Beyond Pre-Planned: Harnessing Real-Time Variability with Erlang-O

Hi, everyone. This is Vicky Harrell, executive director of SWPP, and we’re so excited to have you with us today for our web seminar, Beyond preplanned. We’re gonna learn about harnessing real time variability with Erlang o. And we have had a great response to this web seminar, and we got a lot of you out there today. So we’re excited to hear what these folks have to say about a new Erlang a new Erlang letter as we talked about it before that that, Erlang o, maybe next will be Erlang p. But we’re excited to have you with us today. Excited to have Ted Lango and Dave McCutcheon and Jim Simmons with us from to to get all this going today. Just a couple of other things I wanted to mention, housekeeping announcements. I’m recording this and will send out a recording afterwards with a PDF of the slides for everyone who is registered. So, you may not have to take furious notes. If you want to review the recording, you can also share the recording with other folks in your organization that may not have been able to attend. We want you to use the chat on the bottom right hand side of your screen, as as a way to give feedback and ask questions. Make sure that it says to everyone where your chat box is on the bottom right hand side of your screen so that you will be able to send it and everybody else will be able to see it. So, if you wanna tell us this morning, just put our, cursors in that chat box and let me know where you’re listening from this morning. And, I’m hoping that you guys have some good weather where you are. We’ve got a little bit of hot weather today, but this weekend in Nashville, we’re gonna be in the seventies, and it’s gonna be gorgeous. So we are excited about that. Like, we got people from all over, Panama, Louisiana, Fort Lauderdale, Maine, Franklin, Tennessee, Albuquerque, Peoria, Illinois, Sunset Sunset, Arizona going to a hundred and twelve today. But it’s a dry heat, Brian. Right? There’s an oven. That is an oven. That is definitely an oven. So Philippines is coming in strong too. A lot of Yeah. The Philippines. We got we got Canada. I bet it’s nice in Hillsborough, Oregon, Kristen. Alright. Looks like we got a lot of people in and a lot of different places that are out there. So we’re looking forward to hearing from Ted and Jim and Dave this morning. Ted, I’m gonna turn it over to you and let you get started. Thanks so much for being here with us. Oh, thanks, Vicky. Thanks for having us again. Good morning, good afternoon, good evening. I’m Ted Langos, senior vice president, at IntraDiem and also founder of WFM Labs. I’m excited to share, some work that we’ve been doing around a new approach to Erlang. Jim? Jim Simmons, cofounder of Qulis. I’ve been working in the industry for about twenty six years now and and very excited to talk about what, you’ve been working on, Ted. David. Hey. I’m Dave McCutcheon. I’ve been in the industry twenty plus years. I stopped counting. Yeah. Super excited to see what it’s got and and really dive into the new airline layer, I guess. Newish. Newish. Newish. Great. Great. So, we’re gonna go over three areas, today with everyone, speak a little bit about why, we believe the traditional methods of Erlang fall short, in today’s complex environments. We just think that there’s a better approach. That’s the key thing we wanna share today is a Erlang o, a different way that you can look at your staffing line and how you think about calculating how many people you would need on the phone at any point in time. And then from a a a final thing is how you actually enable it. How do you actually get your staffing enabled with automation so that you can make this new approach work itself? At the end of the presentation, we’re gonna give you a QR code where you can download a paper on the topic itself. In addition, some Excel based tools and Python code for visualizing how Erlang o works itself. If you have questions as we go, please type them in the chat itself. And then based on the time that we have at the very end, we’ll, try to answer as many questions as we can. To get us started, let’s basically, use the chat again. And, why don’t you share if you’re willing to share how you actually calculate your staffing requirements today? Do you use Erlang c? Do you use Erlang a? Do you have your own method of machine learning? I’m curious to hear back from folks in terms of how you’re using, a calculation. It could just be historical averages, to actually calculate how many people you need on the phone at any given time. Some Erlang c’s, some Erlang c’s and a’s, some plug ins. Go ahead and keep typing those, and I’ll share with you, some interesting data from our friends at WeWFM. They did a research paper, about a year ago, a little less than a year ago. Sixty four percent of the industry, is still using Erlang c to calculate your staffing requirements, and I’m seeing a lot of that, pop up here. Historical, actually, which just popped up is the second largest, area. A lot of historical averaging, to calculate the staffing requirements. And then roughly, the last twenty five percent is a mix, between machine learning, linear programming, simulation, different optimization algorithms, proprietary methods. Some people may be using proprietary methods, but the vast majority of people are still using Erlang c. And why is that? There’s an awful lot of us who are answering that right here, but why is it that people are still so focused on Erlang c? And I think a a good reason for it is it’s a simple and effective way if we know, you know, how many calls in a given interval and the volume or and the handle time, for those given calls, and we’ve got a target speed of answer, eighty percent of those calls in thirty seconds, we can use this relatively simple equation to predict the probability of wait time and ultimately to back into the number of agents. And if we plug that in to one of our calculators or plug ins or macros, we can find out that we need twenty agents for that particular interval to service the fifty fifty calls at five hundred and seventy seconds. But this has some real limitations. One of the key things is I don’t know anybody. You can chat chat in the the box here as well. The chat box is does anyone have a contact center that’s that simple with just a single queue, you know, with a a pool of agents? No. Most of us have many queues. We have different goals behind our queues. We have multi skilled agents in almost every call center I talk to these days. We have different channels of work. It’s not just all phone work. Might be primarily phone work, but you may have chat or email that you’re flipping agents back and forth to. We also have abandons. At the end of the day, Erlang c doesn’t address abandons, Erlang a does. But