Shawn Cordner (00:25.71)

Welcome back to the Amplitude of Tech, the podcast that explores the intersection of innovation, strategy, and impact in the technology space. I’m your host, Sean Coordiner, Chief Marketing Officer at Amplix. In this episode, I’m joined by Stephen Fine, Senior Value Consultant at PolyAI, and he’s also now heading up their restaurant division. We dive into one of the hottest topics in AI today, how to measure the return on investment in voice AI. From automation rates and CSAT to first call resolution to agent attrition, Stephen shares expert insights on the KPIs that matter most.

when evaluating AI driven transformation in the context center and beyond. If you’re thinking about deploying AI in the customer experience strategy or struggling to justify the investment, this one’s packed with the data, nuance and actionable takeaways that you need. Let’s get into it.

Steven Fine (01:29.902)

So I am the lead of the value consulting team at Poly AI. So pretty much it’s funny is before I came into this role, I had no idea what value consulting was. It’s like a very tech company role. So I work across all of our clients globally. So US.

UK and Europe and a couple of other. And I help our team build out value cases and value narratives for our customers. Because I mean ultimately we want to help our clients understand the value of the solution and how it fits in.

Sure. You know, when I looked you up on LinkedIn, I saw a different title, which caught my attention.

I have actually recently while dual rolling, I am now also heading up our restaurant division. So general manager of the restaurant division. So that is a recent move. We’re putting a big focus to one of our core verticals is restaurants as a whole. So I’ve been asked to lead the charge on that and grow that out as well.

Yeah, it my attention. I’m a former chef and I read the Spokaneite industry. Did you have any background in that?

Steven Fine (02:47.726)

Absolutely not. So I’m definitely not cooking. We do actually, funnily enough, our kind of thing when we’re going in and doing deals is we just like, let them cook. But no, absolutely no background in restaurants. But it was something I was, know, one of our things that we’ve built out over the last year and a half as we’ve shifted and become.

I very generative first and built that from the ground up as the technology’s gone in that direction. One of the areas that we found we could do a lot more standardized and templatized was restaurants. And so we’ve done a lot of work to really bring that down so that we can deploy that exceptionally quickly and allow clients to start self-serving a lot more. So with that push, we’ve decided that that’s one of the things we’re going to focus on and build out.

Why is that such a critical vertical for you guys? Why is there so much opportunity?

I think it’s twofold. One is the deployment of it is you have a very standardized kind of template almost like everyone calling into restaurants you’re either going to be making a good chance you’re gonna make modify cancel a booking that’s gonna drive typically 35 to 40 percent of call volume and then the rest of it is often very simple questions that people just want to check opening times and these things so you know out of the box

you can drive significant value, and this kind of goes into value conversation why I think it’s valuable is, one, we know we can take away anywhere from 60 to 70 % of calls, like very confidently, we will take out 60 to 70 % of calls, which is massive for the restaurant. Two is, there’s a very strong revenue case, and I always look at what really excites me about the technology is when you find a case where

Steven Fine (04:44.524)

you are focused on driving revenue. It’s not just like this cost cutting exercise or cost efficiency exercise. It’s really about driving revenue. So 24-7-365, you can answer calls and make bookings so you don’t miss any of those bookings out of hours. So it just fits really into a differentiated value case that frankly they didn’t have before.

You know, was no way to solve this for restaurants before. So that’s why it’s become a focus. We’ve been doing it forever. are over 2,000. It was one of our first clients actually, was a restaurant group. And we’ve over 2,000 locations live, taking where we’re automating calls. mean, frankly, I’d expect us to be closer to the 4,000 mark by the end of the year.

It’s a lot of Yeah. I remember being in the industry, restaurants were chronically understaffed, usually chaotically busy, especially during obviously certain hours. And usually underfunded and super slim margins if they’re lucky to be turning a profit at all. So it seems pretty easy to establish that value, right?

Exactly. I mean, look, I know that’s my title now is restaurants, but the core verticals for us as well, we are agnostic. I mean, we worked with you, I think we just brought in a customer as well. I’m not sure I can mention the name, so I’ll hold back on that. But then, know, areas like where we’ve got a strong footing and we know we provide a lot of value is, know, utilities is big, FinServ, so your banking sector.

really really strong, especially like retail banking. And then also healthcare is we’re seeing significant uptick in the healthcare industry looking for what we’re doing. They’re always the later adopters, you our first clients, you know, we were working with banks and utilities, but now we’re starting to really see the healthcare start going like this, this technology, we’re the same problem, chronically understaffed.

Steven Fine (07:01.93)

not super resource like they can’t just throw resources at it. It’s not always the most glamorous work to do. And also a lot of healthcare they’re not set up necessarily to run contact centers. So the phone is, we say it’s like interesting thing with some companies that get set up is they end up with a contact center because they…

we’re really good at business. They were really good at business and they’re like, okay, well now we’re in the cut, we’ve got to run a contact center as well to support all these people. So it’s like a really interesting way that they end up there. So those are like our core verticals, but we do kind of go across all of them.

Yeah, it does sound like a lot of them, those core verticals have a lot of the same problems, right? There’s a lot of overlap there with, you know, margins and, and, being understaffed and, being chaotically busy, right? That’s a, that’s a theme that seems to, to be one of the pillars or some of the pillars of establishing value healthcare in particular. I’ve had the opportunity to speak to some of the CIOs of really large, healthcare payers and networks and providers and,

they’re getting killed right now from all different directions, you know?

Yeah, and I’d like interesting one as well in the healthcare that we found like a really nice and footing in is things with his I’d always turn them like distributed sites so places where there’s at least over the last you know 10 years there’s been quite a lot of consolidation of things like dentists and Dermatology offices where private equities come in they’ve brought all they’ve bought all these little

Steven Fine (08:40.025)

little groups and are bringing them under one roof. And what they’ve done, a big part of that is now you’ve got hundreds, if not thousands of sites with some of our clients across the US, which is predominantly where it’s happened. And when you end up with that is you have all these distributed sites and the way that they deal with the phone often is it’s still at that same thing where it’s going to the front desk. And it’s the same, the challenge is the front desk are not there just to answer the phone.

