Exploring AI’s Impact on the Contact Center

0:05
Welcome to the Bluewave Webinar Series: The Current. This webinar is designed to provide information on the latest technologies and how they can solve your business challenges. Today I will be introducing the artificial intelligence expert over at Talkdesk, Terry Sullivan. He is our guest, and my name is Bob Schweiss. I am a solutions architect with Bluewave Technology Group

0:27
Welcome Terry. So excited to have you here. Could you give us a little bit of an idea of who Talkdesk is?

0:34
Sure, thanks, Bob. I appreciate that and thanks for inviting me on. I’m really happy to speak with you today.

0:39
So Talkdesk for, you know, we’re a leading provider of, you know, contact center as a service in the cloud, right? So cloud-based services for enhancing contact centers. And we’re really focused at Talkdesk on enhancing customer experiences, right? And, and a big part of that strategy is taking advantage of all the new capabilities with artificial intelligence that are out there.

So that’s kind of really who Talkdesk is in a nutshell and definitely focused in on great customer experiences and really specializing on, you know, how we can help with different verticals like healthcare and financial services and retail and things like that.

1:14
Gotcha.

1:15
So I had some conversations the other day with some of my coworkers and we were all trying to figure out what AI is. We were all trying to figure out good use cases. Everybody’s talking about AI.
You see all these stocks that are flying up, but how is AI? What role does AI have in the Talkdesk technologies?

1:35
Yeah, so that’s a great question. So AI has actually been around for quite a while. We’ve just had this insurgents with generative AI and all the great things. But at Talk to us, we look at AI and kind of three categories. Those categories are to automate, to empower agents and to illuminate the business, right? So if you look at those 3 categories, we, we step back for a second.

1:56
The first one is automate. And so that’s the most typical one. You know, in the past we had, you know, I started my career as a, as an IVR developer many, many years ago. So what we’ve been able to do is using conversational AI to create more conversational experiences and automate interactions for, for businesses to help, you know, deflect interactions from the contact center.

2:16
So with that, we’re using concepts like natural language understanding so people can speak right, more naturally to the system.

2:23
And I think the great experience there for, for customers is that when they’re calling in, in the past, you know, I would have to, I’m calling for a reason, right?

2:31
I have to answer a question, and I had to listen to a bunch of menu options, you know, press one for this, press 2 for that.

2:37
And it could be, you know, and I had to think about what am I trying to do and how does that apply to somebody, to the menu options that I’m given.

2:45
With AI, what we’re able to do is we’re actually able to turn flip that around and what we can do is the customer just says whatever it is they’re looking to do, I need to pay my bill.

2:54
I need to find my account balance, whatever it is, they just speak the question that they’re looking to resolve.

2:59
And now the onus is on the AI to kind of figure out, OK, what’s the intent? What are they trying to do?
And if it’s something we can then automate, right, we can then use, you’re right, use speech and capture that information to help automate that interaction. So that’s really one of the big, big hits that’s been around for a while. So it’s a conversational AI and talk to us. We call that autopilot, but you might hear the term like virtual assistant or bots, things like that. And then another area there that we you’ve had a lot of success with is with voice biometrics, right? So that’s that ability for the system to, you know, to use your voice print. My voice is my password, right? And the reason for that, the benefit there is, right? It’s using biometrics, right? Think of it almost like, you know, your face ID on your iPhone, right?

3:46
But now it’s using your voice print, it’s very secure because it can’t be, you know, the voice print cannot be reverse engineered, but it really speeds up that authentication time.

3:54
So when the caller, you know, so we can authenticate in the, in the, in the IVR right before we present information, but also when it goes to the agent, the agent will know that this caller has already been authenticated, right?

4:06
So I don’t need to go through that 30-40 second process of identifying the customer and authenticating.

4:11
So it really provides some great business value there as well.

4:16
So this the second category is what we call empower.

4:19
So naturally you know, we try to automate every call we can, but there are times we’re just not able to do that.

4:25
So calls need to be escalated and you know definitely was more complicated questions.

4:29
So what we’re doing there is to help empower the agent during that interaction in real time.