in addition to abandons, Erlang c thinks that our traffic is just kind of a continuous stream, and we know there’s variance in demand, not just at the interval level, but how calls come in throughout the day, there’s variance. And there’s variance not only on the demand side, but on the supply side. And Erlang c doesn’t know anything about that. It just says put the fifty agents in there. It’s not focused on the fact that, hey, some of those agents may not show up today. So while this traditional method has allowed us a way to to calculate a precise number of agents required to meet the service level, this approach has really taken our entire industry into an ingrained preplanned mindset. We aim to precisely schedule as much of the activities as we can ahead of time to tie back to those calculations. And whether it’s Erlang c or Erlang a or if you’ve got your own machine learning model, this is what you’re doing. You’re saying we need twenty agents at this particular time or a hundred and twenty agents, and then we work around that to try to preplan all of our activities. It’s a noble cause. I mean, we’re we’re trying to do the right thing, but there’s some real implications to precise planning with algorithms like Erlang c. So let’s illustrate, like, what those implications are. If we take a demand line throughout the day, we’ve open at seven o’clock, we close at eleven, and we basically plot out our required staff. We’ve plugged the data in, and we know, like, at eleven thirty, I need a hundred and twenty people. We go out. We schedule those hundred and twenty people, but then demand comes in throughout the day. We have demand variance. If you look at your PDPs, a prior day performance report, sometimes these numbers will line up, but often as intervals go along, like at the eleven thirty interval, we’ll see that we actually needed a hundred and twenty three people. Variance is an important factor here. One of the things I like to share and I’ve shared in past sessions is there’s something called minimal interval variance that you can’t forecast away. It’s an actual math calculation. Herakula from CC math does a great job of explaining this. But, essentially, you can only get so precise based on the size of the queue. There’s always gonna be variance in our our demand side, that we just can’t forecast away. And then once the supply comes in, we also know that there’s variance in how the agents show up. We could have scheduled a hundred and twenty, but maybe only a hundred and seventeen showed up at that interval. And by the way, if we swung to needing a hundred and twenty three and only a hundred and seventeen show up, we all know what happens. Our service level takes a hit and can take a rather serious hit, with just six six agents. At the end, what we end up with are really two types of risk. We have service level risk on one side, and we have an expense risk on another side. Every single day, our organizations are fighting either, over over demand, undersupply, or the opposite. And every one of these areas where I’ve shaded in either an orange, which represents service level risk, There’s also an area of expense risk that’s that’s green here. You sometimes can forecast this away. I mean, to a degree, you’ll every now and then, you’ll land on a perfect spot where these lines cross, and you have just the right amount of staff against just the right amount of calls. But if you look at your PVPs, we really ebb and flow back and forth all day long. This is where Erlang c really only tells half the story. It cares about talk time and it cares about availability. It knows what are those gaps that we need to put in to ultimately have the right type of availability to hit our service level. It really doesn’t know anything about shrinkage. The fact that you may have to schedule time to go have people work another queue to do training, to do, coaching. It doesn’t know about these alternative channels. It doesn’t understand about minimal interval variance and volatility as a whole. And this is an area where it’s like we just think there’s a better approach, here to to looking at your staffing plan rather than trying to pin it right on on the, you know, the head of a needle here. Jim, anything you wanna add to this? Yeah. I think, you know, one of the things in in my time working in the contact centers, and I and I mentioned I’ve done it for twenty six years, and I sort of started in WFM, moved to a few different roles, and then came back into and, finished up with Synchrony in WFM. And one of the things that really surprised me was how little had changed in that period. And I I think that’s true. We were kinda talking a little bit before the webcast that, what would we think mister Erlang would say today about us talking about, his formula. And I think you’d be very surprised that he developed it so many years ago, and we’re still so dependent upon it. And part of the reason is because, you know, if you think about what we do today, we capacity plan. How many people do I ultimately need to get hired and in the system to handle conversations, or chats or what have you? Erlang c has really dealt with over the years, where do I place those folks at any interval section? So the gap that really exists is between developing that headcount and placing that headcount. And that’s one of the things that ever since you started talking about and working on this formula, I found most intriguing. And, you know, kinda looking at the chats and knowing the conversations I’ve had, I think it’s something everyone deals with is how do you bridge that gap between my capacity plan and actually getting somebody scheduled at a particular interval? So I think this is a great bridge for that gap. Well, thanks, Jim. Dave, anything you wanna jump in with? Yeah. Really you really just touched on it, but the way I would sum it up, you know, airline c works great, if everything that you forecasted and predicted comes to pass, but it never ever does. And and is kind of Ted touched on. Right? Things are getting more complex. So just the slightest bit of of variance to any one of your metrics or any one of your forecast or anyone by demand, anything that moves, yeah, you know, airline c is just it’s it’s not allowing you to as as as, had talked about. Right? You’re either gonna miss service or you’re gonna be over capacity. So it’s a very sort of I would say it’s a very rigid way of doing things. And the reality is in the modern call centers, they, like, you need to have as much sort of flexibility as possible and need to be able to to account for that, where whereas airline seat just is very rigid plan. So And so so what’s a better way? We we basically have been working on, what we believe is a superior approach. We’ve called it Erlango. At the end, of this, we’ll