It distracts them from serving the customers in front of them. So they end up with the really really high volumes where they’re missing calls because you always have this tough thing the moment that there’s a customer in front of you or a patient in front of you Your job is to serve them. So if the phone rings that just gets put on the back burner Yeah, and so you end up with a really high volume of missed calls and challenge with missed calls in these places is

people are trying to schedule appointments. People are trying to make appointments and appointments is your downstream ends up in people paying you and lifetime value. So there’s a really strong footing in that kind of distributed market as well.

Yeah, and that’s a bad experience for everybody. It’s bad experience for the people calling in. It’s bad experience for the person standing in front of the front desk operator. it’s bad experience for that employee as well. They don’t want to live in that chaos. And that makes it a tough morning to have to get up and go to work and face that,

I mean they get put under so much pressure and it’s just like they’re overworked often overworked, underpaid and appreciated. So hopefully we’re trying to like I always think of it as like this matrix of we want to serve the customer, right? So our customer, whoever we sell to, we want to serve our customers customer and where those kind of two meet together that’s where we feel like we’re

Steven Fine (10:42.168)

going to create the most value for them as a business.

Yeah, I’m a little new to the CX space. I’m an old telecom guy and then I kind of took a little divergence into marketing. Still in the tech industry, but I got a little removed from the technology, right? So this space has really evolved quite a bit since I circled back and got closer to the front lines of selling the tech. But it seems to me like

In healthcare, example, patient experience is critical, right? It’s a brand thing. It’s a differentiator. It’s something that keeps coming back and increases the lifetime value of that patient, right? Patient compliance is another area where I think CX technology is, there’s an opportunity for it to increase compliance, which helps everybody, right? It helps them get paid better. They have certain metrics that they have to meet and the patient will have more success. Obviously, that’s the

bottom line is what you want is healthy patients. But I think underappreciated is the experience of the employees for the company, right? And not just in healthcare, but the contact center employees, know, that’s a high turnover environment.

It’s a difficult job. a high stress job. They’re getting abused quite a bit, right? So this technology can, I think, deflect some of the emotion and some of the difficult, the transactional, but also kind of challenging calls. What do you think about that?

Steven Fine (12:10.038)

Yeah, I wholeheartedly agree. And I really like the part of EX. It’s like the employee experience matters in all of this. I just want to, there was one of the things you spoke about was attrition there. And maybe it’s worth me giving like a bit of a story to explain how I think about what I consider probably one of the biggest challenges within these contact centers.

and that is attrition, is we were working with a client that during their peak seasons, so during summer, they had to more than double their workforce. So it was moving from, I think, was 600 to 1,200 staff members. And their attrition over the year, though, was 200%. Now, look, some of that is voluntary, some of that’s involuntary, and that’s fine. But what consistently we see across

clients for the attrition challenge is this idea of early life attrition. Where people come in, it’s often younger people, very commonly younger people, they come in, it’s like a job, they’re like, okay, I’m gonna go get paid some money, get this job in a contact center, and then they come in and they realize this wasn’t exactly what they thought it was gonna be.

This is stressful, this is hard, and not always the most fun. And what ends up happening is you have this idea, but in the first 30, 60, 90 days, you end up with most of your attrition happening there. this client that we’re working with, in the services space, is they had to hire three people just to keep one after 90 days.

And now the downstream effect on the organization having to train on board that many people is just painful. the other part of that is, our head of VP of solution consulting always says that he’s like, look, the last, you’re hiring that many, you start in, you’re starting to like, so sounds bad, but like scrape the barrel, you know, you’re,

Steven Fine (14:35.71)

not just hiring the best few, you can’t be nearly as picky, you’re just hiring anything that will come your way. And then it’s just this, like, it’s a self-fulfilling prophecy of you hire people, you’ll hire anyone, because you need to get people on board, they’re not the right people, they’re not committed to this job, hire them, they churn. And then it’s just like a tough high churn, tough working environment, and it’s just like over and over again. And that’s one of the big things that, you know, when we help

take away a lot of the kind of call it simpler transactional and sometimes reasonably complex but repetitive calls you end up not having to hire as many people into those roles and you end up being able to train people for more meaningful work and when you’re training them for more meaningful work and you end up keeping them for longer the calls that they deal with

are not always the same ones. I’d like, I’d like it can be drab in there, where you end up on call after call being the same one over and over again. It’s draining. So, we always, that attrition idea is one of, to me, one of the biggest value drivers that we talk about, you know, costs between five and 15 grand, thousand dollars just to hire someone. And then when they’re leaving that quickly, it’s painful.

Yeah, that turnover has a very high cost and how do you get to $5,000 $15,000? Is that kind of an industry standard or do you go through a process of discovery with your clients and help them figure out what the true cost of their hiring is?

So it’s a good question and I’m gonna give the most consulting answer ever, which is it depends. Some clients, and really what it depends on is if the client has the information. Depending on the scale, size and data that they have, sometimes they’re like, look, this is our cost. We’re…

Steven Fine (16:43.406)

We know we have very high attrition and we’re doing this a lot so they’ve invested the time to figure it out. And they’ll come and say, because Ford, it’s $5,000, we know that’s what it’s gonna take at least to hire recruits, train them up, and get them on the phone. Then what often we help them think about is there’s also the productivity ramp. So you come on, you pay the $5,000 to get them.

to get them on, trained up and on the phones. But then they’re like 25 % productive for their first like month. And maybe 70 % productive by month two. And then at like your month, by month six they’re 100 % productive. The problem with that is if you think about the attrition, early life attrition is you’ve lost most of the ones that are fully productive. So you just end up consistently having these unproductive, like not fully ramped agents.

And most of the way contact centers are going to be structured when bringing people on is the newer agents are going to sit side by side with experienced agents, learning from them, and the experienced agents need to teach them. And then you suck productivity from the experienced agents. like just to, I mean, you have to get these people up to speed. So I I help them often help that on that productivity side once you have it.

