Welcome to ACP

Speaker 10 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. We are also joined this week by Vikas Prasad, senior solutions architect at AWS.

Tom Morgan 0:36

It's time for another ACP.

Meet the Guests

Tom Morgan 0:38

We are all back together. Alex is here. Missed you last episode, Alex. Good to have you back with us

Alex Baker 0:43

Yeah. Hi, Tom. Yeah, I know. It felt weird not being around for an episode. Yeah we had a rare occasion where I was going to see real people on a customer site, which is always nice to go and do. But yeah, sorry I missed you.

Tom Morgan 0:56

No, not at all. And and we're extra lucky this week 'cause we're also joined by Vikas Prasad, who is a senior solutions architect at AWS. Vikas, great to have you with us.

Vikas Prasad 1:05

Hello.

Alex Baker 1:06

to this

Vikas Prasad 1:07

Thanks Tom and Alex. Nice to be with you guys today

FCS Explained Fast

Tom Morgan 1:10

We're gonna be focusing today on Vikas's area of expertise which is FCS. Now, we've talked about FCS a few times before, but not for a little while, and along with quite a lot of other things in Amazon Connect at the moment it is moving very fast. So FCS, forecasting, capacity planning, and scheduling if you didn't know. Vikas, if you were describing FCS to somebody in a lift or not familiar with it, how would you pitch it?

Vikas Prasad 1:37

Yeah, sure. So the way to think about this is how do you know your demand in your contact center, and how do you map it to the supply? That's how would be the way to think about it. And, for contact center agents are one of the biggest cost component, and you want to make sure that you don't overstaff, at the same time you're not understaffed. Both are not the good areas to be.

Tom Morgan 2:01

Yeah, absolutely. And so FCS can help with some of that understanding how many people you need, when you need them, where you need them and helping you keep on top of that

Vikas Prasad 2:10

Yeah, that as well, and it has also to do with things of agent and supervisor quality of life to say, right? We want to give options to agents and supervisors when the schedule allows. Let them pick their schedules let them decide when they do want to work, have flexibility, and have things where it's easier for them to apply time offs and make it less of a cognitive load

Alex Baker 2:35

definitely.

Shift Bidding Launch

Alex Baker 2:36

I was going to ask you actually, you mentioned something there it made me think of. I think a recent feature release has been shift bidding. Is that accurate to say, Vikas?

Vikas Prasad 2:45

That went live yesterday, yeah? You're on top of it, Alex

Alex Baker 2:48

And, keep my eye on all the, the release notes and everything. Can you give us a, a quick overview of what shift bidding is? And 'cause I think it's a feature that I've heard mentioned quite a few times in the past and it sounds like it's quite a high-profile one that people will have been waiting for. So yeah, if you could just tell us a bit more about it.

Vikas Prasad 3:05

Yeah, absolutely. So shift bidding definitely was one of the biggest blockers for large customers. And it's called shift bidding predominantly in US and some other geographies. But many geographies European countries I've seen bidding is not a term that's allowed it's typically referred to as preference-based scheduling. So customers might be talking, "Hey, we need preference-based scheduling." So that's actually shift bidding which got launched yesterday. So it's very exciting.

How Shift Bidding Works

Vikas Prasad 3:35

So if you look at the product until now, before this feature launch, agents didn't have a voice. They don't get to rank the schedules. So shifts. They would get our AI generate shifts for them based on certain rules. We did add rules that, okay, lunch after two hours, break after three hours work. But agents did not had a choice to pick off the available... Suppose you have ten shift profiles available, they don't get to rank them. Now you can have upload your agent's rank based on their seniority. If somebody is very senior, he might have a rank one, meaning he gets to priority. If two people are rank ten agent select the same shift as a rank one person, the rank one is going to get high preference. So you can upload that. There are two options there. One is like you explicitly upload the ranks for your agents, or you can have random ranking. So it's just we have seen like how customers want to do between different groups. But the biggest change here is like the staffing group which is like a, a tab or a feature inside FCS where you define a team. So staffing group is a team for which you generate a schedule, so and then in team has a team lead. Prior to this launch this staffing group do a staffing profile. Staffing profile, think of it like a- schedule-- a daily schedule. So you nine to five could be like a shift profile where you define, okay, shift starts at nine, ends at five. So earlier that was one-to-one mapping. So every team to say was assigned to one shift profile. It could be seven AM to seven PM or nine AM to five PM, whatever. But now with this launch, now this staffing group is one-to-many mapping. So again, if you go to the UI, you can say, "Hey, this team has this ten number of profiles." So now there's choice. Earlier there was just one-to-one mapping and there was a way to work around it. Like it would take more work on the people who are building the system, but now it's out of the box, it's one to many. You can have one group assigned to multiple profiles and that agents get to choose. And you can start a shift bid, you can set that, hey, this shift bid is open from Monday to Friday. Agents go in, pick their, rank them, and when everybody has done their ranking, you regenerate the shifts