4:34
And we do that using some technology we have called copilot, right.

4:37
So your copilot term out there quite a bit.

4:39
So it’s the agent assistant capability.

4:41
And what I mean by that is as the agent’s interacting with the caller, right, or in the chat session, remember all this is Omni channel as well.

4:50
So I forget that, right?

4:52
The system is listening in real time and we’re using AI to search knowledge, right, to generate responses to the agent in real time that they can then use to help answer that question.

5:02
And you know, the real benefit, there’s a couple benefits there, right?

5:04
It helps us reduce the handle time, right? Because agents don’t have to put callers on hold or look at post it notes or look through Word documents and things like that.

5:13
But then it also, you know, really standardizes the responses and the answers, right?

5:18
So it helps kind of give the agents, you know, better responses to give to the callers.

5:25
But the other thing we also do there is around what we call summarization.

5:29
So once the call is complete, another big task that takes a lot of time is the after call work.

5:34
So once the agents done with the call, then they usually have to spend, you know, a minute or two potentially wrapping up that call, right?

5:40
They’re typing notes, they’re updating CRM systems, they’re picking disposition codes, things like that.

5:46
And that takes time, right?

5:47
That’s time that’s not useful.

5:49
So what we’re able to do is we can actually use generative AI where we send the whole transcript of the conversation into the AI model and it comes back with the summarization, right?

5:58
So we’re really saving that time for the agent where they’re not having to, you know, type all that in, but the agent still has the ability to edit it and do things with it as well.

6:07
So that really helps us, you know, reduce that effort and reduce the burden on the agent as well.

6:12
Sure.

6:13
So a lot of people when they go out to the market to look for a contact center solution, you know, they see all these features, but they really sort of lose sight that the single most expensive element of operating a contact center has nothing to do with technology.

6:26
It’s actually the people who are sitting in the seats answering telephones.

6:29
So we see a big push of people wanting to figure out how can we handle those transactions really don’t require a human being.

6:38
And, and what I really hear from you is AI is really the approach to tackling that.

6:43
And instead of putting people through a dialogue driven menuing system, you basically have a platform that listens to the customer and routes them appropriately.

6:54
So, you know, we hear a lot of buzz around AI is going to be the end of the world.

6:59
Terry, is Talkdesk going to end the world with their AI solution?

7:03
Tell us, tell us how you guys are securing the world from that grave threat that we hear on the news all the time.

7:10
With any new technology, right, there’s challenge. I remember back when the web first came out, right? That was the end of the voice call center. No one’s ever going to call. Everything’s going to be done on the web, right. So the web was going to be the death of the call center. Well, of course, it changed how we work, changed what the nature of the work and remove the work, right? So, and that’s the same thing we’re seeing with AI. And I think the big buzz right, you’re seeing right now is around generative AI, right? So this is a giant leap forward capability.

7:35
So previously like a conversational AI and other AI systems, they were really AI geared towards making decisions, right? OK, why is this person calling? But it never actually generated the response. You still have to code the system to on what to say back to the caller right now, a generative AI, the AI is actually responding, right? So it’s actually responding. And it’s important to note that the way it’s working, right, is that that as it’s intercepting this text or whatever the caller is saying, right?

8:05
It’s just listening, it’s generating response, right?

8:07
That’s why it’s generative.

8:08
And it’s called the pre trained transformer GPT, right?

8:11
So it’s generating these responses in real time for the customer.

8:14
So you lose a little bit of control, right?

8:17
So, and the control is different, I should say actually, really, right.

8:20
So in those cases, right, what we’re doing is we have to do this concept of guardrails, right?

8:26
So the concept of guardrails and there’s some, there’s really some principles around AI, responsible AI, because what happens now is we have these giant models, right?

8:35
We take all this data, billions and billions of what are called tokens, which are you think of tokens like words, things and that and what words are in the system and how they’re put together will impact how it responds, right?

8:48
The language model is like a child.

8:50
It learns by what you teach it and what it observes.

8:52
So it’s very important to teach it the right things up front.

8:55
And the things we look at is fairness, right?

8:58
So we want to treat all customers fairly, right?

