Tom Morgan: 0:02

Welcome to ACP, the Amazon connect podcast. This is the show that focuses on Amazon connect and related technologies. I'm your host, Tom Morgan, and I'm joined as usual by my co host AWS solution architect and contact center consultant, Alex Baker. Find out more about Cloud Interact by visiting us at cloudinteract. io. It's time for another ACP. I'm here, Alex is here. Hello, Alex.

Alex Baker: 0:41

Hello.

Tom Morgan: 0:42

Hello. Hello. And this week's going to be interesting cause we've got an opportunity to kind of do a deep dive into a very sort of specific thing. Which is what we always wanted to do. With some of these podcasts episodes, sometimes it's good to look across and see what everyone else is doing and focusing on different sort of broad areas, but sometimes it's good to go deep as well. And so this week we thought we would go into some real detail about some of the billing around Amazon connect. And specifically some of the things that we've seen recently around consumption versus subscription based billing. So that's kind of the overall topic. But before we really dive into that, let's start off sort of right at the beginning with what we mean by all of that and in general terms how we, how we pay for Amazon connect. So if we sort of rewind maybe, I don't know, let's say about a year. If I'd asked you about a year ago, how do you pay for Amazon connect? What would you say?

Alex Baker: 1:36

Yeah, maybe even slightly more than that now, but yeah, so originally when connect was released, it was with that, that pure consumption billing model. And by that we mean you pay on a sort of per minute basis for what you use which was, I think, a real. Disruptor when it first came out contact center as a service and contact center was generally more sort of around a licensing model. So you'd hear per user per month figures for example, and when the consumption based billing came out, yeah, it was a bit of a breath of fresh air. If you have customers or if your own contact center has sort of spike your peaky. Loads in terms of call volumes, actually consumption billing could be quite a cool thing for you because if you have, you know, a real sort of trough of low usage in certain months, whereas with the previous model, you'd be paying your per user per month fees throughout with the consumption based billing. If you're not using it, you're not paying as much for it, which is about the size of it.

Tom Morgan: 2:46

Yeah, and per user per month makes sense in that sort of old world, the infrastructure led world where you've got, you know, servers racked in, cases you're using upload, you're using CPU memory, those types of things. Obviously, Amazon Connect is very much cloud first. And so their model. is, is kind of more aligned to that cloud first model. And I think that's why it was such a disruptor, right? Because not only was the tech quite different, but because of that, it gave them the freedom to be quite different on the pricing as well.

Alex Baker: 3:16

Yeah. Agreed.

Tom Morgan: 3:17

and it was, like you say, it was that kind of real disruptor in the market compared to how things were. And yeah, so one of the things that we always say as well, is that it's very It's having that sort of consumptive bidding model. And to be clear, it's not a one single number that is per minute, depending on what you're doing. So there's all sorts of things that if you do this, if you do that, like everything's got a, a little bit of a price attached to it. but it's all per action. It's all per call or it's per chat or it's per activity. So it's consumptive in that way. And that if you do nothing, if you just don't open your shop on a Wednesday, you won't pay any money on Wednesday. More or less, and there's probably, now that I think about it, I dunno. Are there per monthly charges for the phone lines? Is that a thing?

Alex Baker: 4:02

Yeah, so that, yeah, you're right. There are, you know, it's, it's not exactly that if you close the shop on Wednesday, you'll get billed absolutely nothing for the Wednesday. There are a few, a few little charges that are, I guess, sort of baseline costs. You, you mentioned. The phone numbers, effectively a phone number rental charge. It's on a per number basis. It's pretty low, but if you've got hundreds of numbers, yeah, you know, that baseline cost is always going to be there. There's some other stuff as well, maybe not connect itself, but around the, the supporting services. So things like S3, where you're storing your, your call recordings, obviously the more you're storing, the more capacity you're taking up and therefore. The more that that kind of baseline cost for the storage increases, what you tend to see, though, is that things like s3 actually form a relatively low proportion of the total costs for connect deployment because that they're effectively pretty cheap for a massive amount of storage.