Alex Baker 6:06

Sounds, sounds like there's probably quite a lot of thought that had to go into that, especially around things like the shift profile. So I guess as an agent then I could-- let's say I, I work the night shift typically, I could put my hand up and say, "I'll do a, a day shift to, to cover some some demand next week," that kind of thing. Is that what it supports?

Vikas Prasad 6:26

Yeah, exactly. For that specifically for that night shift scenario one alternative earlier was like just to make sure that it's not always Alex who takes the night shift. Just an example, you could have set up shift rotation. This is also one feature in FCS which some customers use. So they rotate it. So they might rotate back after every two week between Alex and Vikas, they rotate the night shift. So it's not always one person going night shift. That was doable but you don't get to pick the different options to say. So now the shift bidding gives you more options upfront even before you have schedules generated, and you have the same guardrails still in place. So it's only generated. It never gets auto-published. As a supervisor, you get to review it, make any edits you want to. And if some people are like out of office or whatever reason they can't, or maybe a new joinee you don't want to participate in a shift bidding, you can easily say that these three people, I don't want to participate in shift bidding. Let the system auto-assign a shift to them. So that's available to you as well

Tom Morgan 7:31

That's cool. I see that being really useful for... There's a lot of small companies, maybe mid-sized companies as well, that have a dedicated contact center, but then also sometimes need to pull in people that don't normally work in the contact center just to cover busy lunchtimes or Christmas and things like that, where actually having that ability to be like, "I'm just gonna make a schedule that is just a couple of hours a week, and it's, you're not gonna get paid any more because you're here anyway. It's just don't do your actual work. Come and work on the contact center for a few hours because these are the hours where we know we need people," and put it out to the company in a kind of shift bidding scenario. I think that's really interesting

Vikas Prasad 8:05

Yeah, definitely. That's a great use case, Tom. Definitely. On those scenarios, yes, it's definitely... I can easily see a lot of my customers who have reached out to me in past, and then they were hesitant to go with FCS just for this one big feature ship building we missed earlier

Tom Morgan 8:20

Yeah, definitely. It's, it does come up as a feature people really like because it also you're removing a lot of the admin overhead of managing people swapping, you know, and having to have a manager do the work, update the spreadsheet and all those things. If it can just happen without needing that manager approval, so much easier. So yeah

Vikas Prasad 8:38

Definitely. Earlier, like before this, the only way you could have done it is sit down with your agent, talk to them, "Hey, tell me your ship preference. Let me see if I can manually go enter it somehow," and then, yeah. And it's not doable if you have hundred and thousands of agents, right? That's

Adoption and Rapid Growth

Vikas Prasad 8:55

a lot to do

Alex Baker 8:55

Have you actually found that i-i- is it a kind of, has it released that immediate blocker such that some of your customers are straightaway turning around and saying we, we-- that was the thing that was stopping us. Let's go ahead. Let's let's try. Let's pilot FCS maybe"?

Vikas Prasad 9:09

Yeah. That's what I'm hearing. I'm already starting to see a lot of customers and I, I've been on this FCS journey ever since this feature got launched on Amazon Connect, so it's been three plus years. And it's one of the fastest growing feature within Amazon Connect. So it's been more than 120, I think, at this point, features since last three years. So definitely if someone listening to this podcast, if you had seen this like two years back or even one year back, right? So definitely a lot has changed and it's keeps increasing and like adding on new features like anything else in Amazon Connect at a very fast pace.

Tom Morgan 9:47

Today it's part of Amazon Connect Customer, right? And it's not a standalone product but you can see as it grows and we saw it take on tasks a- and emails and tasks particularly are interesting, right? Because now you're covering work that's not really done inside Connect Customer, but you're using the scheduling. So it's you can see tasks do get done inside Connect, but it's like those back office tasks. Do you think it'll always be part of Connect Customer or do you think it could stand by itself?