have a QR code, and also a link where you can download, the paper on this, but also some tools, to help you visualize this. I’m not going to go into all the math behind it here, but I am gonna try to illustrate, within the math what it really means to approach it a little different how we’ve described it through the math. I’ll also say an important concept here is that we’re not throwing Erlang c or Erlang a or your core machine learning algorithm aside. There still is the need for a core algorithm, whether it’s c, a, b, or your own proprietary machine learning to understand the workload, and the the service level goal itself. So don’t look at this as, hey, Erlang c, you know, is obsolete. We just believe it needs to be built on in a different way. So how do I describe this? Well, we’ve no Erlang c or a basically takes your talk time and your availability. At the end of the day, what we’re looking at is your shrinkage, your intraday productive shrinkage. We’re looking at your alternative channel work, emails, chats, other work that may not be as time sensitive as the queues that have to be answered eighty thirty. And then we’re also looking at variability and volatility. The variability is a factor that I I have yet to run into someone who’s planning with minimal interval variance built into their plan, but it’s real. There are intervals that are gonna be more. They’re gonna be less. That’s natural variance that you can’t get away. And by taking the core Erlang c and extending into these other things that we plan for as a part and that should be in our budget, we’re basically looking at those as overhead. And that’s why we we chose the o as as the letter itself is Erlang o is really Erlang c or a or your base algorithm. But now looking at your staff level with your overhead built into it, We do a little bit of this already today. You know, whether we recognize it or not, when we have any given interval like eleven thirty, we’re gonna get four hundred and sixty calls at four hundred seconds. We go through and we calculate, and we say, hey. In order to hit an eighty thirty target, I need a hundred and ten people. But do we schedule a hundred and ten people? No. We we’ve got day of absenteeism. Let’s say it’s Monday, and I know that we’ve got an eight percent absenteeism rate. Our software or if you’re doing it in spreadsheets considers that eight percent absenteeism, and it allows you to correctly adjust your staff line to represent what you expect to have as call outs. If you take the same concept and extend it to other productive off phone activity to other channel work and incorporate your variance, that’s what Erlang o is. And, again, I won’t go into the math behind it here on the Webex, but the paper you download along with some of the tools can show you practically how you factor these things in to create a new more resilient line. Let’s look at what the line, looks like kind of graphically here. As I said before, we’ve got these intervals where we have service level risk and expense risk along the way. What Erlang o allows us to do is sort of rethink about that risk. Originally, we would have staffed right to this based on the talk time and the availability that we need to build in along with the call volume. But what Erlang o does is it allows us to consider shrinkage, alternative channel work, that variance and volatility, and build that overhead into your staffing line itself. At the end of the day, that is Erlang o. It’s a new staff line that incorporates the overhead. In this new staff line now, what have I done? I’ve basically written above my ebbs and flows with service level. So I’ve started to mitigate my service level risk by openly scheduling my people to cover that minimal interval variance, to cover the areas of volatility. Now there’s a second half of this equation, which is, well, what about expense, Ted? You know, at the end of the day, you’ve just, you know, raised my staff line. Now I’ve got just more and more expense risk. And that’s absolutely correct. If you just inflated your staff line, Erlang o wouldn’t be overhead. It would be overstaffed. So how do we address that? At the end of the day, part of that expense risk is eliminated through natural variance. If I basically knew that my call volume was gonna come in higher this interval, I would have staffed for it, but I didn’t know. So part of it is absorbed through the natural variance that you have throughout the day. The other part of it basically requires automation. At the end of the day, we have training. We have communication. We have coaching. We have emails, alternative channel work. If you integrate automation into your function, you can then optimize your net line in real time with productive activities all day long into the evening. Automation runs in the background to deliver all those things that you would otherwise try to preschedule. This basically allows us to have our cake and eat it too. It protects the service level. It still is optimizing your expense. And then a third bonus on this is it really allows us to invest in our people. This is an area where before I joined Intradiem, I deployed, automation at MetLife. The first year after I had taken this approach of moving away from preschedule and delivering these things in real time, I ultimately ended up delivering three times the amount of training and coaching than we could ever deliver when we were trying to preschedule that. So it’s a it’s a it’s a triple triple play here. We end up protecting our service level. We optimize expense, and we ultimately end up investing in people. So there is a piece to this that we need, you know, to to advance on. You know, we just can’t put a magic formula in, and it it comes home to to to to manifest without an expense risk. We really need automation as part of this solution. So how do you get there? At the end of the day, you have to advance your WFM operations themselves. I’ve shared this in the past with probably many of you on the phone. About a year and a half ago, a number of us, including Jim and Dave, developed a new maturity model for workforce management. It’s a five level model that basically starts over here on the left. There’s a number of us on the phone who probably are still in level one where we’re doing Excel based, planning. I’d say a good number of us are in level two where we’ve got WFM software. We’ve moved out of, that that Excel based model. Level three is what starts to incorporate automation. Level four talks about risk assessed capacity planning and goes after solving your budget problems. And level five really is where we all need to get to sometime rapidly because AI is coming. There’s a lot of things that are coming at us that we’re not ready for really in level one or level two. About twenty percent of the industry