Improve outreach with cost-effective contact center solutions

great. so, thank you for coming everybody. my name is evo jensen. i'm a solution architect with aws and the nonprofit team. and today we're gonna talk about customer outreach.

we're gonna talk about customer outreach in the context of, of traditional call centers. modern data strategies for uh a modern messaging time together show how to integrate with um with aws services. uh all gonna do that in 20 minutes.

i'm already counting down. and if i have some time left, we'll talk about uh some gen a i improvements uh to contact centers as well because after all, this is reinventing 2023. so it always fits in there, right?

so i'm part of the nonprofit team. um that means i work with nonprofits, uh often donor based uh uh organizations, membership based organizations and they often have a need for call centers to call in as you can imagine.

so, while this talk has a slight focus on nonprofit, because that's where i come from, it's really applicable to everything, right?

so, let's talk a bit on how we got here, right? so if you look at nonprofits, what are some of the top challenges that we see with, with fundraising, it's donor engagement, donor fatigue, donor retention. and so how do we address those challenges? right.

in that same study, uh 2023 fundraising outlook report, uh we're looking at some technology investments there as well. and what we see is uh um organizations do place increased budget on donor management and marketing automation.

so that's really aaa prime way for a technology for cloud technology for aws to come in and um and make a difference.

so um let's start with some of the more traditional ways of doing customer outreach, which is which is uh a call centers, right?

so some of the common questions that we hear are like, how can i innovate quicker? right? i have this static call center. how can i innovate? how can i introduce new features maybe based on machine learning into um into my call center to improve customer satisfaction? and how can i do this? how can i launch these new features without breaking the bank? how can i do that without um uh buying whole new uh uh modules and, and doing this in a cost effective way.

and so we, we commissioned a study where um we had customers who had already migrated from um uh into our contact center solution, amazon connect. and we asked them. so what are some of the challenges you faced previously?

and some of the common themes that we heard was we're looking at at expensive legacy solutions that are often siloed. so in order to actually get this together, you have to go to multiple vendors and stitch these solutions together yourself.

um it, it, it's, it's, it's hard to uh then also experiment with new technologies and make all that stuff elastic by elastic, i mean, easy to scale up and scale down.

um and as, as you know, at amazon aws elasticity is a core concept of how we, how we deal with uh with our computer.

so how do we get to amazon connect? so let's do a quick step back here.

so, amazon.com at large, uh the motto is we strive to be earth's most customer centric company. and so in the retail business, we had a need to build these uh uh to do customer outreach, customer support.

so we didn't, weren't happy with the status quo. we actually build our own software solution. uh now having about over 100,000 agents worldwide serving millions of customers, dozens of languages, right?

so we build a technology stack and then what we did in 2017 is bring this to market as an a us service and that's amazon connect.

so with amazon connect, um we now try to address all those challenges that we faced previously.

so first of all, it's really easy to get started. um it's just another a bs servers in the console, you go to connect in the console, you can provision in brand new call center in literally minutes, you click a button, start project, your telephony gets provision for you.

you can in a graphical way, then build a call flow with some ivr agent transfers all the features that you need. and so this is really built from the ground up as a as a holistic solution that can address all these challenges.

and in great amazon us fashion, it's paid only for what you use. so there's no upfront cost, you just pay for the call minutes that you need to do and it can easily scale up.

so you start with that small um small poc and it easily scales up to uh to, to thousands of agents.

so in that same study that i mentioned of migrate people who have migrated to to connect, uh we got some great results that customers mentioned uh shortened average handle time up to 15%.

um uh this is due amongst others by having this integrated call analytic stuff. so there's less after call work for the agents to do and we'll dive into that a little bit more how we did that.

and another benefit is uh uh uh a significantly reduced admin effort because we built this product to be very user friendly with graphical drawings of, of your contact flows.

so no wonder that we have a lot of customers using uh uh connect from uh all the way from small companies who love the fact that you can start small experiments, don't have to do any upfront cost to large companies who really enjoy that.

you can experiment with new features and have elasticity to scale up and scale down. i told you i was a nonprofit as a so think about a nonprofit, like for instance, red cross who has a large need for call centers in hurricane season and that scales up quite quickly.

you don't even know when these hurricanes are gonna hit, right? so we need to be able to scale up as demand grows. and then when blue skies come around, we have to scale down as well, right?

so integrated solution, right, where previously you might have to go to different vendors to stitch these these components together. um this is a holistic solution.

so let's let's go through a journey of a call. uh a call comes in even before we go to uh the first agent.

um we can plug in some machine learning to do some analysis on the voice already we can do. for instance, the sentiment analysis is the customer happy. is the customer bad? what are that prompting the ivr system?