Tom Morgan: 5:02

And And the other sort of advantage of doing it this way with, with this kind of consumptive billing, it made it very easy to do proof of concepts, right? And we've talked about this before. You can show up with your credit card, you can get going quite quickly. But it's not just how easy it is to onboard. It's how relatively cheap it is to do like a proof of concept with, you know, half a dozen calls through a couple of different people, because you're not paying that upfront. You know, onboarding cost essentially, or like paying for a month's worth of, of licensing.

Alex Baker: 5:34

Yeah, definitely. That, that was, I think that was really revolutionary to me when I first came across connect. Certainly it was, I've mentioned this before, the barriers to entry for us or legacy contact center solution were much, much higher. And yeah, you, you talk about the POC or ease of doing a POC, there was always that, that thought that, wow, we've got to invest tens or hundreds of thousands up front in things like licensing things like kit in order to just start that POC. And maybe you have the commitment around the licensing as well. That was another biggie. So it wasn't typically easy to sort of get a set of trial licenses. Commitment periods that you had to tie yourself into. Whereas with connect, like you say, put your credit card details in, spin up a POC and whatever you want to do. If, if it doesn't work out for whatever reason, then you can just turn it off and you can stop that billing. There's, there's no monthly or yearly commitments or anything like that.

Tom Morgan: 6:33

Yep, it's very easy come, easy go. So I said, talk to me about what it was like, you know, a year ago. So over the last year or so, we've seen some new features come to Amazon Connect, some exciting features. We've talked about all of them, I think actually on previous episodes. But that has changed the billing model slightly. Hasn't it?

Alex Baker: 6:52

Yes, and I, I guess some of them did, it came as a bit of a shock, I think, to some people when it was you know, cause people have probably got used to, there was that initial change where it was like, wow, Amazon connect is consumption based that that's kind of great for our use case. And then we had a few years of that. And then some of these new features came in and actually it was a bit of a shock to the system to say, okay, so it's mostly. Consumptive, but now we've got a few things that are that sort of per agent per month One thing that it's worth mentioning is that it's still sort of semi consumption based if that's, if that's possible in that if you don't, if you're not using the feature. For a particular month, you don't get billed for it. It's not that sort of license based model where you say, I want to turn on quality monitoring evaluation forms. For example, I've got 100 agents. And then from that point into into the future, you're paying the. per agent per month fee all the time. If you don't use it one month, you don't get billed for it. So we should point that

Tom Morgan: 8:01

more of a, it's not a commitment model, right? It is, but it is rather than being consumptive on activity. It's sort of. Build, it's like a monthly charge that is triggered by activity almost

Alex Baker: 8:13

Yeah. And the, the other, the other thing that's worth pointing out, we've noticed this with a few of our, our customers is the way it appears in the billing, it sometimes makes for certainly the first time you see it a bit of a breathtaking first day of the month, if you're looking at the pricing page. So things like, you know, if you've got. As one of our customers does around a thousand agents using the forecasting capacity, planning and scheduling, that's 27 per scheduled agent per month. And that kind of hits your AWS billing on the first day of the month. So you kind of look on the 1st of October and think, wow, that's a big, a big charge for the first day of the month.

Tom Morgan: 8:52

Well, this is like to the whole point of how different this is to almost, not almost, but lots of things in AWS that are consumptive, because even the cost Explorer can't really deal with this first of the month. There's this big number because it throws out all of the forecasts. It throws out all of the estimated spend. Everything looks terrible. Until you're about a week in and it's like, Oh no, that just happened once. It's not going to happen every day. Like that, you know, there, I mean, like the, even AWS is backend cost modeling and cost Explorer stuff. Doesn't cope well with this kind of one off monthly costs. That's just to kind of highlight how different it is from, from quite a lot of the other AWS services.

Alex Baker: 9:34

I guess people, people that are familiar with AWS, but maybe not as familiar with connect, you might liken it to the way the support. AWS support is billed. I think that that happens in a similar way. So you'll get the, the per month cost on a particular day of the month. But yes, it, you're right. It rather skews the figures, doesn't it? So you get kind of an immediate bill on day one for say, you know, 50, 000. It then, it skews that forecasted figure for the rest of the month, really. Really heavily for the first week, like you say, and you're looking in on day one and going, Oh my goodness, my estimated spend is like half a million dollars or something for this month. And it's only only by the start of week two that that's ramped down quite considerably.