right now is in this level one. The other seventy five percent of the majority is still in level two. We estimate about five percent of the industry has migrated into this level three, four, five. So what does it look like? What does it mean to to move? The biggest things right here, if you’re in level two, if you’re in level one, you need to get out of Excel. But if you’re in level two, it’s looking at all these static activities that are prescheduled ahead of time and us basically firefighting all day long. It’s looking at that mindset and basically replacing it with investments in automation that involve the delivery of training, communications, automated activities where we can basically poach in real time as opposed to prescheduling it. And the automation has all kinds of other use cases itself that you can go after here. That’s where Intradiem comes in. We’re basically a a complimentary component between that lives between your WFM systems and your ACD. There’s really no other solution on the market that integrates like we do back into your ACD in real time to read queue stats and to read agent stats. And we also integrate to your WFM platform, and we write back and we can write back codes to your WFM platform as we deliver activities in real time. We integrate to all the major platforms, standard integrations to cloud and premise based ACD and WFM providers. And at the end of the day, the technology itself, it’s not for IT to write. It’s really for the WFM team to write the automation rules themselves. It’s a simple if then logic that allows you to build your own rules based on your use cases. And finally, at the end of the day, while we integrate to the ACD and the WFM and we talk to those continuously, we also integrate as well to other channels. So because if you’re gonna deliver training in real time to agents, they need to be able to know what’s going on. So we integrate to the desktop along with other channels to communicate when we’re doing things like delivering coaching in real time, delivering training, and making adjustments. Intra Diem has a number of use cases, and this is where I’m gonna have, both, Jim and Dave jump in and share some of their experiences. We’re not gonna touch on all these use cases today, but I do just to bring it home, wanna touch on two of them. The first being training. We all know, at the end of the day, we go after trying to preschedule the training. Half the time, we may end up canceling because we’re fighting service level. It’s difficult to find any good times to deliver these things. So by, leveraging automation, we can build rules that simply look at conditions that you design. Longest call less than ten seconds, calls in queue less than three three in queue, and then we can deliver in real time continuously throughout the day prompts to the agents to conduct off phone productive activities that you otherwise would be trying to preschedule. I’m gonna, take a breath here and just turn it over first to Jim. Jim, ones that stick out in your mind from your time in terms of, what really hit home here with this use case? Yeah. Sure. And and before I hit that, though, I do wanna just take a real quick step back and re reference the Erlang o calculation and and the overhead fact. You know, one of the things that I I know I saw a lot in my prior role, and I know it happens all the time is that we tend to try to intuit to that upper line anyway. I saw I made note of a comment, from Tiffany, when you asked about Erlang, who’s using Erlang, and and her comment was we use Erlang c plus a little extra. So So I think it’s important for folks as we go conceptually and we talk about automation to recognize how we adjusting for it today. Everybody’s got their own way, but in some way, shape, or form, we are trying to intuit this. So one of the things that I really like about the o calculation is that you have an official calculation that you can work from and reference and adjust accordingly. And that really leads into where we were at when I was leveraging, IntraDiem at Synchrony. We tended not to do a lot of preplanning, but we leaned into that sort of overstaffing. And we tried to manage that with humans, right, which is is just impossible. You can’t manage, you know, a contact center of any real size with a small team or even really a big team of people that are trying to immediately place hundreds, if not thousands, of agents at any given five minute interval in the day. So where we struggled was, how do we effectively get people where we need them, whether it’s on the phones or into a critical training or looking at some critical communications? So we really leveraged it primarily out of the gate for training and for communications. Right? We had some specific communications. We’re in a highly regulated environment. So we needed to make sure every day agents were taking just a few minutes, maybe at three minutes, five minutes in any given day to go through some updates. So we leveraged that. We also used it a lot for training. Again, a lot of regulatory training that comes down the pike, very quickly, and you’ve gotta get it done efficiently. And we saw tremendous benefit, in reducing our idle because, again, we had sort of tried to kick the ball and overstaff a little bit and and manage to the top side of that line. And even if we didn’t, to be frank, we had a couple clients that we manage much more closely. The downside there was if you’re managing very closely and you don’t have a lot of idle, it’s even more difficult for humans to find those few extra minutes to get things done. So we were fortunate and at Synchrony when using it to get not only rid of some of the excessive idle, but when we had those skill groups that had low idle, we were able to quickly pick those things off and get those trainings and communications done. And it was a significant benefit. We exceeded our expectations, certainly at year one and going into year two. Thanks, Jim. Dave, anything you wanna add? Yeah. So kind of the the practical piece coming back to the to the, if anybody is from finance on here and and had put that first video up where he showed the extra staff and probably the immediate reaction was. And so and a few people have mentioned that, we we cannot do this anyway to a degree. And Jim touched on it, but you’ll never be as efficient when you’re trying to sort of manage that situation up or down, real time. Everybody, you know, to a certain degree will take a stab at trying to do that and and and do their best to do it. What’s really interesting about this, and it’s really, a a a kind of a a shift in your mindset, like, how do you think about this? It’s very to me, it was a little uncomfortable to not be preplanning everything because that’s just how we’ve done it forever. And so to begin to move