We just go with them. then what they have, and then often we’ll go through on the ones where they’re unsure, we’ll go through a process of trying to figure out wages, costs, and training. So build that up from kind of the ground up. But then make sure, I always say you want two data points to validate something. So can we build it from the ground up?

see that and then does that align with what we’d expect from a benchmark perspective. Because then that’s your like comforting point. One data point is always a little dangerous. Two, you feel a little bit more comfortable.

Shawn Cordner (18:50.286)

Now, we’ll talk about upskilling for a second. Aside from the impact it has on attrition and increasing employee satisfaction over time and just kind of keeping them there, is there another way to quantify the value of having more skilled resources?

I think I’ve always found it quite challenging to quantify in like dollars and cents. Like that’s and maybe I’ve missed something in that. But I’ve always think like what you end up seeing and it’s not like the day the next day it’s all there. But if you’re able to upskill and have better people you end up having higher first call resolution. Like your people when they call in.

They’re more skilled, they know what to do, people get their query solved first time around. And that’s massively, like, first contact resolution is often the gold standard that you want, because first contact resolution drives fewer repeat callers. It drives fewer callers calling back about a similar issue. And so it ends up driving down your call volume, which helps everyone. One is great to the point, great customer experience.

Because they get what they want first time around they walk away happy Great employee experience because they’re not just having people call back and frustrated because when you don’t get a soul first time around and you call back you’re frustrated You know, you’re like I you told you you were not able to help me I’m now annoyed and and that then hurts employee experience and so I think like I’d always look at it in in this more holistic sense of What are we seeing on? first

contact resolution, what are we seeing on customer CSAT, and use those as ways to measure what that impact of upscaling is doing.

Shawn Cordner (20:42.85)

Yeah, maybe I’m a hammer and everything looks like a nail to me, but I look at these things through the lens of brand equity too. First call customer, first call resolution is something that is going to drive customer satisfaction. That’s going to drive customer loyalty. That’s going to drive people sharing their experiences. And going back to the restaurant industry, when I was in that space, the chefs would frequently remind you that a happy customer will tell one or two people, but a dissatisfied customer

is gonna tell everybody that they know that they had a bad experience. That’s way before social media was a thing. So, you know, now amplify that by 10 or 100 times, right? So everyone has a platform now and the experiences that they’re having engaging with your brand and all these key interaction points, especially when they’re in a state of frustration to begin with because they’re calling because something went wrong, right? And they want you…

to not get in their way, they want you to facilitate them being happy and getting the resolution that they need.

Exactly and I mean you bring up important points on brand equity and I mean I’m just gonna I’m gonna side side side track for a little bit because I think it’s a really like important concept that I speak about like when we’re talking to customers like what actually is what is value to them what does that mean and so often like

everyone just boils it down to ROI. You know, it’s the ROI. And the truth is with the ROI, AI is everywhere. Everyone’s getting told to do it. But at the same time, counter to that, there’s budget pressures and purse strings are tight. So the ROI is a way to really help justify making decisions. But ultimately, when you think about value it to me,

Steven Fine (22:43.114)

at least, is just the gap between where you are today and the future state that technology or solutions and services can get you. And that gap to me is value and it’s not always quantifiable. Often what you’ll see is you’ll have these different value levers of, think about competitiveness in the market. I use this of, we have a client

who is smaller than their competitors and is more of a startup. But because we’re able to implement, we’re able to take such a heavy portion of their calls away, it means they’re able to compete with players that are way bigger than them. It’s not just a cosplay. And it’s not just about the dollars of sense that you can save on labor.

it’s being able to compete and gain market share at a much more effective cost base allows them to start leveling up. And ultimately when you can do things from a better cost base and provide that same level of service, you’re reinvest that money to start chipping away at market share and on other investments. And that’s where I think often everyone gets caught up in, just tell me.

tell me the dollar and cents and frankly it matters. It does matter, I get that. I do it all day. But sometimes not losing the trees for the forest or the forest for the trees, whichever way that’s supposed to be. In terms of really looking at where, if this is where we are today, we think we can take us into this future state.

and this is what strategically important for the business and about that future states and lining it up yeah i mean i i i know that kind of just went side side on that but not

Shawn Cordner (24:39.404)

That’s perfect. in my world, establishing ROI on branding activities is the constant struggle, right? You can’t do it. It’s impossible. Even the most sophisticated businesses out there, they come up with these really complex algorithms and

At the end of the day, everyone’s just winging it, right? It’s really hard to quantify the impact of the brand work that you’re doing, but you know that it’s having an impact when you see the ability to penetrate a market quicker or when you see, you know, sales cycles being reduced, right? And when you see a win percentage starting to go up on opportunities, right? So you know it’s having an impact, but it’s hard to quantify it in dollars and cents. And that’s a tough discussion.

every single board meeting, right? Like, what’d you do this quarter? Well, you know, we refined the brand messaging and we updated the visual identity and we rolled out this new thing and great, how many leads did you get from that? I don’t know.

Yeah, I mean it is is hard because ultimately you have it has to go it has to flow down to to revenue and and and profit and but it’s it’s it’s more of an attribution problem than anything else does it work it’s an attribution issue.

But I think it’s a critical piece of the puzzle because if you don’t have something to differentiate you and the way that I think about brand strategy is a Venn diagram. So one circle is what the market needs and customers demand.

Shawn Cordner (26:23.746)

The other circle is what your competition fails to deliver. And the other circle is what you do best. And it’s the overlap of those three things that you really need to focus your message on. And that’s how you’re going to penetrate because you know the customer needs it. You know your competition is not delivering it. And you know that you can deliver it exceedingly well. And in my experience, especially in B2C, customer service is very often that thing that the competition is not delivering. And to your point about being a smaller business that’s trying to compete with larger businesses,

you need to wage asymmetrical warfare. And the best way to do that is to use their strength against them as a weakness. And the strength of a big company is their size and their scale and possibly competitive pricing, right? But the weakness of that size…

is it becomes an impersonal experience. The volume makes it difficult to make it a really good experience. And people just don’t feel a sense of connection and community to a business like that. So if you as a startup can leverage AI or any other technology as a way to make your point of entry that weakness that they have, I think that gives you a lot of power and leverage. at the same time, it’s something that you have to be able to continue to scale.

over time, otherwise you lose that competitive advantage.