Tom Morgan: 10:20

Yeah, exactly, and it's probably always incorrect as well, if it's an average of all the days, because you're never going to have that value. So high again. So your estimated costs are probably always artificially high for the rest of the month

Alex Baker: 10:32

Yeah. Yeah. Agreed. Yeah.

Tom Morgan: 10:34

So let's talk about exactly what we, what we are talking about. So forecasting and scheduling as a feature is. It's something that is charged per agent per month. Evaluation forms is something that is done per evaluated agent per month. And then Amazon queue, which is their AI engine and all those kinds of AI capabilities is, is done per agent per month as well. So those three things are not consumptive in that way. They are sort of subscription based in that when you use them, you'll pay, per month. Now, this is, this is still not named users, is it? It's triggered off who does it. So it's, it's a kind of total and I guess it's a total, is it? Yeah. So I've got so many questions it's total discrete users, right? I think, I think I'm right in saying that it's not like overall usage, like you know, load based or anything like that. It's total discrete users, but you don't have to name them as just as soon as they start using it. That's when it triggers the bill. And it's, I think it's, there's not much grace period either. It's the first time they use it that triggers the charge for the month. Silence.

Alex Baker: 11:49

agent per month. So like you say, If you run one evaluation on all of your agents, that's going to be your 12 that you will incur per agent per month. So it does, you can see where the potential benefits lie in, and we've talked about in the episode about QM and evaluation forms. Not just doing it kind of the old way and doing manual evaluations, but taking advantage of some of the. The potential for automation in the platform, because it then starts to make more sense if you're kind of automating the evaluation of more like 80 or 90 percent of your, your calls, it sort of justifies that 12 per user per month, a bit better.

Tom Morgan: 12:33

Yes, I think that that's true for a few of these as well, because taken, if you just look at those headline numbers, you know, the forecasting and scheduling 27 per agent per month. I mean, that's an expensive price if you compare it to a standalone. agent scheduling tool, right? Like a staff scheduling tool. You know, 27 per per agent per month is fairly pricey. Same with evaluations. If you were just talking about kind of manual evaluations but these are more integrated into the platform, aren't they? And so you get that benefit of. You know, it, it being built into, so for instance, you know, with the scheduling and stuff, the forecasting, it can see the current activity. So it can do a good job of forecasting without too much messing around. It kind of works out of the box, same as scheduling it, you know, it will, it will work with the existing agent experience. So, and, and the stuff you're saying about evaluations is true as well. And you've already got the integration with, where you've got Amazon already got the integration. Into the agent desktop and stuff like that. You've got those benefits, I suppose that may justify that price of a kind of all in one single pane of glass for, for, for agents. I mean, there's, yeah, I mean, maybe it could be better integrated. I don't know, but I mean, it's, it is at least start, you know, it is at least integrated more so than a standalone standalone tools would be.

Alex Baker: 13:52

Yeah, agreed. And that sort of, yeah, because, because it's their natively, you haven't got that overhead of trying to integrate a third party tool, which there's a lot to be said for that on the forecasting and scheduling, arguably. It's probably cheaper than a lot of the, this sort of third party vendor solutions out there. Also, though it's worth pointing out is probably a bit more basic in some areas. Amazon as always are working on improving that. And there's a pretty, pretty big roadmap of items of improvements that they're making to it But yes, probably in places it's, it's a little bit more basic. I, I guess a lot of it's down to, and we've talked about this around the Q functionality, the, the AI generated agent assist stuff. Yes, it's, it seemed, it's a, the headline price is quite expensive. It's how you can build a business case for it. And can you, can you make it justify itself by getting more than your, your 40 worth of, benefit from it. And I think probably that's still, still in its infancy and that there's a lot of, a lot of companies, a lot of partners like us that are experimenting with it, with a few, a few customers and trying to figure out where the real proof of value is for the, for some of these features.

