likely need to start recalibrating their workforce management to ensure they're not squeezing the balloon at both ends, that they're accommodating for that extra agent time to handle those tougher problems, to be able to take time to engage with their customers, show the empathy and ultimately too, they may consider how that affects their KPIs like speed of answer and things like that.

[05:01 Brett Weigl]

Yeah, I couldn't agree more, Aron. I think that the idea that calls may be more complex, chats may be more complex, they take a longer time, so it's never been more important to make sure that you're managing the quality end of things.

[05:19 Brett Weigl]

What experience do those customers who reach out actually have with those advisers? And the idea of applying AI to that is now possible as well, right? Let's use the AI not only on the front end, but also to evaluate how things are going.

[05:42 Brett Weigl]

And then it's really important to also have a lens into what journeys our customers are taking who reach out if they've had a part of that interaction be self-service and then part with an agent. How efficient was that? How much effort did they have to expend? And were the paths convenient and optimized to the problem type they actually have?

[05:53 Weston Morris]

I don't think I've ever thought about that because, you know, a lot of people are using journey analytics to measure how people work their way through a system. It's being used in a lot of areas, but especially in service desk.

[06:06 Weston Morris]

Historically, I think we're looking at it solely from the interaction with the live agents. But what I hear you saying is we also need to start thinking about measuring the experience that people have with virtual agents as well.

[06:16 Brett Weigl]

Yeah. It's really convenient, by the way, if a virtual agent can output the same kind of data that helps you to measure the quality of that as well, right? So, it's not only the sequential path, but that's also a conversation. And was it a good conversation? The more that gen AI lets self-service become more human, we're going to really have to look at the quality of
those conversations too.

[06:44 Weston Morris]

And Aron, you talked about workforce management and training. As we give harder problems to the service desk, do you see a time where really that level one agent, what we call level one today, disappears, or do they have a different role? How are things changing with your different skill levels within the service desk?

[07:06 Aron Meyer]

Yes. Ultimately, I think we're seeing more of what may have been traditionally level one agents really becoming level one point five. And I think it's just going to continue to elevate what we may have considered to be level two before. It's going to kind of become that front door of human agent interactions.

[07:26 Weston Morris]

So far, Brett, we've been talking about how AI is used in the most obvious sense, right? It's that touch point that I, as a caller or a person looking for help and reaching out to the service desk, and my first contact is with this virtual agent, a generative AI-type chatbot.

[07:45 Weston Morris]

But I've got to believe AI and generative AI can play a role behind the scenes, too. So, how is that being used today, or will it be used?

[07:55 Brett Weigl]

Yeah, I mean, I think that the amount of assistive technology is really growing. Agent Copilot kind of technologies can really help to whisper in the agents’ ears what they should be saying, what content they can use.

[08:15 Brett Weigl]

I think a lot of your prior discussion on knowledge is really relevant here because a lot of content-based service experiences really require a way to generate answers out of whole manuals, long knowledge articles, things like that.

[08:29 Brett Weigl]

So, I think what we really want to do is prompt with suggested solutions, and we want to make things more efficient, sort of at that point of need when the agent might actually have to hunt for answers.

[08:40 Aron Meyer]

Yeah, I think just tagging on that, using things like Agent Assist and Agent Copilot, that's doing that automatic surfacing of knowledge and helping to highlight those relevant sections and, in some cases, generalizing a summarized answer for that agent.

[09:02 Aron Meyer]

It allows them to spend more time engaged with the end user, engaged with the customer and less time searching the knowledge base and looking up information. You know, it's certainly that thing that we're seeing so far with these technologies that helps to offset a bit of that higher level of difficulty the agents are handling.

[09:31 Aron Meyer]

There's some reduction in handle time and especially a reduction in hold time, putting the customers on hold to go do that research and then come back to them with an answer. Now they're getting that right up front. So, in a sense, it's kind of making them bionic or
giving the agents a bit of superpowers.

[09:52 Weston Morris]

And I know you've been digging into this a little bit in the real world. What kind of reductions in time are you seeing as you employ the virtual agent whispering in the ear of live agents?

[10:06 Aron Meyer]

Yeah. As the technology continues to evolve, the improvements continue to get better. But we're seeing things right now in the order of like an 8% to 10% improvement in overall handle time.