away from that way of thinking about it was like, well, what what if we can’t get this done or or, we have to do this, but we’re not gonna preplan it. And so, you know, as as Ted mentions, there’s certain rules that you can set within, within the software, within the system that allow you to sort of still have that staffing and still be able to make sure that you’re, achieving whatever you know, if maybe you have a compliance training, it must must happen, then that that can still be driven regardless of service impact. What I found was really sort of interesting with this as well, though. You know, there’s there’s such a push for efficiency now. So the first component of our line was, like, where we’re gonna put more people in in theory. That’s not necessarily what we’re seeing here. What you’re what you’re able to do with this, as I said, is is drive the efficiency. So think about just training right now, and think about your associates. If you set a fifteen minute training up or a thirty minute training, they’re going to use the whole time. What’s really kinda neat about this, you can ask you know, maybe somebody gets done in ten minutes. You can put them right back in or you can give them an extra break. You know, you can help with engagement from that perspective. Maybe somebody takes a little bit longer. It’s kinda painful as a workforce management group when somebody doesn’t complete a training and you have to adjust the segment and you need to update this. And so all of this can again coming back to the point about efficiency, you can drive this data efficiency through that and and through the automation. So, as as Ted talked about on that maturity curve, that’s where we need to head in the industry, really drive more of this automation in our place. And then we’ll be able to realize these these changes to how we think about staffing and and really move away from this preplanning. So Yeah. Thanks, Dave. I and at the end of the day, it I think, Dave, you hit it right on the head as well. We’ve I look at it as we’ve been kind of led down this path where it’s a noble cause. We’ve got this precise calculation of how many agents we need with our core algorithm, and then we do everything around it to try to ensure that we have just those right number of people on the phone. But every time variance hits us in a negative way where we’ve got a mismatch and over demand under supply, then we’re ending up having to cancel, reschedule, cancel, reschedule. So all that work, while noble, it’s not the most effective way to deliver training or coaching or other productive off phone activity. It’s a prime use case for automation. As I mentioned before, I was taken back by the sheer volume of training and coaching that I delivered the first year after we put automation in. And at the end, there’s less work for my WFM team, but we didn’t have to go through and do all that scheduling. Second use case that’s very similar is coaching. You know, you go out and you attempt to find the right times to match up the supervisor and the agent. At the end of the day, this is another prime use case for automation. In our product, we actually enable the supervisor themselves to set the list of people who they wanna coach, who’s first, who’s second, who’s third, how long do they wanna coach, but then we don’t go and schedule it. We allow Intradiem to find the right time based on queue condition to match that agent and that supervisor to together. There’s an override if all of a sudden I need to coach now because Jim basically just said something he shouldn’t on the phone, and I really need to interrupt him. But at the end of the day, if I wanna, you know, coach Adrian or Ben, those are my first two. Even if it’s just a quick touch point, I could put two minutes in there for my time. And then Intradiem’s gonna go out and say, hey. No calls in queue right now would be a great time. Deliver the pop up to the supervisor and the agent. Say now is the time that you guys can hook up together, ultimately, harvesting those pockets of variance where supply and demand are in your favor. Delivering it then at the end of the day and allowing you basically to to, ultimately just deliver more coaching, support your agents. Ultimately, this for me, I saw a longer term impact on attrition. We were able to really invest in agents themselves and and support them the way they needed to. I’ll just, pause again, too and and flip it back to Jim and Dave. Any other use cases or items you wanna point out, before we wrap up and go to questions here? I do. And, actually, it’s it’s prompted by one of the questions that that Jordan asked. Can IntraDMB configured to prevent agent burnout, no overutilization? And and that one spoke to me, because that’s another big use case we had. And and, really, there were a couple examples, and I know there are more out there that that we could talk about. But for our implementation, one of the first ones was, I mentioned, for those agents that were really struggling with, high occupancy, being able to get them off the phone, some of the feedback we got was, wow, it’s nice to finally be able to get some of the training people in those other skills we’re getting or to get time to read through communications. However, what we did when we launched the system was we gave everybody one additional five minute break in a month just to make sure that they were looking for prompts. And we really honestly didn’t necessarily think the employees would see it as a big benefit. We just wanted to make sure that there was enough of a hook that whenever a message popped up, they looked at it and paid attention. The feedback we got on that extra five minutes a month was the most significant positive feedback we got about our implementation. And then finally, we partnered with a wellness, app, company, Thrive, and and brought them together. And and they are now partnered with Intra Diem that actually sends out wellness breaks, throughout the day. And sometimes, you know, you can configure it when there’s a little bit of idle, but oftentimes, you need to configure it when it’s not idle. Right? When it’s really busy and most hectic. And what we got from that, again, aside from the feedback from the agents, what we found is we really weren’t suffering on the servicing side either because it was so efficient at getting those in that the agents were getting this little break that they were taking anyway. Right? Let’s face it. If if you’ve got agents that are in a high stress situation, they’re finding ways to get that. They’re using ACW. They’re going into unaccounted for a x. There’s a myriad of ways. So for us going, hey. We’re gonna put your mental health first and make sure that you get those breaks was a big win for us. So those are just three examples of what I