Completely and I think what you said there’s something interesting in that it just made me think about and You know you can scale that competitive advantage and with AI You it’s often actually what we’re now I like trying to help the enterprises doing is right is you say you know a lot of them potentially become impersonal a little bit more impersonal

Steven Fine (28:10.35)

And that’s a level of service isn’t what they needed to be because they know that they’re comfortable and like losing one customer isn’t the biggest deal because it’s just a tiny portion of their pie. It’s really what we’re helping to scale that personalization as much as possible and personalization at least with the AI personalization. So it’s why we put such a radical focus on making their

voice assistants and AI agents sound lifelike and feel lifelike. It’s not just sounding, it’s feeling lifelike. Like someone who understands you, puts pulling in all your data to understand actually what, this person is, who is Sean on the other side of the phone when he calls in. So I can tailor the conversation to you. I’m able to actually respond effectively and get you where you want. Because ultimately at the end of the day, like,

people call to get a job done and our responsibility as a company is to help them get that job done in the most efficient way possible. you’ve kind of like, that’s the first part of that is making sure that they actually engage at all. And so it’s always been the challenge is people hear the bots and they go, give me a representative. And that’s why we…

make such an effort to have a lifelike conversation because at every single point you need to build trust. Trust drives engagement. Engagement drives resolution. And that’s the path that we follow in every single time they speak, every single time we try to understand them, every single time they respond, there’s a potential opportunity for us to lose trust with the customer and all of our tech and the proprietary tech that we have.

The models that we train and the way that we design conversations is all built to have to make sure that at every single point in the conversation we deserve to have the customer and our customers customers trust.

Shawn Cordner (30:26.222)

Every interaction is the potential to lose trust, but it’s also the potential to gain trust and loyalty if you can exceed their expectations in that moment. When I, as a consumer, when I call into customer service, my expectation is that it’s going to be a miserable experience.

Yeah, I’m going to it is not sad so

I’m already in a bad mood. And then as soon as I hear that voice that sounds like a robot, I’m like, here we go. You know, so then I’m already up another notch and then I’m zero button to see if I can, if I can get out to somebody live, right. But I have a, my mobile provider, um, I don’t have to call them often. It’s, know, it works just fine. But once in a while I do and I call in and there’s this super friendly, uh, AI voice and he’s just like, Hey Sean, what’s going on today? And I’m like,

Let’s talk, you know, I’m happy to have that conversation with that robot, you know, so it’s interesting how, you know, something as little as just making it feel real and believable and just natural can lower defenses and increase adoption. And do the metrics bear that out?

Yeah.

Steven Fine (31:38.35)

Yes, yes, very much so. So I’ll give you an example of how we’ve seen it play out with some of our clients. recently, like when we were, we always used to operate in a more of an intent-based world. And as we’ve shifted a lot more to generative is the conversations can be more fluid, they’re more flexible.

you can move out and in conversations in a much more natural way. It’s able to respond more accurately with, say, Sean, say, Steven, and it doesn’t feel bad. And with the quality of voices that you can do in text-to-speech now is unbelievable. And what we saw with the client is, in the old world, they were seeing their automation rates and CSAT rates.

at the kind of CSATs were in the 80s and containment rates or this was resolution rates with them, the resolution rates were in the 60 % range. And literally almost overnight when we flipped from intent-based to generative, we saw the CSATs move straight into the 90s because customers were able to get what they wanted every single time and that

getting what they wanted was resolution rates where it bumped, it moved up to this high 70s, 80s, and on one day 90s, which I was really hoping it stayed there forever. But that’s, you’ve just seen this uptick in ability to be flexible and guide and speak to people with larger knowledge bases. it’s like you instantly see CSAT jump, conversion rates jump, and resolution jump, which is really what you want at the end of the day.

Are those metrics, do they usually move linearly? In other words, CSAT score goes up, then you can predict that deflection rates are down and, or sorry, let me ask that question again. Are they connected in the sense that if you see CSAT scores going up, then you’re gonna see first call resolution go up and other key metrics, do they kind of align linearly?

Steven Fine (34:01.757)

You know what, metrics are really hard sometimes and I will tell you why. One of challenges that we’ve always had with CSAT is most of the time, often people give CSAT on calls that they haven’t hung up on or they… So there’s like this survivorship bias. We often see that ends up with higher numbers because people who are frustrated either we…

If it’s something we’re not able to help with, we transfer them through. So you have this survivorship bias. And it’s always this balance. I used to say this to clients every now and then. It’s like, we could give you 100 % containment. Or we can give you 100 % containment. We’re just gonna tell every single person to go away. And we’ve worked with clients where before us coming in,

they’d go down this IVR tree and they’d end up a number and it just hangs up on them. And that’s still very common practice. Like very, very common practice is that you go down and if there’s like, you end up at number three, you click one, three, three, three, five, and you eventually you’ve worked so hard to get to the point and then they just hang up on you. They’re like, that’s not something we want to deal with. And so like we can always get really high containment, but then what you really want to do is balance that with, so really high containment.

low CSAT in that example. You want to balance that down and going like let’s automate and resolve the maximum number of calls that we should automate. I use the word should because the truth with automation is sometimes you do need humans. Like humans are weird, wonderful, quirky, but we’re also able to be exceptionally nimble.

I can go run over to my colleague and get something done. Sometimes it is a very, very personal matter that you don’t feel comfortable with the robots. If your house gets flooded and you’re calling your insurance, you’re without a home at that time, it’s a very vulnerable situation and sometimes it doesn’t make sense to automate those.

