Latest / The Adults In The AI Room / The Adults in the AI Room – Episode 102: AI in the Contact Center, Why Implementations Fail, and the Business Model Reckoning with Michael Darwal

Transcript

The Adults in the AI Room – Episode 102: AI in the Contact Center, Why Implementations Fail, and the Business Model Reckoning with Michael Darwal · The Adults In The AI Room

• 0:02
  SPEAKER_00: AI is moving fast, but inside the enterprise, fast is not enough.
• 0:08
  SPEAKER_00: This is the Adults in the AI Room podcast, where we cut through the hype and get into what happens when AI meets the real world.
• 0:16
  SPEAKER_00: We bring together the leaders, builders, researchers, and practitioners who have put AI to work from the pilot to production.
• 0:24
  SPEAKER_00: We'll challenge assumptions, expose the gaps, and unpack what it takes to turn messy data, complex organizations, and ambitious AI strategies into results.
• 0:34
  SPEAKER_00: No hype, no shortcuts, just the conversations that help you make AI work.
• 0:39
  SPEAKER_00: Here is your host, Seth Early.
• 0:46
  SPEAKER_02: Welcome to uh the what was formerly the Early AI Podcast, now rebranded as Adults in the AI Room.
• 0:54
  SPEAKER_02: This is our first episode with a new branding.
• 0:56
  SPEAKER_02: I meant to mention that to you, Michael.
• 0:58
  SPEAKER_02: Um, it's brought to you by Vector.
• 1:00
  SPEAKER_02: I'm your host, Seth Early.
• 1:01
  SPEAKER_02: Uh, and in each episode, we're going to be talking about a uh about how business uh and and and operations are being reshaped by artificial intelligence.
• 1:11
  SPEAKER_02: And uh today we're going to be talking about uh AI in the contact center, a space where the stakes are really immediate and high volume uh and expectations from executives and customers are are very different.
• 1:23
  SPEAKER_02: Uh when generated AI arrived, the dominant narrative was that it was going to automate away most of the traditional call center work.
• 1:30
  SPEAKER_02: What's happened is different, definitely more nuanced and much more interesting.
• 1:34
  SPEAKER_02: So joining me today is Mike Hardal, who is chief AI officer at IBEX, a global customer experience management company with nearly 40,000 employees operating 31 call centers across the US, Jamaica, Nicaragua, Honduras, uh, the Philippines, Pakistan, and India.
• 1:49
  SPEAKER_02: Michael has spent 13 years at IBEX across finance, digital marketing, and now AI, building the company's Wave IX platform for seamlessly integrating AI agents into traditional customer experience workflow.
• 2:00
  SPEAKER_02: Michael, welcome to the show.
• 2:02
  SPEAKER_01: Thank you so much.
• 2:03
  SPEAKER_01: It's a pleasure to get to speak with you, and I appreciate getting the opportunity.
• 2:07
  SPEAKER_02: Absolutely.
• 2:08
  SPEAKER_02: So um, let's start off with some of the common misconceptions about um uh about AI and the cost center.
• 2:15
  SPEAKER_02: When AA generative AI first arrived, the dominant narrative was that it's going to automate away 70% of traditional cost center.
• 2:21
  SPEAKER_02: What has actually happened and and and what do executives still get consistently wrong about what AI can do in the space?
• 2:30
  SPEAKER_01: That's a great question.
• 2:31
  SPEAKER_01: And I think if we take a step back, think exactly as you mentioned, when generative AI really started coming to the forefront a few years ago, people were under the misconception that it would almost immediately go after 70 plus percent of the contacts.
• 2:44
  SPEAKER_01: And the reality of it is when we look back now, I think MIT has published a study and there's been a number of studies that are out there, but it's you know, over nine out of ten, ninety-five percent of AI implementations fail.
• 2:54
  SPEAKER_01: And when you say fail, it means fail to deliver on the ROI or the promise of what they initially went out to achieve.
• 3:00
  SPEAKER_01: And that's for a variety of reasons.