[10:19 Brett Weigl]

I was just going to jump in, Weston and Aron. Sorry about that. I think another thought is that a lot of service experiences arrive at something transactional that then needs to happen. Up front, I'm trying to figure out what the problem is.

[10:34 Brett Weigl]

Gen AI can really help with that knowledge article. All these things are helpful. Ultimately, I've got to go do something. A lot of times what you can do is prove out those processes with agents, getting them more efficient just with scripting.

[10:47 Brett Weigl]
Some of that can also apply to self-service as well. You prove it out with humans in the loop, but eventually, why can't a customer do it directly? There's a real opportunity there to look at that overall process improvement lens as we apply these technologies.

[11:03 Weston Morris]

I'd like to dig into that just a little bit more and understand how that works. Is that the type of thing we have with the service desk? There are the agents themselves, and then there's the person overseeing a group of them or coaching them. That level in the organization, that person that's overseeing multiple agents, is AI helping them as well?

[11:24 Brett Weigl]

Oh yeah, absolutely. I mean, I think that the problem of being a team lead or a supervisor is difficult at best, especially if there are live calls, live chats, you're trying to monitor everything. You really want alerts, right?

[11:47 Brett Weigl]

But you also need digests, summaries and interpretations of what's actually happening.The traditional method is you cherry-pick, and you have to read or listen to all of the verbatim. Then, you make your own judgments via a tagging system and then invoking
workflow.

[11:59 Brett Weigl]

But AI can take over some of that job as well and provide more of the ability to scale for the supervisor.

[12:06 Weston Morris]

Cool. So far, we've actually identified three different areas where AI and generative AI can play a role in the service desk. Just to recap what I'm hearing: Number one, obviously talking to the service desk, the virtual agent; Number two, the agents themselves, having this Agent Copilot whispering in their ear, saying, "Hey, talk nicer. Here's a solution," that kind of thing; but the third one you just brought up is taking another step back, and that's the team lead overseeing multiple agents and providing some guidance as to how to help them improve. I really love that.

[12:42 Weston Morris]

Let's see if we get one more. This was, I think, one I'd like to tie back to the podcast that I did with Alan Shen, the most recent episode where we were looking at knowledge management and curating knowledge management, something that he has really put a lot of thought into.

[12:54 Weston Morris]

But let's move from theory to the real world. Aron, what are your thoughts on how we can use AI to help us curate the knowledge that then the generative AI or even live agents are making use of?

[13:09 Aron Meyer]

Yeah. I think this is one of the super cool things that our AI team is doing because we have those interactions that are happening up front with our gen AI bot. And then we have that wealth of expert interactions that our agents are going through as well, both in their
transcripts as well as what's documented in tickets.

[13:23 Aron Meyer]

We're taking the AI and pointing it at that information as well. And so it's able to start surfacing for us where the gaps are in that knowledge that we fed it on the front end that leaked through to agents, and it's able to then make recommendations to us based on
what it found in that expert conversation that happened with the live agent in Genesys through to the tickets that we're documenting.

[13:45 Aron Meyer]

And we just have to add the knowledge manager human in the loop process to review that suggestion and then approve it to be published, and it now becomes available immediately both for the bot, for those next interactions as well as for the agent.

[14:18 Brett Weigl]

Yeah, I would say from a Genesys perspective, we really focus on the idea that knowledge is an asset that gets applied across channels. It gets applied across experiences in a flexible way.

[14:40 Brett Weigl]

The idea of leveraging knowledge for self-service and for agent-assisted support, and even thinking about infusing it into web or mobile experiences up front so search can take place at the front end --- those types of things are all great use cases. And I would just say also that I think a lot of organizations have modernization that’s needed about where all of their knowledge lives in the first place.

[15:09 Brett Weigl]

To get more of a simplified view of that and to be conscious about those curation and approval pipelines, which I know Unisys is really focused on and has a great approach to.

[15:10 Weston Morris]

Yeah, you hit on something there that I really do agree with. Traditionally, a lot of the knowledge is in knowledge articles, and it's in an ITSM platform and people view that as the place and it's the only place, but things are changing so quickly.

[15:25 Weston Morris]

We have IT, the digital workplace, the tools we're using change, the apps we're using change, where we work, how we work, and how we get help needs to change and be reactive like that. And that knowledge, a lot of times, is in someone's head.