Steven Fine (36:26.99)

You don’t automate that. But sometimes automating a claims form makes a lot of sense. So I mean I’ve kind of gone off track with the containment but you want to level of automation with the customer experience. And I’m trying to remember Gartner had this great cube, or it was Contact Babel, had this great cube showing like

of the, like, it was a matrix showing like percent, like calls that you should automate. And ultimately, like, if you’re able to, you take the calls that you’re able to automate with very high levels of success, go automate those transactional and repetitive ones. Like that’s really what you want to go after because you’re going to provide a very good customer experience for those. So go after, automate them, but don’t try

fit every single call into, we should automate this. At least for now. Eventually AI will be ridiculously all-knowing and smart and we can just call in and it does everything for us, but we’re not there yet.

Thankfully, you said something that I think is so important that I want to go back to it. You were just kind of talking about emotionally loaded calls, right? There’s a guy named Antonio Damasio wrote a book called Descartes Error and his thesis was

You Descartes said, I think therefore I am. And he says, I feel therefore I am. Right. And his argument is that we’re emotional creatures first, not rational, logical based creatures. And I think that’s that’s very true. Right. So there’s something to be said for, you know, the the path.

Shawn Cordner (38:21.174)

The shortest path to resolution being a straight line through AI is not always the best path to resolution. And back to the thing about trust and loyalty and ultimately brand equity is understanding that there are some topics or some purposes that people are calling in that are emotionally charged or are complex to the point where…

Even if AI is able to understand it, someone feels like they need to express it, right? So I think being able to differentiate between those use cases that people are calling in for and route them appropriately, but also giving people a quick exit when they feel like they’re going down the wrong path to get them into the right path is important.

Yeah, I mean I agree and I wish that there was often, you know, companies when doing this would take a more prudent approach. And again, the reason why automation has been so horrible and people don’t often like, they’re, your experience when you call into places to expect that if there’s any automation it’s gonna suck.

because it’s just been used historically as this tool to just try to get rid of people as quickly as possible and not necessarily serve them. And I think that that’s where there’s been a shift. You don’t have to choose anymore. You don’t have to necessarily choose between getting rid, just like sending customers away.

and not giving them what they want and serving them. That is, you can now serve your customers and not have to have them speak to a human. And that’s where the choice is just becoming more and more obvious that it makes sense that if you give the AI the right data, you program it in a way

Steven Fine (40:32.632)

that is engaging and makes sense and you teach it what it needs to know is you can drive a great customer experience and drive the efficiencies and goals that you want with call reduction. Because ultimately one of the things you’re always going to get is you’re always going to get calls reduced going to humans. And that does a couple things. One is it gives you this

huge bucket of capacity gain that you can either use if you’re missing calls to start answering your calls quicker. Get to your calls quicker, make sure that calls are routed to very specific teams, and all of that’s just about then customer experience. You get back to the point of you get your customers through to the right person first time around as quickly as possible, and that’s really what we always want to do.

Yeah, but it requires using AI, the tool as a scalpel and not a blunt object.

Yes, that’s fair. That’s fair.

So let’s take a left turn. think we went down some awesome roads here, but everything we’ve talked about here from an ROI perspective, I think we’re kind of the gray area of ROI, right? So let’s give the listeners some hard data that they can work off. So let’s start off with what do you think are the key metrics that businesses should be looking at or

Shawn Cordner (42:09.632)

tracking in order to establish a good ROI case for AI.

Yeah, so first let me say that I think it’s really important to one, build the case upfront based on benchmarks and data and what you can expect and what you think is going to happen. But then just as importantly, make sure that you’re tracking those metrics as you move through into value realization. Really, really important part of it. So I think like really what you, you wanna track a couple different metrics on either side.

One is you want to make sure just to get a holistic view. So one, you really want to start tracking your efficiency. So things where it’s about how much time are you removing from people. And I’m just going to give a typical contact center that’s not a revenue generating contact center. Otherwise, you can look at revenue, the revenue side of things in conversion, but just using a typical support.

non-revenue generating center is you want to look at your one side on your levels of automation, right? So how many calls are coming in and you’re removing those. All right, so calls answered. That’s going to give you an idea of how many calls you are receiving. At the same time, very importantly, you want to track your average handle time of your agents. And for two reasons. One is if you’re deploying something where

you know, you’re using AI to do identification and verification. Often what you’ll see is then a reduction in handle time. But on the flip side, often what you’ll see with AI, which is counter, is you end up with actually your handle time slightly going up for your agents. You take out all the repetitive and a lot of the simple calls, you end up with your agents getting the more complex calls. Handle times just by law of averages end up going higher.

Steven Fine (44:09.134)

So you wanna track both of those things, because you wanna get an idea of what is the length of call that AI is automating, because that’s gonna give you your level of comparison. So handle times and levels of automation on that side of things is critical. Then next thing on the efficiency side, you wanna track agent productivity.

look at occupancy and shrinkage. There’s two things that I’ve always wanted to look at, just to make sure that those are staying consistent or there’s a potential chance that you’re able to increase or sometimes the right thing is to decrease your agent’s occupancy because you can actually now give them a little bit of time, little bit of time back to go have tea and a biscuit if they want. So,

track that productivity side of things. And then on the other side is the attrition rate. We spoke about that at the beginning. Track your attrition rates religiously. Most of the time you can track it every single month, every week if you’re a larger contact center. But just to give you an idea of where that’s trending. And then one that I’m…

always very very bullish on everyone looking at beforehand and after is looking at internal transfers. So how many calls are getting transferred around the contact sensor because one of the huge things we see is calls coming into the contact sensor, they hit an agent, they go I’m not the right person for that and they get transferred across. Now normally that’s a function of the IVR just

having limited options of where to send people. And that is frustrating from the customer side, but also the agent side is just like, call, can’t handle transfers. you want to track that because with AI, what you’re able to do at least in the way that we deployed, where we deploy typically at the front of the over and above where the old IVR used to do, is we’re going to pinpoint routes people to the right place.