• 3:02
  SPEAKER_01: And I think um one of the ones that I'd highlight is just a misunderstanding from the get-go of what it's trying to accomplish or what it's able to accomplish, then you know how it really fits into the customer journey.
• 3:14
  SPEAKER_01: And are we delivering a better result for the customer on the phone as well as for the agents and well for the company who's trying to deploy this?
• 3:22
  SPEAKER_01: And then also really what the what the ultimate goal of it is.
• 3:24
  SPEAKER_01: And are you getting a true resolution through using the deployment of AI, or are you simply um measuring off a single point in time and then getting a contact that recurs later on or in the very near future?
• 3:36
  SPEAKER_02: Yeah.
• 3:37
  SPEAKER_02: No, that that's interesting.
• 3:38
  SPEAKER_02: And and and uh as you saying, uh, we're gonna get back to the the failure, the nine out of ten implementations that fail.
• 3:45
  SPEAKER_02: But a lot of organizations are really looking at AI in the contact center as kind of a binary choice, right?
• 3:51
  SPEAKER_02: Either AI handles the call or human does, but why is that framing incorrect?
• 3:56
  SPEAKER_02: And what what what is uh looking at this as a spectrum uh produced that that binary thinking uh does not?
• 4:03
  SPEAKER_01: That's uh it's a great question.
• 4:04
  SPEAKER_01: And that means like when we look at it in a binary, that just simply is a pass fail.
• 4:08
  SPEAKER_01: Like, did an AI agent handle this all alone, or did it eventually did that person have to be transferred to a human agent, or did they did they have to call back in later on because they were dissatisfied with that?
• 4:18
  SPEAKER_01: And I think that's where a lot of the failure or the the failure to produce the initially estimated results are, and where the reality of and we're looking at it on the whole spectrum is are we trying to accomplish all of our three goals or at least one of those three goals?
• 4:32
  SPEAKER_01: And the three goals that I would think of there are are we ending up with a higher C set?
• 4:36
  SPEAKER_01: Is the customer on the phone getting the answer they want or need or getting resolution faster, more effectively, or more efficiently than they would have without the use of AI?
• 4:47
  SPEAKER_01: Are we able to do this in uh in a way that that produces positive ROI both for the customer, you know, in terms of having that positive CSAT as well as saving money?
• 4:58
  SPEAKER_01: And are we able to serve more customers with the same more consumers with the same number of human agents we were before, which is the factor of are we are we resolving this the first time that they're talking to us in an AI agent?
• 5:09
  SPEAKER_01: Are they having to come back in?
• 5:10
  SPEAKER_01: Are we able to, you know, be able to take some of the more road or easier, less complex interactions and automate those with an AI agent and then allow our human agents to take on those more complex journeys faster and get to a resolution more quickly by leveraging AI tools in the contact center on the on the agent side as well.
• 5:28
  SPEAKER_02: So, so uh so that makes a lot of sense.
• 5:30
  SPEAKER_02: So you're really you're really facilitating the the whole experience and you're you're injecting uh the AI uh assistant at various parts of that, right?
• 5:39
  SPEAKER_02: By you know, cat gathering your basic information, passing it on to uh a call center, a rep, uh looking up uh knowledge uh and and having uh answers in an anticipatory way, um, being able to make that um make that uh experience more effective by uh by uh uh making the the uh call center agent more effective and of course avoiding them if you can, right?
• 6:03
  SPEAKER_02: If uh if possible.
• 6:05
  SPEAKER_02: Um but um let's also talk about, you know, you know, you we talked a little bit about um you know nine out of ten or ninety-five percent of the uh fail of the implementations not achieving their objectives.
• 6:17
  SPEAKER_02: And uh talk a little bit about um what what is driving that and and what are the successful deployments uh what do they have in common?
• 6:29
  SPEAKER_02: What what are what if what's the formula for uh uh success?
• 6:32
  SPEAKER_01: Great.