9:01
So there we’re working to take bias out of the models.

9:04
We don’t want the model bias to certain things responding certain ways, right?

9:08
We want to be fair and equitable, right? That’s a very, very key thing. And so there’s a lot of things we’re doing around putting guard downs to do that. The other key is interpretability. Why did the AI respond the way it did, right? Many people think that’s a black box, right? You just put in as magic happens and here comes your response.

9:27
Well, really the idea is providing that visibility, right?

9:30
So understanding the thought process of the AI took to make that response is very important, right?

9:36
Because I could help you when we do things to help train and things like that, that’s how we know how to adjust the model, right?

9:42
So it, it, it meets those goals of fairness.

9:46
And then another area too is privacy and security, right?

9:48
Everybody’s worried because if you go to like ChatGPT today and you just type in your questions, well, cheap, that’s open domain, right?

9:56
That’s out in the public domain.

9:57
So we’ve got steps in place where we actually protect data, right?

10:01
We don’t allow customer data to go into models. We don’t retrain with customer data. So making sure we protect that. And then it’s always using human centric design. So we always keep the human in the loop. You’ll hear that term the human in the loop.

10:14
So the human is still the instructor.

10:15
It’s still observing and guiding the AI, right?

10:18
We don’t let the AI just go off on its own.

10:20
We’re not going to do the Terminator, right?

10:22
Skynet’s not going to become self aware, right?

10:25
So we always have that human in the loop capability and, and you know, the tools that you need to really monitor that as well.

10:32
Got you.

10:33
So where do the AI tools that you’re pointing out to customers, where do they get their context right? Is that through integrations and to their line of business applications?

10:43
I don’t want people to have the perception that they buy Talkdesk and now they have to basically raise a child, right?

10:51
A lot of the footwork’s already done.

10:52
It’s just connecting it to the right data sources, correct?

10:56
Yeah, very good point.

10:57
So there’s a couple things, right?

10:58
The models are we have pre trained models and we have industry specific models.

11:02
So we’ve done a lot of legwork, right?

11:03
So we’ve got pre trained models, but then yes, and we have pre built integration.

11:07
So if you’re a healthcare system in the like Epic or Cerner back end systems for financial services, right, Salesforce integration, the most common one, right?

11:15
That’s all pre the system’s pre integrated there, right?

11:18
So you’re ready to take advantage and use those external systems because when you look at context, right, you want to do really hyper personalized service.

11:26
If you think about it, if you look at Ozzy Osbourne and Prince Charles, if you were to look at them from a demographic perspective, they’re the same, right?

11:33
They’re not the same age, both live in castles, both have been divorced, right?

11:38
But they’re clearly not the same.

11:40
If you were personalized to them, you’d want to provide different experiences.

11:43
So that’s why it’s important to have all that rich context of the back-end systems, their previous interactions, and what the AI is hearing and responding to.

11:51
Combining all that together to make a great experience as well.

11:55
Circling back to Talkdesk in general though, Talkdesk has really established itself, you know, as a core technology solution that that businesses need to have it, it’s sort of the glue that interconnects a lot of disparate platforms.

12:10
So, you know, I always refer to it as the platform of interaction, right?

12:15
That’s where we’re, we’re capturing people’s intents and outcomes.

12:18
But the date, you know, that the, the actual inquiry is getting memorialized somewhere else.

12:24
A lot of people today are concerned that, hey, AI is going to displace the workforce, right?

12:30
You’re going to be able to come in today and flip this magical switch that the reality though, is call centers today are hard to staff, right?

12:40
It’s hard to get people to sit in chairs and answer telephone calls, whether it be in a building or out in somebody’s home.

12:45
So AI to me is really driving a solution into the marketplace that allows you to be able to sort of level out your customer service levels, take those interactions that don’t require high level of human engagement, you know, a lot of thought, those utility type interactions and be able to allow customers to self-service.

13:07
Yeah.

13:08
So in in that vein of the thought, right, when I look at what I’m so I do a lot of work with customers where I, I do an analysis, I look at their call metrics and try and understand the behavior, what’s happening and where I can look at.