Steven Fine (46:33.678)

first time around. So you can get hyper specific on where you want to send people. So really important thing to track because you’re typically going to see that number go down, which is really impactful to the business. Other like ones that you obviously just want to track, but these are pretty, almost everyone’s always tracking them. It’s just your abandonment rate and average being to answer. And reason I say that is ultimately like, I always have this debate with people. They’re like, but

average speed to answer is 100 % with AI. So I get that, like AI answers, but the reality is there’s still the calls that AI doesn’t answer and you need to transfer those through to people and there’s still gonna be wait times and speed to answers on those. So you wanna just track that for two reasons. you, seeing that now your average speed to answer is just way like instant, there’s a decent chance that you’re now over-staffing and it’s giving you room to…

kind of let some natural attrition take place. Or if your average beat to answer is exceptionally high, probably need to like look at like looking at what other areas you could automate or frankly staffing up a little bit more. And then I mean those are like the core ones and then from an experience perspective, normally like this it’s the same struggles of what should you be looking at.

but CSAT is good but absolutely flawed. If you are tracking NPS, it’s good, again, like flawed, but these are all just like indications that you just wanna make sure that if you’re seeing CSAT on the agent’s side go down, it could be that your AI is doing some things wrong upfront, they’re not handing off the right information to agents.

they could be frustrated and they’re getting passed through frustrated customers. So I’d kind of like look at those. The one thing like, this is not a metric to track, but it’s worth like not forgetting, and that not everything is percentages and hard numbers. It’s like, a feel for, get your front, your agents to start telling you people their experience.

Steven Fine (49:00.216)

Like ask your agents to ask the question about the AI. Like get them to ask the customer what their experience with the AI was. Because they’re going to give you the… That’s where you get that first hand, it’s not just numbers, this is anecdotal information about what people’s experiences are. And you can’t overlook that.

I bet there’s a way that you can leverage AI to listen into those calls to capture those sentiments and then analyze them as hard data points,

Yeah, so it’s been really interesting sentiment analysis because we’ve played around with it quite a lot. working across the world from the Southern drawl to all the way to the Scottish accents. Every single person, every single accent…

sometimes pick up sentiments totally differently when they don’t mean it. So actually what we’ve found a lot of the time is, sentiment is reasonably accurate, but at what point does it actually just become dangerous you making decisions off of data that’s not accurate enough? Or what we’ve found is, often keywords in there as an indication is actually a very effective method of looking at sentiment, not trying to over-engineer something crazy, crazy fancy.

I don’t know, people are out there selling it. But sentiment, does work. But we’ve found for what we do, looking at the sentiment on the bot portion, the AI agent portion, it’s a bit more helpful just to use some keywords botting mixed with a little bit of other like.

Shawn Cordner (50:53.734)

Well, I think what I’m hearing from you, you’ve said a few things in this conversation that I think reinforce this idea, but data can be flawed, right? Like data can tell any story that you wanted to tell. It can also mislead you in a lot of different ways. And taking any one or two data points in a vacuum is not really going to tell you much of anything. So much of the data points that you were just talking about in the KPI that you’re tracking are kind of interdependent and related to each other and contextual.

And if you don’t have that full picture, and if you’re not really doing a good job of painting that full picture, you’re not really learning what the true story is.

And it’s not easy. Everyone wants this perfect world where it’s like you’re gonna end up with this measure that gives you the answer that exactly how everything’s performed. But the truth is we live in, a lot of stuff is a little bit in the gray and you’ve gotta use anecdotes and some data points to make…

Use them to make decisions, but you don’t just need more and more data. Sometimes a few data points are true to kind of guide you in the direction that you want, and measuring too much just becomes draining. always, I remember an old mentor of mine said, the point of measuring stuff is not to measure. The purpose is not to measure. That’s not why you do it. You don’t.

You do it so that you can get insights that can drive actions and actions that drive impact. And I always think like use as few data points, measure what really matters, as few data points as possible to try to the information that you need to make action.

Shawn Cordner (52:44.406)

I think there’s also the potential for a little bit of the observer effect, which is obviously a physics thing, but in the sense that I mean it is if people know that they’re being measured and that they live and die by those measurements, it’s going to influence behavior and maybe not in a positive way.

Yeah, I mean we’ve had some challenges that we’ve seen that that exact thing in contact centers is measuring agents on average handle time and very very common in the BPO market because often they’re paid, they’re getting, the BPO is getting paid per call sometimes and when they get paid per call they want, or even per person, there’s a finite number of times, is you end up the

BPO wants to maximize their profits.

Yeah, there’s an incentive for inefficiency.

Yeah, and they just like answer and get off the phone, like try to get off the phone. average handle time, bring it down because that means we make more money. And ultimately is often what that ends up happening is they bring the average handle time down because and they achieve the goal. They achieve the objective of what they were trying to get done, but then they don’t resolve the call. People call back and their frustrated customers call back.

Steven Fine (54:03.838)

and they go through another call of handle time and it ends up not being the most efficient, actually the way. It’s like trying to be so efficient actually stuffs up their efficiency.

But you know, it’s hard to have that conversation with the stakeholders that are above you, that data is not everything and ROI is not that clear. Do you have any tips for how a contact center operator can message that up to the executive and board level?

Yeah, so it’s I wish I had the golden the golden answer because then I’d make sure that we closed absolutely every deal we’re ever in. But what I have learned after doing probably a couple a few hundred business cases and deals now is one is like put on your executive hat and make sure you’re very attuned to what they care about.

and whoever you’re taking up this, if there’s a ladder to go up, that you’re very attuned to what they care about. Because if you’re not, if you’re just building numbers without linking that to strategic objectives and outcomes that the business cares about, it just lands flat. Most more importantly than the dollar number and the, how big the ROI is.

is that’s linked to a priority that executives care about. A priority in my mind, priority trumps ROI. Every day of the week, executives will invest and teams will invest where their priorities lie and not where there’s potentially just a very big ROI number. So I’d say, top one I guess, is make sure that you’re very attuned to what the…

Steven Fine (56:00.376)

team is and executives and whoever leads the division and is able to give you the money is aligned to. The second thing that I think is very helpful is make sure it’s wrapped in a narrative. that links to point number one is make sure that it’s wrapped in a narrative that builds a story that gives the context, the setting, explains what you’re going to do and then on the back of what you’re going to do, what that payoff is.