• 6:32
  SPEAKER_01: So I think some of what, you know, to answer your first question, some of what's driving those failures and the failure to to achieve those objectives is you know a disconnect between the systems, a dis you know, a disconnect or a misunderstanding of what that customer, that ideal customer journey or what historical customer journeys look like, and perhaps uh failure to really know align what the agents, the AI agent's objective is with what the organization's objective is.
• 6:56
  SPEAKER_01: So um, if you were to say you had a subscription service that uh you automated disconnecting or canceling very easily, well, great, you could automate the heck out of every single contact where somebody just said in the first minute, I want to disconnect and cancel my service, and the AI agent just simply said, Great, we'll cancel your service right now.
• 7:13
  SPEAKER_01: But if the AI agent never, that's not a great business outcome because the AI agent, if it's not trained and it doesn't understand to probe or understand what the issues are or why this person might cancel or what could be done to save, how long or how persistent that customer has been historically?
• 7:28
  SPEAKER_01: Is this somebody that's worth really offering a discount for or helping to save and understand?
• 7:32
  SPEAKER_01: You're not gathering the right information.
• 7:34
  SPEAKER_01: And alternatively, if the simple measure of success is, you know, on a more complex call, did the AI agent handle this on its own?
• 7:41
  SPEAKER_01: Well, if it's not capable of fully handling it on its own because there's multiple issues that occur on the call or something else, like is there a different way to measure success?
• 7:50
  SPEAKER_01: And I'd use that in the case of there's a lot of call types that might have a handle time that is 20 plus minutes long, 40 minutes long for some very complex things.
• 7:58
  SPEAKER_01: Well, success doesn't just have to mean that that AI agent handled the 100% of it.
• 8:02
  SPEAKER_01: It could very well mean that I took what used to be a 20-minute handle time call, and I that AI agent did the front end of the call, talking to the consumer, gathering their information, verifying them, accessing their account information, clearing up the first four minutes of that and turning it into two minutes or or what have you, and then transferring it over to a human agent who, because they have all this information verified, the information's been summarized and identified for the human agent, they might be able to resolve those last few steps in a matter of 10 minutes or eight minutes.
• 8:30
  SPEAKER_01: You can shave off 50% or 40% of the time.
• 8:33
  SPEAKER_01: The reality is if you're able to save off, shave off any percentage of that, and there's less friction, less repeating of yourself, a quicker identification of the issue through the use of the combination of an AI agent and a human agent, that's a better experience for the customer on the phone, and obviously a huge cost savings as well through using that marriage and that seamless integration between the AI agent and the human agent.
• 8:54
  SPEAKER_01: And to your second question of what is more successful, what is making implementations more successful?
• 9:02
  SPEAKER_01: And I think that's really a combination of, again, technology and understanding and know-how.
• 9:08
  SPEAKER_01: So, you know, obviously on the IVEC side, we partner with, we believe is the far and away leading AI agent development conversational AI technology company and Sierra AI.
• 9:17
  SPEAKER_01: Their technology is tremendous.
• 9:20
  SPEAKER_01: It is incredibly intuitive, empathetic, a great voice.
• 9:23
  SPEAKER_01: But what we bring to the table and what we help and what really leads to successful implementations for a lot of our clients is being able to go in with that historical understanding from having handled millions upon millions of calls over the last two decades of what does this customer journey look like?
• 9:40
  SPEAKER_01: Mapping it out with our business insights and analytics teams, looking at all the tools and technologies that have to be integrated into that AI agent set, mapping out exactly what that customer journey looks like in an ideal situation, in a middling situation, and in what to do when it doesn't go well, and being able to have all of those things mapped out before launching the agent, but not having to do it for 100% of context.
• 10:05
  SPEAKER_01: We're not trying to boil the entire ocean to start with, but taking it off in digestible chunks and saying, great, these 30% or this quadrant of your call volume is going to be easier to automate this way.
• 10:15
  SPEAKER_01: Let's start with figuring out how we identify it.
• 10:17
  SPEAKER_01: Maybe we replace the upfront IVR with an AI agent and just get that routing done more effectively and more efficiently and don't make a consumer sit through that phone menu on the IVR and having to do that or having to repeat their voice or get frustrated.