and the way that i talk to um to an automated response even before you go uh to an agent. and so then by the time that gets transferred to an agent, we can pull in the customer's uh uh system of records, we can uh maybe transfer the agent to transfer the the collar to an agent appropriate with the sentiment of the collar.

and then when the agent sits in there, it can integrate with the system of records to have a full view of the customer during that call.

uh adamski, in a keynote on tuesday, he mentioned uh our great partnership with salesforce. so that's a good sales force integration between amazon connected sales force to have that salesforce data right there on the agent screen.

and then after the call is over, um we have some innovations and in doing post call work where all the logs, the call transcripts, the call logs, all the information about the call cal duration go automatically into uh into three for form analysis.

so this whole journey is basically uh handled by a amazon connected pays you pricing and um and easy scalability skip through this one. ok?

so that's traditional call centers, right? it, it, it's, it's, it's phone calls and and, and as you probably are yourself, um you know, and as proven by 85% of smartphone users, um you don't like to be called anymore.

so we gotta talk about modern digital ways of outreach.

so uh instead of doing calls, find new digital channels, um build that content, but also the message that you send them have to be personalized. otherwise you're sending a bland generic message and that doesn't help with, with, with positive outreach.

and so when you do that and you build all that um build these digital campaigns, you need to measure your um your goal against your goals as well.

so to do that, we have a complementary product to amazon connect called amazon pinpoint.

so amazon connect is the, is the call center. the the voice calls pinpoint focuses on digital communications through text through push through email, uh custom channels and yes, also through voice.

so where we can do where we can connect to where we can interact with amazon, connect to uh to do outbound calling in a very targeted way, right?

so pinpoint can be used in many ways, reminders, notifications, uh uh maybe uh a marketing campaign in there and then uh uh maybe even a stepping stone for a two way conversation um into a support chat.

so let's talk a little bit more about that targeting of, of audiences, right?

so it starts with who, who gets these communications. so, segmentation is very important and we'll get in the next slide a little bit closer on how we do that segmentation.

so uh uh um you have your customer list in your system of record and gonna figure out which part of oo of your system is part of this next campaign.

and then what do you send them? step two. it's got to be a personalized message, right? and the next slide, we'll dive a little bit deeper on how we have some of these approaches to personalization.

so between the who and the what we can then build these campaigns um that, that build this outreach and can be a journey, a workflow over time that reaches out to customers uh in a holistic workflow.

and of course, everything we do can be measured. so there's insights coming out of that as well.

so a little bit more about segmentation, think about uh uh properties of people that can be static, it could be a location where they live, it could be dynamic or evolving over time, for instance, their interests, right?

so if you have a sports app and a person likes soccer, but mostly on the weekend and seems to view a lot of brazilian soccer, then we can feed that back into the tagging structure for that segmentation to say like, ok, we're going to target that user with international soccer news but not so much during the week.

uh and, and we'll focus on some brazilian players there, right? so we do that. we built that campaign over time in a workflow.

so, excuse me, so think about a hotel booking as example. um the the the straightforward way would be yes, we will, we will send you a confirmation message when your hotel is booked.

but that work flow, that journey stretches out all through the whole life cycle of that, of that, of that stay. so day of check in, we get some messages, day of check out, we get some messages and then even maybe a week after the follow up, we can figure out like, ok, let's do a survey or something, right?

so that whole journey can be done to, to pinpoint as a holistic flow with engagement numbers.

um and, and, and yes, pinpoint has quite a bit of volume. we send billions of messages. hundreds of push notifications, there is uh we know how to do scale at amazon.

so again to uh um wrap that up, right? audience management, personalization, segmentation, build it in your campaigns and outreach through your digital channels through email through s ms um or maybe even through voice.

i'm gonna give you an example for that. um lots of customers, i i'm a nonprofit so i focus on public sector, but this is really applicable to all industries. let me give you an example of how we could, how we could um combine these, right?

so how can we can combine these various services together? where in pinpoint, you build the segmentation and you build the outreach campaign that campaign can then feed into amazon connect, outbound dialing.

so amazon is not just connect is not just inbound, this is much as about outbound dialing. so you build the journey in the campaign and pinpoint and then you hand it over, connect for outbound and then the outbound can be uh uh dialing.

we can have a contact flow with checking for a voicemail box. if it's a voicemail, we might leave a message or do nothing at all and then eventually transfer to a live agent, right?

so that integration in the amazon ecosystem is important, right? so you can stay and take advantage of all the components in the amazon aws ecosystem.

uh so not just pinpoint and connect, but there's a couple on um couple of services here that i already mentioned actually.

so amazon comprehend for about the first amazon transcribe for speed to text, right? so to build those cal call logs, amazon comprehend, which is the mahi learning services to do sentiment analysis and keyword matching.