Tom Morgan: 15:09

Yeah, definitely. And I, yeah, I do want to talk about that with you actually, cause I know that's a lot of, you know, what you're, what you're doing working with, you know, customers is, is helping them sort of prove that value before we do that. How do you, if you're a, you know, you own a contact center, you manage a contact center. that you know this, you know, if your agents start using forecasting or evaluation forms or Amazon queue, you are going to get charged this flat rate, not consumptive. So you want to make sure you know what's going on and you're in control of that. What's the best way to stay in control of. Knowing when users or like preventing users, if you don't want them to do those things or just knowing when they do it, is it, is it security groups or is there, is there another way of doing it here? That is like turn on, turn off or caps or I don't know.

Alex Baker: 15:59

Yeah, so yes security groups is, is a great option. So for example, if you wanted to do a smaller proof of concept on one of these things, you can just turn them on for groups of users. So Q, for example, you could have a specific security group for agents, which enables the Q functionality and just assign that security group to your 10 POC users or whatever it might be. Okay. In terms of keeping an eye on what the. What you expect the cost to be. And I suppose one thing about Amazon connect and AWS billing is it can be a bit complicated and it also, because it's consumptive,

Tom Morgan: 16:39

Okay.

Alex Baker: 16:42

of budgetary figures, how much we're going to spend over the next six months. There is a, a fairly detailed Amazon connect cost calculator spreadsheet where you can plug some numbers in, in terms of your, your estimated usage and get some. Some, some figures around costing and what it should be. But yes, it, it always helps to have that, I suppose, have that preview to, to what you, what you might be spending, but then you, you do need to sort of keep a close eye on it in cost Explorer, look out for any of the sort of unexpected stuff. And as you, as you allude to try and lock it down so that everyone doesn't have that expensive feature if you don't want them to,

Tom Morgan: 17:22

Yes. Yeah. And I guess there's, there's probably reporting you can do to work out who is incurring those charges. Who is it? Cause you, I can imagine there's, there's several layers in a big contact center as well. There's sort of a few different layers of autonomy and management there around the people looking at the costs, but then also. You know, the, the shift supervisors choosing the agents that they're scheduling for that month. And, you know, they know there's a busy period. So they're onboarding 20 more people and they just want to roll them into scheduling because it makes their life easier. But that's a significant cost. They've just taken on for that month. Things, things like that, you know, that just need to be worked out. I think.

Alex Baker: 18:01

yeah, and it helps if I guess everyone throughout the organization has some understanding of which of these are cost incurring functions and features or back to the security group thing. So the people that are sort of doing the day to day running of the operation, maybe they don't have the permissions to. So just kind of add in one of these features, if they, if they feel like it is down to sort of an admin staff member, possibly.

Tom Morgan: 18:29

Yeah, definitely. So I think what I'm kind of interested in as well, that's the kind of the what, like the what's what we're talking about, really, it's it's taken us nearly 20 minutes to get to the, what we're talking about, which is good. But the why is interesting, you know, why would Amazon connect do this? I've got some thoughts. Some of it, I was trying to think, how else would you do it? You know, how would you consumptively charge for forecasting and scheduling? I think that's quite hard. I don't know what you think. Would you do it? I don't know what you do it by.

Alex Baker: 19:01

Yeah, I, so I can, I can see some of the notes that you've written down and I think what you've, what you've mentioned does some of these, they're probably making wider use of things like AI, things like LLMs, which are probably massively more expensive to run than some of the previous underlying

Tom Morgan: 19:19

that's definitely true for Amazon queue. You couldn't like just knowing how those LLMs work, right? There's stuff in the background, keeping those LLMs up to date. And you kind of, you need to, it's kind of all or nothing on their side. I think. I don't think they can really be selective about maybe they can about which calls get included for like, I know that's the experience when you're using Amazon queue, but like the LLM that gets built on the backend. So. It might that might be a way of justifying those costs because it they almost need to have it ready. Whether you use it or not, so that it's there in case you do want to use it. And so in that model, consumptive doesn't work very well because you'd have to charge so much. You know, so,

Alex Baker: 19:58

we, we found it with some of the others where I. AI tools, Bedrock, for example it can, it can quite quickly for some of the things underlying it. So I'm thinking the sort of the linking up to knowledge bases and open search in the, in the background, they can quite quickly start to produce big bills and I guess having it rolled into Connect, what am I trying to say that they sort of needed some sort of guardrails around making it available to everyone. But then just sort of hammering it and, and you know, maybe, maybe it's like a sort of barrier to entry thing. You, you, by seeing this cost, you appreciate that you are using something that costs a lot to run. And therefore you, you kind of giving your, your commitment to it.