• 10:30
  SPEAKER_01: Let's make the front end more successful and a little bit easier to navigate.
• 10:33
  SPEAKER_01: Then let's parse out that 30% of volume that's going to be easier to automate and we'll send that to the AI agent, and the rest will start capturing information and transfer over.
• 10:42
  SPEAKER_01: So the key part is using the best in class technology and marrying that with that unparalleled know-how expertise and in the mapping and analytics and insights as to what this customer journey is to look like.
• 10:54
  SPEAKER_02: Well, so just so let's uh uh talk a little bit more about that.
• 10:57
  SPEAKER_02: So you've had decades of customer journey data across all sorts of interactions and intents.
• 11:04
  SPEAKER_02: And how does that help?
• 11:06
  SPEAKER_02: How does that change what you can do with AI and what are organizations missing when they try to deploy an agent without that?
• 11:12
  SPEAKER_02: I can't imagine they would, but uh I imagine some uh have less of an understanding, less of a detailed and high fidelity understanding of their uh customer journeys.
• 11:22
  SPEAKER_01: Yeah, what's great is by looking at all that data and insights and then looking at the you know the client-specific data and insights and understanding great, this is what percentage of your contacts look like this type.
• 11:32
  SPEAKER_01: This is what fall into this queue or this type of contact.
• 11:35
  SPEAKER_01: These are the ones that we know that have a handle time that is shorter, so those are likelier to be easier to resolve.
• 11:40
  SPEAKER_01: Let's listen to some of these, let's listen to how those get QA'd, and let's listen to and understand you know what of those we should identify as most easy and readily available to automate with an AI agent.
• 11:51
  SPEAKER_01: And then let's look at the stuff where there's high customer dissatisfaction as well.
• 11:55
  SPEAKER_01: Where do you have customers who have to get transferred historically through multiple human agents who have longer wait times while things get verified?
• 12:02
  SPEAKER_01: And we say, great, how do we go through and we start automating and tie into systems to authenticate who this user is faster with technology up front, allow the AI agent to capture that information and queue it up and summarize and have it already queued through and already queried through a knowledge base, an automated searchable knowledge base that's indexed through the AI agent.
• 12:21
  SPEAKER_01: So that when they do, if and when they do get transferred to a human agent, that's there faster.
• 12:25
  SPEAKER_01: So being able to look at all of that data and insights leads you to a much better starting point of setting reasonable expectations, knowing what you should be targeting, and then knowing what the next step of it is.
• 12:35
  SPEAKER_01: So you it allows you to set the set the mark for what success looks like or failure looks like up front, modify and optimize off of that, and then know what you start chasing next or pursuing next.
• 12:46
  SPEAKER_02: And so that journey mapping is so important and understanding what all the data sources are at each point in the journey and what the pain points are and what their uh objectives are, what do they need to accomplish, what do they need next, anticipating that and then making sure that you're surfacing the right data either to the customer or to the uh uh call center agent or to the AI uh tool uh in that context.
• 13:09
  SPEAKER_02: So one of the challenges, one of the big things here is that um you know the business model for call centers was really uh based on price per minute, price per hour.
• 13:20
  SPEAKER_02: Oh, AI resolves issues faster.
• 13:22
  SPEAKER_02: That's intention with that, right?
• 13:23
  SPEAKER_02: It's like back in the days when I would talk to attorneys, and I'd say, well, you're gonna be billing fewer hours, right?
• 13:29
  SPEAKER_02: You know, save time and view bill, uh no, I don't want to do that, right?
• 13:33
  SPEAKER_02: That's but but then it raises the bar for everybody, right?
• 13:35
  SPEAKER_02: So AI, you know, resolves issues faster.
• 13:38
  SPEAKER_02: Uh, how has the pricing model evolved?
• 13:40
  SPEAKER_02: And what is that uh change about um you know cost per resolution?
• 13:44
  SPEAKER_02: How does that change uh the industry?
• 13:47
  SPEAKER_01: Great insight.