so is the customer happy? is the customer met? will actually get a score that you can use for your routing and amazon personalized, which is uh think about amazon.com.

um hey, you've bought this, other customers have bought that. are you interested in this? you can apply the same principles through our a i service. amazon personalized to build these personalized campaigns and pinpoints.

give you another example. this is a little bit more non nonprofit focused. again, uh again, i'm a nonprofit essay.

so how can you, how, how could the integrated architecture look like if you have a website for donor engagement or if you're membership based organization for memberships? right?

so a lot of services here uh we'll actually go through them one by one real quick.

so user um a user logs in uh we have an identity provision provider service called needle uh as an amazon service

um, user logs into our system and then lands on our website website could be built. for instance, with amazon amplify, could be your on prime hosted website. we don't really care. right.

and then once you land in the website, there's an option to go either a chatbot route or contact a live agent.

so if you go to chatbot route, amazon lex is the service here to integrate with, that's a chatbot service. if you want to go the, the live caller side, we can plug into amazon connect.

and again, since the user is already logged in the system, knows who that user is when the call transfers, so we can pass that information on and do all the things that we talked about earlier.

um amazon lex can integrate with like a lambda to go to the system of record, like a sales force or a b boomerang or, or a blackboard or anything like that.

and of course, analytics in there as well. this is the time to take a picture and i i, i'm complete. ok, a couple of minutes left.

so i thought this is again, re invent 2023. what are we doing with j fa i in this space?

um actually, um adamski touched on this a little bit during his keynote. uh we of course launched amazon q as a huge um um new service that integrates with a whole bunch of amazon a us services.

but i want to focus on on three ja i innovations where we can provide value in customer outreach.

first one, expanding the asian capabilities, right? so when the asian is on the uh on the call with the customer in a call,

um the customer text can be prompted in these large language models in the in these ja i models to provide feedback.

now, i would recommend at this point in stage to keep that agent in the loop. you know, we still have some issues but we find these models to make sure they don't, they give accurate data, they give safe data.

um and, and, and, and, and it all stays. uh um um we don't, we don't give you wrong data basically.

so that's important to keep that agent in the loop for now.

um second, um um outcome is enhancing the manager task, right? so this ties into lowering that average handle time for the customer.

if the agent doesn't have to summarize the call, we can just dump the whole transcript straight into aj a i model that summarization can happen under the hood.

the agent cannot, doesn't have to worry about that.

and then third, um this is, i think a great interactive experience innovation that is gonna change the world in terms of chatbots.

um a lot of um um we all deal with chad bots in our real life a little clunky and i think ja i is really gonna change that landscape.

so how do we do that? basically dump all the data that you have and, and, and, and train and fine tune some models with that, right?

so all your customer data, all of your stuff, your system of record, you can add that to, to the model and in, in the us fashion.

uh and you've heard this in the various keynotes, a big point here is privacy.

um when you fine tune an fm or lm, that data stays in your vpc in your account, it won't leak into the base model.

uh other companies won't suddenly on a prompt, suddenly get your ip out.

so, so that's the amazon security is a job number zero. so we're very focused on that.

so dump all your company data in it and be assured it stays private, right?

all the previous customer conversations as well, the model can learn from that.

um and then of course, all your business data, if you're, if you're uh uh a nonprofit like a blood bank, uh all the previous donation data,

uh if you are uh uh a, a travel agency, uh all the routing data for the airlines, right? and then build that to solve those use cases that we talked about.

um for a couple of considerations, the first one i already talked about, right, maintain customer trust and compliance, keep that human in the loop for now.

so you can make sure the responses are given are safe um and, and trustworthy and, and it's the right data that the agent can come back.

i think we innovate over time in this particular area.

um leverage and protect your ip. so, um again, um put your private ip in there and, and, and feel safe and trusted that, that um that your data is safe with us, right?

and then the last one i would say ja i still new use it for your high profile use cases.

a lot of cases of simple chatbot interaction like looking up a reservation or looking up when you can donate dot again, can be handled by the existing amazon lax chatbot type type systems.

that's the same slide. so we're here to help. i'm a solution architect, your entry point in a to l aws has a large ecosystem from, from uh training programs.

and then when you're at implementation phase, we have uh professional services to help you out uh or one of our trusted amazon connect accredited implementation partners as well. Here is the remaining transcript formatted:

i'll put this, i'll leave this up for a little bit after the talk for just like a minute or so.

but the two reference architectures i showed are, are listed here in the middle.

you can read a little bit more about some of the gen a i improvements that we're making to uh to customer contact centers.

so with that, i wanna wrap it up and then thank you for your attention.

um name is eva jensen, come talk to me and have a great re invent.

thank you.

thanks.