[15:44 Weston Morris]

So how do we capture the knowledge that is not just on paper or digital paper written down somewhere, structured knowledge, but capturing maybe even conversations in a chat, capturing knowledge from an expert and accurately feeding that in.

[15:52 Weston Morris]

This is an area where I like what you described, both of you here. AI can play a role in finding it, but then there is a curation process step where we definitely want to have the human looking over the shoulder of the bot here and making sure that it's done properly.
Yeah, great points.

[16:13 Weston Morris]

Well, gentlemen, this is an exciting space. There are just so many cool things that are helping to reduce the digital friction that employees run into on a daily basis.

[16:24 Weston Morris]

But I know one thing is true, and that is you just don't throw a switch, turn all this on and all of a sudden AI, the bots are all just working. There is some additional effort that's required, and we'll also need to keep our eye on the future. What are your thoughts about what's next
in this space?

[16:42 Brett Weigl]

The one thing I would throw out that we haven't really talked about is that I think the age of being able to realize the next best action framework in a more comprehensive and human way is going to be possible so that we can get customers what they need in a really
personalized way, sensitive to their journey.

[17:02 Brett Weigl]

That's going to require, yes, gen AI, but also more traditional AI to the point of your prior podcast. That's not all gen AI. Predictive technology has a huge role to play in terms of, how do we actually thread that needle and get the customer not only their answer, but their answer sensitive to the journey that they're on?

[17:23 Brett Weigl]

And then, you know, I think even further out, how can AI start to make its own decisions, become more autonomous? I'm a service advisor who needs a job to be done in the back office. Can AI run away with that and come back and report back when it's done?

[17:44 Brett Weigl]

That is a huge area of focus with agentic frameworks. A lot of AI companies are looking really seriously at that, and so there are going to be great things to come on that front too.

[17:49 Weston Morris]

I'm going to steal that "next best action" phrase. That's a good one. Really, if we don't get to a resolution or an action, to me this is all a waste of time, right? It's just all theoretical.

[18:05 Weston Morris]

Aron, what are your thoughts about what somebody needs or an enterprise needs to be doing and thinking about as they are jumping into this Wild West of AI here?

[18:14 Aron Meyer]

To make sure that organizations are not having it just be the Wild West for them is really thinking about the data. And I'm sure many organizations have likely already put into place some kind of levels of data classification, whether it's public or confidential or restricted.

[18:36 Aron Meyer]

But they also need to be considering the temporal nature of that data as well, right? Is this
article out of date? Is it still relevant?

[18:46 Aron Meyer]

And then thirdly, are the persona roles within the organization. Should this individual have access to this data or should this individual outside the organization have access to this data?

[19:05 Aron Meyer]

And so when you think about those three key attributes of every one of those data sources that you're going to use to feed, whether it's the AI bot on the front end or it's your assistive AI that's helping your employees and your agents, my recommendation is to make sure that you're starting with that data classification as it will really pay off when you are feeding AI those data sets.

[19:32 Weston Morris]

Well, Brett Weigl, Aron Meyer, it has been great talking to you both today. We've captured at least four distinct areas where AI and gen AI can play a role in the service desk. It's realworld stuff, but also some cautions and lessons learned. I really appreciate it. I will now thank you both for joining us here today.

[19:52 Brett Weigl]

Thanks so much, Weston. Thanks for having me, and great to talk with you, Aron. As always, if folks have questions, feel free to reach out on LinkedIn. My handle is slash Brett Weigl, all one string. You can get me there, and I look forward to continuing the
discussions.

[20:13 Aron Meyer]

Yeah, and thanks, Weston. Likewise, I love answering questions about all things customer experience and AI, so feel free to reach out on LinkedIn. I'm also just like my name, slash Aron Meyer, all one string.

[20:26 Weston Morris]

Excellent. Well, we'll share those links in the podcast notes as well. I really appreciate you taking some time out of your speaking schedule to speak with me on our podcast today. Well, folks, you've been listening to Brett Weigl, the senior vice president of product for AI at Genesys, and Aron Meyer, the solution manager for digital workplace solutions here at Unisys. I'm your host, Weston Morris. Thanks for listening.