• 13:48
  SPEAKER_01: So um, you know, when we think about that historically, as you mentioned, the vast majority of traditional BPO or CX interactions are measured on a cost per hour or a cost per minute or even a cost per contact basis.
• 13:58
  SPEAKER_01: Whereas today we've been able to, instead of just delivering a cost per cost per minute or hour, or a cost per usage, like a pure play technology, when we deliver an AI agent solution, it's a cost per resolution.
• 14:11
  SPEAKER_01: So what we are able to do is look at, okay, this is what your typical cost per resolution looked like based on average handle time and what your usual bill rates are, and know what that looks like for an onshore, a nearshore, or an offshore deployment, and then look at what this can look like using either standalone AI agents or AI agents integrated into the human agents.
• 14:30
  SPEAKER_01: And what we're seeing is that obviously this creates an opportunity for you know 50% plus savings on a cost per resolution basis for most of the most of the clients with whom we're dealing.
• 14:40
  SPEAKER_01: What we're not seeing on our business, well, there's definitely an impact to a lot of companies inbound or the amount of traditional BPO uh revenue or CH revenue they do on an agent-based side, we're not seeing that adversely impacting us today because we're able to really absorb more calls or more contacts than we used to with the same number of human agents.
• 14:59
  SPEAKER_01: So that's creating, you know, creating revenue opportunities.
• 15:02
  SPEAKER_01: And we're seeing that as by being a leader and by pushing forward with presenting these and driving these AI agent solutions, capturing more volume from our competitors or capturing their share.
• 15:11
  SPEAKER_01: Um yeah, you're aging the bar.
• 15:14
  SPEAKER_02: And then and then others have to compete with you on those on that basis.
• 15:19
  SPEAKER_02: And and uh, you know, there's there's no turning back, right?
• 15:22
  SPEAKER_02: Once you solder those things at a lower cost per incident and lower time per incident and fewer callbacks and all of that.
• 15:30
  SPEAKER_02: So um you have seen AI increase throughput uh by agents, by human agents, by a significant amount, 40%.
• 15:37
  SPEAKER_02: That's enabling less experienced agents to handle more complex uh interactions.
• 15:42
  SPEAKER_02: What does that mean for how you recruit and how you train and how you develop people?
• 15:47
  SPEAKER_01: Well it's great is that you know, through leveraging that technology to make the agent more powerful or stronger faster, it allows us to have a great, you know, very solid increase as you said, 40% on who we're able to, the efficacy from recruiting to getting an agent live on the floor faster.
• 16:03
  SPEAKER_01: So it's a it's made it much more effective for us to be able to go out and recruit from a wider base of potential agents, uh, train them much faster, and get them to proficiency much more quickly, which also leads to a happier agent.
• 16:16
  SPEAKER_01: Like I think uh we went from what we already believe was it was you know at the high-end best in class 77 employee NPS, you know, a year and a half ago, to the most recent data that we published at the end of our fiscal year the summer was an employee NPS of 82.
• 16:33
  SPEAKER_01: Obviously, that's you know a combination of what we do for our employees, how we you know, how we treat them, how we promote and create opportunities for them, but a lot of it is also the job is becoming more enjoyable or better for the agent to do because of these tools and technologies.
• 16:48
  SPEAKER_01: So not only are we able to see better throughput from the recruiting all the way to getting them on the floor, having them be more effective on the speed to for efficiency side, but it's also a better experience for the agents.
• 16:58
  SPEAKER_01: They're ultimately happier, which leads to much better retention, leads to lower churn, leads to recruiting fewer people to replace the same job that you lost.
• 17:06
  SPEAKER_01: So that's we're seeing great impact on that.
• 17:08
  SPEAKER_01: Five point, you know, 77 from an employee MPS for a BPO is off the charts.
• 17:13
  SPEAKER_01: 82 is we're uncharted territory for that, which is fantastic.
• 17:18
  SPEAKER_02: So coming back to that in terms of um entry-level employees and contact centers have been where people get a lot of their first uh professional experience, develop foundational skills as AIN does more of the work.
• 17:29
  SPEAKER_02: Um, where does the next generation of experienced operators come from?