Pricing Basic vs Customer

Vikas Prasad 10:14

A small correction here. So it is included in both basic and customer. So you can use FCS both. You don't need to be on customer. The only thing that changes is the pricing. With the basic, you need to pay per agent per month. That's the model that has always been there since the launch, so we have not retired that model. With customer, it's a no-brainer because when you are on customer, FCS is, you don't need to pay for it. It's free, it's included, it's part of that package. And a lot of customers are like I'm seeing like when they are at this on the fence making decisions, should I go to, from basic to customer? So FCS is one which is adding a big value add to make that financial decisions because you might be really using some other vendor, WFM space, workforce management space, correct, and paying them maintaining that vendor relationship building that IT pipeline to stream historical real-time data from Amazon Connect. So you need a IT team who maintains and builds that IT pipeline and all. So those all heavy lifting is taken away if you are on FCS because you don't need to build any IT pipeline. Amazon Connect knows that data. If your calls are landing on Amazon Connect, it knows that data. But yeah, that, that's the distinction today. So it's mainly the pricing. If you are on basic it's per agent per month, and if you are on customer, it's all included with the pricing

Alex Baker 11:43

that was really interesting. So when the, there was the sort of pricing model split out the move to the, I think it was called unlimited AI for a bit, but the what's now Amazon Connect customer we had one of our customers that was already using FCS, so they were paying the per agent per month fee. And then it was, it for them, it became a no-brainer to go to the Connect customer pricing 'cause they were already paying for the the FCS, but they got so much more on top with the Connect customer pricing. So it was a re- a really interesting use case for them just to say, "Yeah, it's a total box tick. We're definitely gonna go with Connect customer 'cause we get so much thrown in plus everything else that we're already paying for with FCS."

Vikas Prasad 12:25

Yeah it does help significantly to build this business case when you need to send that data to your CFO or lead whoever, right? That okay, so if I use this, like how the pricing is going to change, so it does make sense. Yeah

Tom Morgan 12:39

Yeah. Okay, that makes sense.

AI Impact on Forecasting

Tom Morgan 12:41

One of the biggest disruptors at the moment is obviously agentic capabilities and Connect Customer brings, a whole load of those and we see that in the contact center at the moment, and it is changing how, how people manage calls and how calls get serviced. I'm just trying to think through with FCS if, if you've got an AI agent that you're working on and you're, like, you're managing and you're improving over time, right? And it handles, it contains 40% of calls next month. It contains 30% calls this month, and next month it's gonna contain 40% of calls. And how does the forecast keep up with that? Like, how does it know? Are you thinking about that in terms of how AI is changing FCS and how it changes that mix of where you need human capacity? Just interested on your thoughts on that.

Vikas Prasad 13:24

Yeah. I can see three places within FCS where AI is playing a very key role. So first of all definitely the AI model that's built in within FCS is trained on Amazon Connect Retail and supply chain, long history of the way we have handled contact centers at amazon.com. It's, it has been always a talking point when talking with some customers who were initially used to picking their own AI models to do things. But the key differentiator here is you don't have to do that. You can focus more on your business and have this AI engine, which AWS science team is constantly working on im-improving things and the accuracy. But we don't want to work in the silos and just make an assumption that this is like generally the best output as possible. So all our customers who have onboarded have given us a lot of great feedbacks when they don't get certain outputs certain way, and it has been constantly evolving. One of the features which very recently got alive and which was a feedback from one of the UK-based banks was our model earlier had this average handling time, AHT. And if you look at the the spreadsheet which you use to upload the data itself used to have one AHT for the day Which was giving good outputs at that point when feature got launched, but it was not the best output because there's no way your AHT is going to be same for the entire day. For this bank, typically what we heard is when people are early in the morning just heading to their office, they want to keep the call short. They are in a rush. But so the AHT is short, shorter. But towards the end of the day when they're done with the day, maybe just relaxing at home, reached their home, they might be o-okay to talk a bit longer just to understand things better. So the AHT might be higher towards the end of the day. So now this variable AHT has been taken into account by our science team, which is available to all customers now, and you don't need to do anything. It's part of the model, so it's inherently available to everyone. So that's one area of the AI FC. One we have one AI thing that is coming very soon, and it's in beta any customers might be okay to talk to their AWS team and we get onboarded to that beta. So the beta feature is more related to pretty much you might have seen Manager Assist within AWS console. But this one is very much focused on FCS and it takes away a lot of cognitive load. So FCS is a very niche area when we talk to people, and it does require significant cognitive load to say. You might have a lot of data but you still need to parse and understand that data. So think of this AI teammate who will be with you once this goes live where it will be sitting with you and you can ask questions in natural language. Like I can ask, "Hey, for this group can you tell me how many of my agents are at lunch between one to two PM?" So you can figure this data on yourself as well, but it does require navigating to different tables, different dashboards, and figuring out things and... that might be okay for someone who has been long enough, but still spending more time. But it makes this data now available to people who are even new and not very long in WFM area to talk naturally, understand. And not only can it will be able to understand and answer your questions about what's happening But it can also take actions on your half- behalf, right? So those are the things that's coming up.