and be clear that what the risks are and the time that we’ll need to go into this. Because it just shows you’ve thought through it a bit more, that you’ve considered every single thing that we do has risk. Every single project that we go into, every single relationship that we form, they all have risks. And going in and showing that you’ve thought through the risks and the potential things that can go wrong.

is very very impactful as you as you presented to shows you’ve done your work the way as it gives you a way to thinking through the risks allows you to think through the risk mitigation strategies and which is always like it just shows that you’ve done the work and you understand why you and should it so i guess first uh… first one is make sure you’re a chintz you know uh… what your execs wants a strategic priorities to make sure it’s wrapped in a a narrative and i’d say the third one is like

don’t make the number ridiculously big. It’s so funny is like, I’ve never, I’m like always, we’ve won deals on this. I’m very realistic on what I think we’re able to do. Whereas I know that we’ve gone in and where there’s quite a few deals and it happened the other day is, competitor said, the competitor came in.

Yeah.

Steven Fine (57:58.338)

and they said they were going to do 3X your return on investment. And they were like, but that’s just nonsense. And we don’t trust them because it is nonsense. And ultimately is like, I always say this is, so I guess the point is make it realistic, right? Make it in the realm of reality and an easy logic to follow. Like we can get so caught up in this idea of like,

needs to be this super complex model that understand like is like super super built up over three years every single month mapped out and most of the time like it doesn’t need to be like that it needs to show clear logic on how you got to the numbers that you got and a few assumptions that build that are the building blocks to that outcome so make it real like realistic and

I’ve always gone in with the attitude of when we help customers build business cases and value narratives, it’s their business case. Like us as a team, we’re just there to enable them. Like ultimately, like we’re there to help them think through how to build this out. But at the end of the day, it’s not, it’s not poly AI, it’s not Amplix’s business case. It’s their business case to get a project.

And there’s a great person, shout out to the guy on LinkedIn, Nate Nasrallah, who has shaped a lot of my thinking on stuff over the last few years. So it’s like helping people understand the recommended approach to solving a problem. And I think that’s a really good way to think through things is every single way of solving things, there’s an alternative to doing it. It’s like you can…

you can implement if your goal is around scaling your contact center at a lower cost, you can take outsource to the Philippines. That is an alternative to investing in AI. But then think through what is the benefits and challenges with each of those alternatives. Because it shows that

Steven Fine (01:00:25.964)

that you’ve thought through these things. I mean, I always think that you should think through different ways to approach solving the problem, because there’s not only one.

Yeah, I think with technology, it’s really easy to get tactical fast and lose sight of the bigger picture of what technology is in service of, You don’t have this contact center.

just so that you can have an efficient contact center. It’s in service of something greater, whether that’s providing an outstanding customer experience because that’s how you’re gonna differentiate yourself in the marketplace and be more competitive. Or if it’s a healthcare application, it’s in service of providing better care and increasing patient compliance so that people can be healthier and live happier lives. There’s something that’s at the top.

that you really need to not lose sight of when you’re getting engaged in these kind of projects.

Yeah, I mean, and it’s not always easy because you do get stuck. Sometimes you get stuck in the tactical stuff and you just zoom out a little bit and remember what you’re trying. We’re all here to do. And it’s got to be end of the day serving the customer. They’re the ones at the end of the day who have our customers, they’re the ones that pay our wages. They’re the ones that mean we can do the work that we do.

Shawn Cordner (01:01:51.054)

A couple more questions. Where do people go wrong when they’re trying to set up this ROI case before they go into actually purchasing this AI technology?

That’s a good question. I think it’s very often, I’ll give you two sides of it. One is often when they’re thinking about the costs of the AI, they assume that it’s going to talk for the same length as a human. And the reality is you typically, the longest calls are not going to be done with AI, they’re going to be done with humans.

And so you assuming that the number of the handle time of an AI is going to be the same as a human is just massively incorrect because often you’ll get a huge chunk of calls that you shouldn’t be automating where you just route those straight through to straight through to the group that can handle them or team that can handle them. So just like making sure that you’re very aware attuned to that side of things. And big one is

not taking into account fully loaded costs. so a salary is only a tiny portion of what makes up the cost of the total cost of your contact center. On top of every single agent, there’s software, there’s benefits, there’s PTO, all these things that health, mean, health insurance is a huge one. And all these things.

that go on top of the cost of an agent and making sure that you bake that into the overall cost model that you’re looking at. So that gives you the one component on cost. The other thing which I see way too often left out is productivity. When you’re AI and humans, often you wanna do some sort of apples to apples comparison. Now, through this AI is,

Steven Fine (01:04:01.65)

24-7, 365, unless you turn it off at a certain point. It doesn’t take a break. It doesn’t get to to the bar. It doesn’t go to the bathroom. It never arrives late. And we kind of often, forget that humans are messy. They often come late. They rock up a little bit late. They had a big night the night before they come in a little late to work, you know? They need to go to the bathroom. They take paid time off.

all of these things, they’ve got to do training. Like they’ve got to be taken away from being able to do their job. And all of like getting taken away for training, they have to do the training and you need them to be good at their job. But it does mean that they’re in that time, they’re unable to answer any questions and help. So you always want to take into account shrinkage and occupancy when doing some cost analysis. And that’s, say that those are the two like

calculation parts of ROI that I think is very often missed. The next thing that I’m really cognizant of is making sure that you also put in the softer costs. So put in the softer costs about experience, about abandonment, whether it’s about helping on compliance risk because it’s giving a consistent answer every single time.

Whether it’s helping on your competitive positioning and giving you scalability is like those things do matter and Just because they’re you’re not going to attach a dollar and cents thing goes to the point of the value narrative and like building what’s important to the business is Weave in the softer benefits into your value case because often those are part of what is making the decision

And often it’s like the cherry on the cake. You you gotta have the ROI needs to be there because otherwise it’s very hard to justify upwards if you have to go through the strict ROI process. But then having the soft things is like, just adds it and look, it’s like, always think of it as like, this thing washes its face. You know it’s gonna pay back within six months. You can be fairly confident of that.