Tom Morgan: 20:42

weirdly and perversely by cost capping it. You might increase rollout rather than decrease rollout. My initial thought would like this would hamper rollout because it's a big sticker shock. Do I really want to chart, you know, add 40 per every agent every month, but actually maybe if the alternative is like, well, it's consumptive and it could easily be quite a lot more than that. If you want to get into this AI game having it capped actually, maybe that is A nicer experience. It's tricky to argue, if you're Amazon Connect, it's kind of tricky to argue the benefits, the complete benefits of known flat fee. Pricing, right. Having spent so long arguing for consumptive, but in some use cases like AI, maybe it does make more sense. I don't know. I think there's something as well, just because of the market we're in at the moment that this stuff needs to be profitable. And so this is well known upfront pricing that is dependable. So there's that kind of shareholder profit thing. I don't know how real that is, you know versus sort of perceived, but like these business units need to succeed and they need to be shown to be succeeding. You know, we've had a couple of years of tech redundancies and everybody tightening their belts and every, you know, every area of the business needing to be profitable. And this might be some of that as well. Yeah.

Alex Baker: 21:59

and I, it is not, it is not sort of unheard of. Is it? It's you look at maybe copilot or something.

Tom Morgan: 22:05

Exactly the same,

Alex Baker: 22:07

fairly

Tom Morgan: 22:07

exactly the same reasons. Yeah.

Alex Baker: 22:09

fairly chunky sort of costs per agent per month, which I know, I guess maybe some of them, the talking about Microsoft a bit now, but some of the, the billing has always been more toward that, towards that licensing. Based model, but yes, in the, in the sphere of AI, it's definitely not unheard of. Is it? I know, I think it does seeing that headline cost does make you carefully consider how you want to implement it as well. If it was just something, I mean, we take, take contact lens, for example, which is. Fantastic native native functionality. Sometimes I feel that it can be almost too easy to turn it on and have the, don't get me wrong, the great benefits that you get just from turning it on. But there are also some other things that you can do with it. Maybe things like creating rules, things like the sort of topic modeling that you need to give a little bit more thought rather than just turning it on. You need to do other stuff.

Tom Morgan: 23:05

Yeah, and actually, like your copilot example is a good, is really good one because Microsoft traditionally don't do very much consumptive billing. But the reason for that is enterprises really don't like it very much. They would much rather have the sort of flat fee and looking at the things that are now subscription based forecasting, scheduling, evaluation forms. AI, maybe less so, but those first two are very enterprising features right there. So maybe some of that is market driven, you know, that's what the focus groups said they wanted. But also as well, like if you look and compare Amazon connect to the competition, it's still, you know, It's still very competitive. So maybe they're just thinking, well, we're just leaving some money on the table here. We could, you know, we could do better at this and just, you know and, and use these new features as a way to do that is always the problem with these kinds of. Evergreen services, right? Because you, as consumers, we expect more and better exciting features to be added every month for paying the same price. Eventually that doesn't work. So that's where you see these kind of new features come in with their own pricing models. And we've seen it on the Microsoft side as well with, you know, features being added to the E5 but they're separate, you know, things like Teams premium And then copilot, you know, they're separately priced because that model doesn't run forever. You know that I pay the same price as I paid five years ago or marginally different, you know, 2%, 3 percent each year, but I get so many more features all the time. Just doesn't, doesn't stack very well. Yep.

Alex Baker: 24:30

agreed. And, and they're, they're quite. Distinctive sets of functionality as well, aren't they? So forecasting and scheduling isn't for everyone, right? You can see. You can see why some people would just say, right, yeah, not interested equally. Then if you are interested and you want to want to get on board with it, you kind of know where you are in terms of we've got a thousand agents by bundling that in, we know that this is going to be our, our baseline cost for it.

Tom Morgan: 24:56

no, definitely. Any risks to this approach, do you think? I mean, will it make it harder to do POCs or do we just have to be a bit more careful about what gets POC and what doesn't?