know, and I know there’s some additional products out there as well, that address that one. So Thanks, James. Dave? Yeah. I I wanna kind of, take a bit further into what Jim was seeing. One of so, and I know there’s a lot of questions, and we will get them all, but it’s it’s and, Jordan, you sort of, touched on a good point. That is one of the things that I think really, with this kind of model drives a lot about. Since we think about and you may be different in your call centers, but a lot of, like, the sort of level one, level two traditional way, of thinking about staffing and running your call center. Schedule adherence can be an ugly word and there’s not it can be something that really stresses our associates depending on, you know, how heavily you focus on schedules. Maybe it’s part of your bonus structure or maybe it’s not. But it is something that that can be a stressor for, for your associates. And so one of the things that’s kinda nice about this sort of model is and so, again, going back to the charts that that Ed was showing as as we get the the variability, through automation, you can begin to move things like breaks, lunches, and you can do it automatically. So taking a lot of the burden off of your workforce team. If somebody gets stuck on a long call, well, they’re stuck in a long call. In some cases, that might be an impact to your scheduled appearance, which then might be a stressful good associate then they gotta talk to supervisor. You gotta get that fixed. And, and so a lot of extra work, but really no value in theory, and actually it’s adding stress to the associate. So you completely remove that in this example where it’s automatically adjusted. So that’s a benefit. And then the other piece I think about is, again, from a stress perspective, a lot of lot of kinda really COVID changed the game in many cases. And so a lot of centers, maybe you’re still remote. Some of you might be back in office, some of you might be remote. But I think about a lot of the places where there’s still a lot of remote work, and there were some benefits to doing that. But if you think about your traditional call center now where maybe your supervisor is sitting with your team, and they can see they they can see your associates. They can see what’s happening. They can see maybe if somebody’s having a tough call or a really long call or something difficult, which, again, leads to burnout, leads to a negative associate experience. How do you do that in a virtual world? And so one of the things that we’ve thought about from a rural perspective, is is if we we have a a really long call, it’s probably been a difficult, tricky call. And so we make that assumption, and you can put in surprise breaks. You can put in additional sort of benefits to that associate and just give them a breather. To Jim’s point, they might take that anyway, but you’re doing it more proactively, and that comes across in a much more positive way. And it shows that you really care about the associates. So, again, you’re able to do that with the with the variability, and you’re able to offer these types of things. So Yep. Yep. Great great points. Thanks, Dave. So, I’ve explained kind of visually what Erlang o is about. We’ve got some assets for you that you’re welcome to, go and and grab. I’m actually gonna put, the link in the chat, because if you take the picture on your phone, you can go ahead and download them, from your phone, but you can also just click on this link in the chat if you wanna, gather the assets. The assets include, a paper that explains the math in a bit more detail. I’ve also, provided two Excel files. One is, not macro enabled. The second is macro enabled that shows you the math behind. If you can’t run macros, you can use the other version, but you have to manually put in the, headcount number to to generate the, service level results. And then the third one is a actual Python, program. If you’re into Python and you wanna take a a shot at running Erlang o, that assets in there as well so that you can see what the new Erlang o staff model would look like. But more important, it it’s illustrative of what we’re doing here mathematically. You won’t wanna use these to actually schedule people, but they’re just, tools that I think help people in the WFM mindset really visualize the math behind the data. Finally, before we kind of switch over to questions, and answers are here is that, implementing, automation isn’t something where you’re, you you know, you’re giving something back or having to spend more. There’s a huge ROI benefit to bringing automation into your organization itself. At MetLife, I saw three x. It was over three hundred percent ROI that I achieved. Our top ten customers are are represent over a six x ROI and are on, tracked across, like, over a billion dollars of savings, over the last four years cumulative savings over the last four years. So this the investment in automation, you know, you win on the service level standpoint. You can do more protection there. You are, not only optimizing expenses, but potentially lowering expenses, dramatically and investing in your employees as well. So there’s a real strong ROI behind, you know, investing in your WFM organization to bring automation to the table. So that’s all we had to to present with you today. We do have a a few questions here that we wanna go over. And if you have other questions you wanna ask, just plop them in. There’s a couple that I grabbed as well. We addressed one out there, but there was a question about does Erlang o, consider factors related to multiskill agents. We all you know, almost all of us are dealing with multiskill. The answer is no. It’s not the part that’s considering the multiskilling in that perhaps the calculation of just how many agents I need based on that workload. You still need a core algorithm that’s gonna look at how many calls, how many agents, how many calls, what’s the handle time, what’s my service level objective, and how many agents that I need. The people who’ve migrated away from c or a, Erlang c, Erlang a over to machine learning models are doing so because they’re getting more accurate predictions around how many people you need on the phone based on that multi skilling. Erlang o is an extension of that. It’s saying don’t staff. You you may think you need precisely a hundred and twenty people, but don’t staff to that hundred and twenty. Consider the variance, the volatility, and the fact that if you deliver coaching and training when the ebbs are in your favor, you basically are gonna be better off. So, no, it’s it’s still required of your core algorithm. Erlang o is an extension to that particular algorithm. I think we kind of addressed the the burnout question. The only thing I wanted to add, and I’ll, see if Jim or Dave wanna add