Intraday Forecasting

Vikas Prasad 17:06

And the one more AI feature I think worth talking is the intraday, which is a very exciting feature which has been live for a while, more than a year, I believe. And intraday is very interesting. So let's think we are in September, and let's see take a scenario like Tom, you have a vacation planned for 15th of October. You can look at the-- All right, that's great.

Tom Morgan 17:28

That's weird. Yeah.

Vikas Prasad 17:30

Yeah, I love vacations,

Alex Baker 17:31

really good this forecasting capacity planning, isn't it? It knows already that Tom's got his vacation

Vikas Prasad 17:36

So let's look at that scenario. So you today went to weather.com or whatever, right? You figured out, okay on fifteenth October it's a bright sunny day. Let me plan out that only I could start at nine AM, go whatever place you want to go and on. And that's how the forecasting tool works. You normally plan one month in advance what you want to do next month. That's very standard. So typically, like mid of September I will plan for October. That's very common I've seen across customers. So yes, that's great. And then the whatever the forecast, everything is, everything gets done. Like it's just like a weather thing. So on fifteenth October morning, things might have changed. It might not be a really sunny day. And what tells you about it is a day before is your intraday. So intraday is specifically AI models which are focused on last twenty-four hours. So it gives output for what's coming in the next twenty-four hours, and it refreshes every fifty minutes. So it's highly optimized on that front. And so it-- think of it like this, the contact centers have been always all hands on deck. Hey, there is a huge call volume in this queue. What should we do? Let's meet up in this meeting room, and then let's move people around." But what this really does is on that day, on fifteenth October at seven in the morning when the supervisors are there, it will tell you, "Hey, be prepared. At two PM there is a huge spike." So that team leads at that seven AM will make a decision, "Let's put that team meeting we had planned at two PM, because there is going to be a huge spike at that time." So it's more helping you act proactively, which is a huge trend shift. I am like I I don't like to do a lot of comparison with the other market players, but yes, it i- it is a very true differentiator there

Tom Morgan 19:20

I think you said something really interesting earlier as well, whi-which was about the sort of value... and we often think about, Connect Customer as the, the power of like a, a cloud-based, a SaaS, so it's constantly evolving. You're getting these new features, and that's great. What I didn't really appreciate until you, used it as an example there was it's not just that you're getting the features, there's people working all the time on making the algorithms you already use today better. So you're, you-- It's the same feature. The feature is the forecasting, right? But actually the, how it does the forecasting is improving over time because there's smart people working on it all the time. I didn't really think that through until you said it I think that's another really valuable sort of piece of evergreen software like this versus something that you buy once and sits on a server somewhere and, doesn't get updated over time.

Vikas Prasad 20:05

Yeah, ex-

Tom Morgan 20:06

because also, yeah, a-as also, like this, the, coming back to this, the AI thing and agentic AI is as that happens and more and more calls get contained or different types of calls get contained and depending on the business and different flows and different scenarios, it will be able to react to that much better because you'll be feeding that in because you'll be seeing it because agentic AI flows through Connect Customer as well. So it's, that completely makes sense.

Vikas Prasad 20:30

Yeah, definitely. You definitely do get that advantage of speed of innovation. So whatever the science team and the AI team at AWS is already improving, innovating on those S- FCS models and everything. So you get those benefit out of the box. You don't have to do anything. That's something very similar to we just talked about on the variable HD thing. Yep.

Workforce Planner Role

Alex Baker 20:54

Do you see AI really changing the role of the, the workforce planner? 'Cause that always used to strike me as it is quite a complex role that you've gotta, have a handle of what's going on operationally in the contact center but have a good sort of head for figures and know your forecasts and be able to put a, a plan and a schedule together around that. It sounds like AI is only gonna help there, but it, is there still a role for the workforce planner going forward? Are they gonna be just freed up to do more valuable work? What's your thoughts?