Alex Baker: 25:06

Yeah, I think a bit being careful and selective about it. Certainly. We mentioned Q for example, don't just turn it on for everyone and start incrementing, your costs per agent by 40 extra a month. Do that POC and a ring fence group of people and figure out where it adds value.

Tom Morgan: 25:23

Yeah. Okay. Okay. Okay.

Alex Baker: 25:39

one evaluation per agent per month. Look at how you can automate that process and move towards more like 100 percent of your contacts being evaluated in some way. Even if then you've got your human QM team that's doing a lot of the work as well.

Tom Morgan: 25:56

Yeah. And again, I think this is where partners can be helpful because I think there's a risk, connect can be expensive to set up because you know, it's the setting up it's getting everything right. And that's where partners can help. But obviously you've got that upfront cost. But then the operational month to month costs, you know, very low. Now, if you add these things in your month to month costs are going to go up. And I think there is that risk that you, you, you might just want to sort of save on the setup stuff and not do quite so much, but I think actually what you just said is a good example of where partners can really help if you're going to pay. 27 a month for every agent to do forecasting and scheduling. Or if you're going to pay 12 per agent for your evaluation forms, you want to get a hundred percent of value out of that. And so you want to do as much as you can. So to your point, you want to do all the automation because, you know, you're not going to get charged any less for not doing it. And so you might as well. And that's where partners can really help and understand to make sure you use what you've bought to the fullest extent. I think.

Alex Baker: 26:53

Yeah, I agree. Yeah.

Tom Morgan: 26:55

I guess it makes it harder as well. To price up deployments, does it, or is it just more questions, you know, in sort of thinking and planning out a deployment for a customer you have to consider, okay, not everybody is going to use evaluation forms. And so you're asking them questions that maybe they haven't really thought of yet.

Alex Baker: 27:12

Yeah. We mentioned you're right. To some extent it is difficult to plan for what you're going to be spending and especially if somebody is moving from that sort of legacy per user per month billing, it's sometimes difficult to get them to want to adapt to that, that new way of working and saying, actually, we don't really know how much you're going to be spending. It will be roughly this much based on the numbers you've given us, but we can't say it's going to be 1000 times a hundred dollars a month, every single month. Back to your point on the partners though, we have experience of doing this for a lot of customers. We can do our very best to take the inputs that you give us and figure out roughly what your spend is going to be and justify why that's the case.

Tom Morgan: 27:58

it's almost split into two, isn't it? It's like, here's what you're definitely going to pay every month. And then here's what we think you might pay every month, which is the consumption bit, whereas before. Oh, okay. Cool. You know, it would have just been like, here's what we think based on your, you know, your activity and stuff. So

Alex Baker: 28:11

Yeah. Yeah. Agreed.

Tom Morgan: 28:12

it's just, yeah, it's just slightly more complicated. All right. No, that's been, that's been really interesting. We've managed to yammer on about billing costs and billing types for 30 minutes. That's pretty impressive.

Alex Baker: 28:23

Interesting area and one that I think we've alluded to it throughout, but you, you can If you're not careful, you can get a few surprises and those sort of breathtaking moments that, you know, I think we've probably all had it where we've been playing in our own AWS account and we've left something on. And then you come back to it in a couple of days time and think, Oh, wow. I've just sort of, I don't know, I've left on like a Redshift database or something, and it's

Tom Morgan: 28:45

now I need to remortgage the house. Yeah. Yeah.

Alex Baker: 28:47

that's it. And you don't want that in an enterprise. Setting the way I've, I've left, I've left forecasting and scheduling on for a thousand users or something. We're

Tom Morgan: 28:57

Yeah, no, exactly. Yeah,

Alex Baker: 28:58

really careful about that kind of thing and be cognizant of what it is that you're, you're sort of signing up to.

Tom Morgan: 29:04

absolutely. Oh, well, we could talk about this all day, it turns out, but no, we should bring this episode to an end. Thank you very much, Alex, for your insights and your knowledge. And thanks everyone for listening. Be sure to subscribe in your favorite podcast player. That way you won't miss it whilst you're there. We'd love it if you would rate and review us. And as a new podcast, if you have colleagues that you think would benefit from this content, please let them know. To find out more about how cloud interact can help you on your contact center journey. Visit cloudinteract. io. We're wrapping this call up now and we'll connect with you next time.