anything, is in addition to doing things like offering surprise breaks, we have another product now, that’s predictive of attrition. We’re just in the process of rolling that out with a couple of clients. We now can see leveraging machine learning, the signal in the noise of when an agent is starting to burn out so that we can take activities to prevent that. So it’s kind of a yes and a yes. We have, automations you can build that ultimately help prevent burnout. I could write a rule that said, if Jim took six calls in a row and they were all over five minutes, let’s offer a surprise break. I mean, that’s configurable to your own design. And then there’s other ways as well that we address it. Anything that either you gentlemen wanted to add on that particular question? Just to that product, I I always try to point out, you know, keep in mind, if we know what drives attrition, which the product does, we also know, conversely, what helps keep people happy. So I I love the fact that you can play both sides of that with with the new product. Not to add, but I I I was gonna keep us moving and maybe take the next question. This is quite a few questions in the market. Yeah. So sweet spot, for preplanned and saved day was the next one. And, I wanna get here both your guys’ answers on this for the the question is, is there a sweet spot ratio between doing preplanned and just delivering everything in an automated fashion. And and there’s probably a few dimensions you can look at this for me. One of the big dimensions was the length of time of the activity. A second dimension was the number of people. Am I delivering something that’s just to Jim, or do I really need a ten person, meeting that’s gonna be twenty minutes off the phone? Those are difficult to pull together with automation. So, yes, there are some activities. Or if I’ve got a training that is two hours long that somebody has to take, you might as well just schedule that. But we found anything that was, like, less than thirty minutes, we would break down into maybe two fifteen minute chunks. We’d load it into the automation tool, and, ultimately, we could just deliver far more content training, coaching sessions than than prescheduling. I’d just like to hear from Jim and Dave, though, if you’ve got another view on that question. Yeah. I don’t know, Jim. I’m jumping in. I would say so I would sort of say it kinda depends, on your center and your advocates. If there’s a concern that, like, you know, advocate feedback is, well, nothing’s prescheduled, then that that’s giving them anxiety. If you’re thinking of going down this path, then possibly you bring them in as part of that process and you you get that direct feedback. There might be things where, like, I I’m really I need to see my brakes. I have to always have my brakes. That’s fine. You can you can schedule those brakes. But as Ken mentioned earlier, what if you get stuck on a call or something happens, and so you can also implement automation to move those brakes automatically, which helps. And then, you know, other other things like train ins or check ins, they they can kind of be there if necessary if it’s if it creates angst. But, again, I think I think really having the conversation with your associates and explaining why and and and heading away from the sort of that traditional way and how it benefits them. So they understand they’re not losing their hair. In fact, they should be gaining from this. So it’s just for me, it’s like, well, may I don’t know if there’s a sweet spot. I actually think it just depends, depends how much you wanna lean into this. You know, it’s you’ve got a maybe there’s something that’s not right for automation. It’s an and our long term hope, maybe that’s not easy to move, and and we just have to have it at that time, and and then that goes in. But other ones, I would say, probably, as long as you’re talking with your associates, you have them understand, then I don’t know, Jim. When when we looked at it, when we first implemented and we looked at everything that could, kind of excluding what Ted and and Dave have already mentioned, we were, you know, in the thirty to forty percent range that could easily and then we had set a goal well over fifty percent that we wanted to get to, and that was really working with training in HR, giving agents more flexibility on breaks and lunches and things of that nature. So we started at one point and aspired to a greater number. Makes sense. Another question about the size. So is Erlang o a beneficial technique for a small call center? It’s beneficial for any size call center. If you can do things in real time, there is a, an aspect of you need to introduce automation to make this work. Otherwise, as I said, your Erlang o will stand for just overstaffed if you can’t pull the staff out in real time to optimize your expense line. In terms of interdiem, we generally look at two hundred or more in terms of the ROI making sense. You do need to have a little bit of scale to deploy a tool like Intradium for the automation, but, the short answer is it’s applicable for anyone. If you can build your automation to deliver things in real time, you know, whether you’ve got ten people, a hundred, or ten thousand, it’s a applicable, concept here. We answered, I think, the question about can things be a hundred percent automated. It really depends. You wanna automate as much as you can, but there are things that you can’t. There’s a question here. Can we get different simulation based on different KPIs through Erlang o, and how will it deal with teams queues of cross skilled? That’s actually my same answer. Again, Erlang o is more an extension of a core algorithm that you’re gonna use in your your center. If it’s c or a or machine learning, you need to think of Erlang o more as an extension. How does Erlang o carry forward into influencing the intraday requirements defined by WFM? How does Erlang o carry forward in influencing the intraday requirements defined? At the end, I guess my answer on that would be you really have to look at Erlanger as a separate calculation than your core WFM, requirements calculation. If you are using IX or Verint and it’s basically saying, okay. Here’s our algorithm. That’s doing an attempt to line up the precise number of people that it thinks you need based on the volume, the handle time, and your service level goal. Okay. So if you’re gonna go after Erlanger and make it real, you have to think about it similar to how we put an absenteeism rate into these tools. You’re building your overhead instead of prescheduling into your WFM tool to inflate that line rather than sitting there and trying to preschedule all the coaching and training and activities themselves. Let me see if there’s another. Can Erlang o calculate FTE for a dedicated email