Steven Fine (01:06:27.918)

like your abandonment rates are gonna drop by 20%, you’ll see SATs, you’re see 5 % uplift there. And I’m not gonna try and attach a number to that, a dollar and cent number, because I just, don’t think it’s, Bain made a lot of money by bringing NPS and putting a dollar and cent number to it, they made a killing out of that. But I’m just, I don’t back it enough to put that into my cases.

Yeah, yeah, something that I have heard in other AI applications, not necessarily in your space so much, but I imagine it applies. tell me people underestimate the, the ongoing training of the AI model, right? And, and so there’s kind of some early inefficiencies and some bumps in the road, but you still need to dedicate resources to continually train that model and fine tune it.

And yes, I think there’s two, this is a very good point and it’s our CTO just wrote an article around the idea of us using forward deployed engineers and to businesses. There’s two things. First, the stuff is not easy to just get stood up. Everyone thinks like it gets just stood up in a day and it is perfect. For enterprise applications, let me read like for enterprise.

applications, there’s a good chunk of work to get this stood up and where it’s about designing conversations and using that and getting the technology stood up. It is hard to do and it takes resource. that is and it takes a very skilled and like a very specific set of expertise.

And that’s why from a setup perspective, we’ve always had used this idea of forward deployed engineers, which is where we’re actually deploying engineers from our side to come help you get this stood up, help make sure that the conversations are designed in a way that flows and is engaging. And then also they’re there to constantly look, monitor, and improve. Because it does need improving. mean, the way that we’ve always thought about it is you’ve got to get it deployed.

Steven Fine (01:08:51.214)

And then you’ve got to learn from the data. You’ve got to see where there’s gaps and then constantly make improvements. can fine tune some of the technology for automatic speech recognition stuff. You add new frequently asked questions. You change the way the flows are designed. So that you end up increasing overall resolution in CSAT. So it really takes a lot of work and is why we’ve always deployed with like our in…

in-house engineers because we have all the battle scars of things that have gone wrong that permeates the organization. you make sure that they don’t happen again. I I think it’s worth touching on this because it’s along the same lines is there’s a graveyard of high potential AI applications that have had to be parked and never gone into production.

because people have tried to do it all themselves without the expertise or they’ve been promised the world by this team that’s never stood up an AI application. They have an in-house tech team who’s like, I can build this. I’ve built a little demo. Look, it works. And it’s very easy to build a demo these days. But there’s so many where…

they just end up in a graveyard and they never see the light of production at scale because it’s way harder at scale. It’s way harder to do this reliably at scale compared to building a single demo.

There’s a saying in sales, never lose a loan. And what that means is leverage all the resources that you have as a salesperson, right? Bring in the engineers, bring in the executives, go in as a team, because at least you know that you did everything that you could to wrap your resources around this opportunity. And if you end up losing the opportunity, you know.

Shawn Cordner (01:10:57.996)

There’s nothing that you could have done differently. There’s nothing you could have done better. You didn’t leave anything on the table. And it sounds like that’s what you’re saying here as well, right? Don’t set yourself up for failure by trying to figure this out on your own if you don’t have the experience and the expertise. Leverage the resources of company like yours where you’ve got these engineers that have done this more times in a week than any one practitioner is gonna have done in a career.

Yeah, and I always say this is, it’s one that goes to our sales side, but it’s also the exact same on the deployments, is often in a customer’s life, they’ll buy this technology once. They will buy this technology once in their life. And where our job from a sales perspective is to help them make a decision that they only get to make potentially once in their life. And it’s the same with deploying this is.

Someone who goes and deploys this is a good chance this is the only time they deploy this kind of technology if they go and do it in-house. And that’s where we think it’s really important to use a team and use a platform that makes sense because often when it’s in-house, first they’ve got to spin up a whole tech team pretty much to get this self-deployed at scale.

really expensive first. But two is there’s so much dependency on that team of, let’s say they spin up a team of three. So you lose one person, you’re just in trouble. Like you are in trouble when you lose one person. It’s 33 % of your team is gone. you really like the benefit of…

for deployed engineers and having a company behind to support is that there’s not just a single dependency on one person who’s there. Like we’ve got a company who is built around people who can do this, who do this type of thing all day, every day.

Shawn Cordner (01:13:02.562)

Yeah, that leads me to my last question. So I think that what we just touched on is probably one of the answers to this question, but that is how can they set themselves up for success going into this and improving the ROI on these type of projects?

Yeah. Make sure that you’ve set up your case upfront and all the benchmarks that you’re to use as you go through. So from a value realization case, take that business case and keep tracking that and use the same numbers just for consistency. That’s like one, make sure you have that path and there’s a cadence on doing that. Make sure that you…

you have a team that’s able to, like, this is, guess, ensuring AI, ensuring this actually works, is like, make sure that you have a team who, when things go wrong, they’re like, going to be able to be there to solve challenges, because like, we’ve done enough deployments of these, and they don’t, they’re not every single one, as much as everyone would like to paint is like,

smooth sailing from day one and there’s never a challenge like it works and we do it really well but there’s often challenges that you have to overcome where something doesn’t go perfectly right on day one. So I’d say like make sure that you’ve got some sort of tracking framework in place, make sure that you’re updating whoever needs to see the tracking so that they’re aware of what’s going on and don’t over complicate it. Like keep it to a few

core metrics that essentially build up the business case because it’s much easier to track four or five metrics than it is to build out a business case every single time. So just keep your inputs reasonable, the number of inputs that you just can update rather than having to rebuild this thing from scratch each time.

Shawn Cordner (01:15:09.272)

That’s great advice, man. Great conversation, really enjoyed it. Thank you so much for your time and your expertise. Appreciate it.

And thank you so much for thanks to you, the Amplix team and everyone. I really enjoyed being here. Nice and hopefully I’ll get to see you soon.

Yeah, listen, if I ever open a restaurant, I will be calling Polly and on a case.

Or if you open a bank or a hospital.

Shawn Cordner (01:15:38.094)

All right, thanks man. Appreciate it.