13:20
So for example, if I look at a contact center and I see declining service levels, right.

13:24
So they’re not.

13:25
So a lot of calls are sitting in queue for long periods of time. That’s indicative of not having enough resources to handle those calls.

13:34
And the most expensive and difficult way to solve for that, right is hire more agents, right?

13:39
It’s not a not a scalable model.

13:42
So there, that’s where we look at using conversation AI to help kind of deflect those interactions, right?

13:46
So because what you’ll see too is I’ll look at call servers.

13:50
This seems counter intuitive, but if the call center has really short call durations like the agents not talking, that tells me that the agents might be handling a lot of simple tasks, right.

14:00
So those are potential things we could look at automating, so account balances or just in healthcare, like confirming an appointment by calling just I can’t remember when where my appointment is or when exactly it is right.

14:12
You know, you shouldn’t need a skilled agent to handle that.

14:14
Let them handle the more complex interactions of, you know, managed care and things like that, right, that require that human touch that AI still can’t solve right to do that.

14:24
So that’s, that’s a key area where we look at efficiency.

14:27
So it’s really about understanding the contact center.

14:30
And that’s another area we do with AI is we have an analytics capability, right too.

14:34
So we can also use AI, we can look at things like what we call topic discovery.

14:39
So the AI can monitor all the interactions in the contact center without any training and identify.

14:45
These are the topics that people are asking you about.

14:47
So you can identify, oh, here’s a problem.

14:49
Here’s things people are calling us about that we didn’t even know that we didn’t understand.

14:53
It was a, a big issue for us.

14:55
And then also doing things around sentiment, right.

14:58
Another thing is like, how do you keep make sure maybe we’re doing all these things?

15:02
We’ve always got to make sure we’re, we’re, you know, achieving good Net Promoter scores and good customer set service and things like that.

15:08
So one of the things we’ve done, AI is going to do is something we’re excited about called mood analysis.

15:14
So traditional sentiment kind of looks at a call and it’ll say, oh, these, all these words mean negative sentiment, right?

15:22
And so that, and if this call had 80% of those words, but at the end, the agent saved that call, right?

15:27
That still might show up in sentiment analysis as a negative call.

15:30
Well, a, with AI, we can actually listen to that whole interaction.

15:34
We can understand, oh, it started off as a negative interaction.

15:36
The customer was frustrated.

15:38
We understand their mood and we can see the transition, right, and how that call transitioned to potentially a positive interaction, right?

15:44
And we can report on that and the AI will actually tell us how that happened.

15:48
So it’s a great tool for getting training back in and it’s really taking, you know, insights into the call center actually to that next level.

15:56
Yeah.

15:56
And from my, my experience having worked with contact Centers for years, one of the biggest challenges was quality management, right?

16:05
How do we take all these interactions that are coming in, whether it’s my voice or chat or, or, or SMS and, and how do we look at it?

16:14
How do we get a clear picture of what’s transpiring within our contact center?

16:18
Why are our customers calling?

16:19
Are they happy when they hang up, you know, that their call gets resolved the first time?

16:23
So it really seems to me that, that that’s an area where AI could definitely improve.

16:27
And you know, our approach in the past was we’d go out and do samplings, right?

16:31
We’d have 10,000 calls that would come in on a day and we would, we would listen to 10 of them, right?

16:37
And, and we had to pay a whole team of people to sit there and listen to those calls and work hard about and, and, and now with AI tools built in the contact center stacks, you know, talk to that a little bit about how it’s improving that process.

16:51
It it’s giving people a clearer, actionable picture of what’s actually happening with their customers.

16:56
Yeah, yeah, sure.

16:57
So the, I was meeting with the customer the other day and one of the, one of the people in the room because she was responsible for that quality management of grading the agents, right?

17:08
And I asked her, I said, well, how many calls do you score, right?

17:11
She said on average we do 3 to 5% of the calls, right?

17:15
So you think about that’s a very low not, is that really a statistical valid sampling, right?

17:20
Well, so I asked her, So what if with AI, I can score every call? Because remember, AI doesn’t get tired.
AI can work constantly, it can work faster. So we can do that basic scoring and then again using human in the loop. So the AI will score every call. And then I’ll call out the ones that really need a human to look at.