Vikas Prasad 21:23

I think this is one area where I believe that the planners still have a big role to play. And I can talk about one scenario just to set the perspective. Like the forecast itself the AI model can do a great job. It will generate your forecast. But I have typically seen customers do some adjustments based on their business knowledge, which the AI model not, might not really have. Something suppose mid of next month, I have a new shoe lineup being launched, and I do anticipate a ten percent spike in volume. So those, that, those are very business knowledge that planners and business users know, and they need to be there to fine-tune that and play that role. So the, I still see the planners are going to have a significant role in this part of the future. So there's many things that you need to take care of. A lot of businesses that I talk to have a lot of labor laws and other laws and everything, right? So you can't have the AI run on its own. You have to look at everything. So everything on the exercise is gated in the sense that it gets generated. It never gets published on its own. You have to be there. You review it, you make some changes, and then you do it, I think. So yeah.

Alex Baker 22:42

Yeah, makes sense. Thank you

Tom Morgan 22:44

There's always gonna be inputs that outside of the, the domain of Connect Customer that influence it. Like for your example, like the know- the knowledge of like the shoe launch or whatever else it is that you can feed in to help it understand that more.

Vikas Prasad 22:57

Yeah. The way I look at it, it's going to provide you two things. One is definitely speed at things having a AI teammate working along with you. And it's also a lot geared towards the quality of life.

New Features and Data Lake

Vikas Prasad 23:11

One feature that I very recently launched is called adherence thresholds to say. And the granularity, it's a four levels granularity without going too much into technical details, but it gives you a granularity at different levels. But think of it like this prior to that you went to lunch 10 minutes late because your customer call was still on the call, which was the right thing to do. You should not hang up and go to your lunch just because the schedule says it's lunch. But you will be marked red as out of adherence, and your supervisor will be alerted, "Hey, this guy is out of adherence. He did not follow his schedule." So that was not the right thing to do. So that's where adherence thresholds have now live, which allows you to say, "Hey, it's okay to start your schedule activity." You can configure it at activity level, and in this case, you could have set it up like it's okay to start lunch 10 minutes late. You would be marked out of adherence. And then it's good for both agent as well as a supervisor. Supervisor not getting that false alerts, something that he should not look at and spend his time, as well as from the agent side who's doing the right thing there not to get flagged as out of adherence, right? So th- those are the two areas I'd see this being helping out the contact center.

Alex Baker 24:33

Yeah, it sounds like that's really important to take into account the example you gave. It's very rare that a call finishes exactly at your 12:30 lunch break or whatever it might be. Sounds like you need that extra flexibility.

Tom Morgan 24:45

I was thinking about different workloads in, the contact center and how the, the kind of workflows planner needs to be aware of these different things. Then I thought actually, there's the Amazon QuickSight as well," right? Actually, Amazon QuickSight probably does know all these things, and so then do you ever see FCS coming out of Connect completely and being more something that runs across your whole business with humans and AI? Because that challenge of, it's not so much, at that point, it's not so much shifts and management, it's more about, making sure people take their breaks and tracking more of an HR level function, than just shifts. But plenty of companies do run in shifts, and they don't run in contact centers, right? You just run elsewhere in shifts. Yeah, I don't know. I it's interesting to see how these things go

Vikas Prasad 25:27

Yeah very interesting, Tom. I haven't seen anything going that direction yet. There might be use cases that might come up with all this back office thing that's now available to task and everything. So now this year we have two more channels added to FCS, which is task and email. So now things you will be able to do not only the like the personas who take calls and chats there might be back office people and email handling personas who might be able. So it might very well go there. And it's worth calling out one of the differentiators here as well for FCS where I think it's one of the only products I've seen in my twenty-plus years of contact center journey where I see the entire data is exposed to you, the raw data in data lake. So everything is defined, available in our admin docs. The table structure, each field is defined, and each data is allowed to be sent directly into Amazon kind of data lake. And that's where I see like the things that you're talking about Tom, like it can be sent to QuickSight or any other BI tool that you want, and then you can like Snowflake some customers use and others, right? But you can use native AWS tools as well, where you can send this data and do whatever AI thing and analysis on those fronts as well.

Tom Morgan 26:51

Yes. Yeah that's quite right. That's, yeah. The openness of the data for FCS but also in Connect generally is really good and opens up lots of, opportunities for partners as well, let's be honest. It's good, like closes lots of requirements gaps for customers in doing that as well, so it's good for everyone.

Wrap Up and Subscribe

Vikas Prasad 27:08

Yep

Tom Morgan 27:08

We could talk about this all day, but it is time to bring this episode to an end. Thank you, Vikas. Thank you, Alex, and thank you for listening. Be sure to subscribe in your favorite podcast player. That way you won't miss the next episode. Whilst you're there, we'd love it if you would rate and review us. 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