queues where service level is eighty forty? I would still say you wanna go back and use an Erlang x or Erlang, core workload calculation. No. The short answer is no. It wouldn’t do that. You would still wanna use the mathematics for that, of whether it’s Erlang o, Erlang c, or Erlang x, that works with different channels. Can you use it to push VTO? Yes. Absolutely, Mark. I think that’s probably Mark Williams. How are you doing? At the end of the day, VTO is a great example, that can be delivered in real time based on your net line going forward. Automating those activities is a part of what, our automation can do as well. We got a lot of questions here. Can Erlang, can you push we did that one. Does in oh, no. We’re getting toward the end here. Does InterDiem look at the forecast for the next few intervals in real time before letting rules run, or is it currently looking at only current queue activity? Oh, now I need a Larry on the phone for that one unless you gentlemen know that. I’ve used it all for real time. Yeah. It was so it will look you can set that standard up. Right? Yeah. I I was gonna say that the end of the day, the the looking forward can be triggered off of things like VTO, but there’s a degree to which you don’t wanna go too far ahead of yourself because we are doing things in the moment. So you can look ahead of time and say, well, maybe it’s four o’clock that we should train. But the reality of it is if it’s two o’clock, by the time we get to four o’clock, the things may have changed. So, really, in most use cases to answer that question, ultimately, you’re best off pointing it at real time data to deliver the activities or not deliver them in real time. I’d say VTO is the exception to that, where you’re looking ahead at the health because you’re trying to make a decision around whether or not you wanna recruit more over time or release more people. And and just to be clear too, one of the things that’s really nice about the system is for things like training, you can set it up. It it can and will do it automatically. For things like VTO, you can set it up to either do it automatically and offer that to the agent, or it will check with a particular person. In our case, it was a WFM specialist. And the system would recommend, hey. Do you want me to offer VTO? And it essentially had to get permission. So that was a really nice component of the system, that so, yeah, it’s you don’t get necessarily locked into it doing something without approval if you don’t want it. Makes sense. Would it be a disservice to still preschedule things that you’re aware of? As in no. There are use cases. It it absolutely makes sense to preschedule certain things. But the reality of us delivering things in real time is the automation is always on, always assessing the queue health, and WFM teams We’re a small, nimble group of people who always has more work than what we need than what we know to do with. And at the end, that automation can sit there and grab pockets of availability that are too small for us to be able to see and react too quickly enough. So at the end of the day, it’s not a disservice, necessarily, but can you do it better than the automation? And the answer is generally no. The automation is gonna be able to to do things more efficiently, more effectively than individual scans. Boy, we got more questions than we have time. We’re gonna have to have a discussion. I was gonna say, I’m sure within the link, you know, folks wanna reach out. Yeah. Yeah. Perfect. So, actually, I’m I’m gonna hold off on on more questions. They keep coming, and we’re running up against top of time. I do wanna thank everyone, encourage you to click the link. My email’s in there. There’s also if if you do wanna learn more about Intradium and you have deeper questions, that link will take you to our page where you can also do a a contact us, and we’re happy to set up sessions to explain that. If anyone wants to talk more about Erlang o itself, my email’s in that that document. You’re welcome to reach out, and I’m happy to to talk to people at more. Jim, any closing thoughts? I just wanna thank Vicky and and the folks who joined. I think, you know, I will say some of this stuff is really new, concepts, and we’ve been so static for so many years. So it’s great that there’s an organization like this Because the only way we can really change some of this is to to have these type of meetings and thing places like SWPP. So go out and talk to your peers, talk to your leaders, and make sure they’re aware of of the things that are being discussed in WFM. Dave? Yeah. Just just reiterate. Thanks everybody for for your time and investing the time in this. I think it was two years ago in Chicago talking with Jim and Ted. We need to change how we’ve been doing work. Of course, we’ve been doing this forever. And the contact center has evolved so much in that time. So this has kinda been part of that journey, thinking a little bit differently. And and what’s cool about it is, and Jim mentioned, it’s it’s not the three of us. It’s it’s everybody. And so I I actually saw in the chat, which was cool. Some people were answering some of the questions on behalf of us, which I appreciate. But, yeah, we are a community. We all have very similar challenges, and things are continuing to evolve and will well, continuing to evolve as AI comes in more and more prevalent in the call center. So it’s just really kind of what I really like about these sessions and like about the community that we’re in is we kinda get together and try to solve stuff together. So, just a huge share and thank you to everybody for for taking. Thank you. Yep. Thanks, Dave. Thanks, Jim, and thank you, Vicky, as well for having us. Happy to share this and happy to follow-up with people. Alright. Ted, in that last link that you sent, Erlang, is misspelled. So I don’t know if that’s Oh. The wrong link. Actually, no. It it’s the right it’s the right link, but it’s, it’s, it’s misspelled on our website. But use the link. Oh my goodness. The link will still work even though it’s misspelled there. Agner Erlang would be appalled. Yes. I apologize, Agner. Alphabet problem again. I like that. Ugly thing. Good at math, not spelling. So Alright. Thank you for choosing. Well, thank you guys so much for, all of this great information. I sure appreciate you being here with us today. And, we will be sending out a link to the recording and a link and a also a PDF of the slides to everyone who registered. And feel free to share that with whoever in your organization may not have been able to attend or any of your other friends in WFM. Thanks so much again to Ted and to Dave and to Jim for all their great information today. And for now, we’ll go ahead and close this web seminar, and hope you guys have a great day.