17:36
Like these calls were all compliant, you know, they look good, but here’s some ones that are outliers that you need to probably review and, you know, do training with.

17:44
So, yeah, by using the AI to do that right, we’re able to score, you know, to have much broader insight and score all of those interactions, whether they’re voice or digital, right?

17:54
And, and really allow the, the human workforce to focus on the problem child’s the, the, the, the interactions that need that special attention.

18:05
Gotcha.

18:05
And, and when you sort of look at the, the, the, the mix of clients that you have coming in today, how, how much of an impact is AI having in, in, in, in those contact centers?

18:17
I mean, is it, hey, this would be nice to have, but or, you know, this is starting the mainstream and the conversation’s pivoting from how the contact center work to how the AI is going to work.

18:26
You know, give me a, give me a gauge for what the adoption looks like on, on these AI platforms today.

18:32
So the adoption is, is really taking off, right?

18:34
So conversation AI is, is definitely taken off in especially in certain markets like finance.

18:39
But if you’re looking at it, AI is becoming so important and integral in the contact center that I think it’s going to be almost as important as providing dial tone, right?

18:47
It’s going to be that core to the platform just because of the, the adoption that’s in the benefits we’re seeing.

18:54
Our CEO recently put a post out on LinkedIn showing how we were able to drive abandonment rates, but through better call handling down by like, I think 60-70%, right? That’s a huge, huge win there, right? Just by doing these. So I think it’s overtime, right?

19:10
There’s a lot of adopters and I think we’re still some capabilities for someone in the early adopters, especially with generative AI, but a lot of AI capabilities are very mature being adopted quite a bit.

19:21
So and I think over time it’s, you’re not going to see a contact center without it, right?

19:26
And just like, you know, doing the Omni channel back in the I remember 10 years ago, nobody did Omni channel contact centers, right?

19:32
It was, it was agents here, it was this.

19:34
And then we had web chat, but the web chat was in a different channel, but now we’re consolidating it together.

19:39
AI is another key core piece of that, right?

19:41
That’s why it’s a core piece of our platform as well, right?

19:44
It’s not an add on, it’s a core component there.

19:48
Gotcha.

19:48
So as, as the contact center becomes much more complex and, and touches many more elements of, of the business, who are the key stakeholders that you’re seeing that are engaged in these conversations?

20:00
You know it, it used to be when people made telephony decisions, it was the IT department that handled it, right?

20:06
But now, since it’s touching so many other facets of the business, I’d have to imagine that, that that plate of, of, of participants has expanded.

20:14
So who do you guys envision being on a team when you’re getting ready to roll out it?

20:19
Yeah, that’s a great question. There are a couple levels there, right? So if we’re looking at the, the operational efficiency of the metrics driving core KPI is right. That’s, that’s the contact center team. So no longer the IT guys out there doing that. It’s the contact center team that’s telling us these are the goals we’re trying to achieve, right? So it’s the supervisors, the managers, we definitely are spending a lot more time, we’re getting higher up in the sea levels because of the impact that’s going to have. Because we’re seeing a lot of industries people are looking with AI, we can take call centers from just being cost centered. Even though it’s at the cost center, right? So just try and reduce costs, reduce costs to being an Innovation Center and a profit center. Because remember, this is where all your customers are having interactions, right?

21:01
So we’re seeing more like chief marketing, chief experience officers that are getting more involved with these decisions as well. Because with AI, especially on the automation side, right, we can put persona and personality into the system. It’s no longer we don’t want to be that rigid mechanical system anymore, right? So we want to reflect the brand, right?

21:20
What are the things that are important?

21:22
So definitely seeing that then obviously that, you know, we still get the CFOs and the, the, the other financial people there as well, making sure that right, everything we do has a strong return on investment. And that’s a key part of any, any making any technology decision. But AI for sure, right?

21:38
Sometimes AI can be great, but if the ROI isn’t there, right, then you need to reconsider. Is this worth the effort to do?

21:45
Sure. Thank you for joining us today. Glad we were able to have this conversation.