• 17:33
  SPEAKER_01: So I still think it comes up through people who have had that experience on that side because the the folks like we have a great track record of taking people from agent to supervisor to floor manager to director, and in many cases, multiple places.
• 17:46
  SPEAKER_01: We have country managers who started for us as agents or started for predecessor companies as agents.
• 17:51
  SPEAKER_01: And I think that's true.
• 17:52
  SPEAKER_01: What we do is really leverage an understanding of, you know, from the agents who move on to then the QA or the business insights and the analytics team, the workforce management team, to understand how the AI technology is benefiting those agents, seeing how whether it's when they were an agent or the teams they're managing and helping drive the iterations and the improvements of how we're using that.
• 18:13
  SPEAKER_01: Like a lot of the tools and technologies we internally develop for our AI agent lifecycle have come from those people who started off as agents for us.
• 18:21
  SPEAKER_01: And I think that continues to be the case.
• 18:22
  SPEAKER_01: Those are the people who best understand how the technology is impacting them and not just a person who might have been forcing a solution down them.
• 18:30
  SPEAKER_02: Sure, sure, sure.
• 18:31
  SPEAKER_02: So one of the things that um you know we we talked a lot about, and uh it's it's very important for any of these things and and for generative AI is is the you know the AI agents are only as good as the knowledge behind them and and search is the retrieval layer for AI, right?
• 18:47
  SPEAKER_02: For generative AI, right?
• 18:49
  SPEAKER_02: What is building and maintaining a uh a knowledge uh base look like at scale?
• 18:53
  SPEAKER_02: What is the foundational work that that you can't shorten regardless of how capable the model is?
• 18:59
  SPEAKER_01: Well, that's the what's becoming really exciting in you know in recent times is with the strength of the AI models that are being developed and deployed now, we're able to not have to wait for a knowledge base to be perfect or perfect access to it, but actually help compile and build the knowledge bases for the AI agents faster by being able to say, great, let's look at all of the historical data you have on calls and contacts, the entire customer journey.
• 19:22
  SPEAKER_01: Let's look at whatever segments of knowledge bases you have and access them into using tools like Ghost Rider through Sierra or some of the tools that we've developed internally and say, great, let's pull data from where you've got, pull it all forward and say, great, how does this actually map out for good customer experiences and recompile and rebuild a quicker reference guide for those knowledge bases?
• 19:43
  SPEAKER_01: So AI is helping us be able to actually utilize whatever disparate systems the clients have uh to build a more perfect knowledge base faster rather than having to get everything, you know, having to go into each of those individual systems and build it the the more legacy way.
• 19:59
  SPEAKER_02: Sure, sure.
• 20:00
  SPEAKER_02: No, that makes sense.
• 20:01
  SPEAKER_02: Um, you know, I think this has been a great discussion.
• 20:04
  SPEAKER_02: Thank you very much for uh joining me, helping us understand our audience understand why AI and custom experience isn't just a uh switch and flip, right?
• 20:13
  SPEAKER_02: And what it takes to get this right.
• 20:14
  SPEAKER_02: And to our listeners, thanks for tuning in to adults in the AI room.
• 20:18
  SPEAKER_02: Be sure to subscribe for more conversations about how AI is shaping the future of business.
• 20:23
  SPEAKER_02: Finally, thank you to our sponsor, uh Vector.
• 20:26
  SPEAKER_02: Uh, you can also find us at thelogos.xyz.
• 20:29
  SPEAKER_02: And um uh Michael, where can people find you?
• 20:32
  SPEAKER_01: Uh yeah, they can find us at uh ibex ibex.co.
• 20:37
  SPEAKER_01: And uh really appreciate the time and the opportunity today, Seth.
• 20:40
  SPEAKER_01: It was wonderful.
• 20:41
  SPEAKER_01: They had very thoughtful questions and uh yeah, happy to reconnect anytime you get a chance.
• 20:46
  SPEAKER_02: Sounds great.
• 20:46
  SPEAKER_02: Thank you again.
• 20:47
  SPEAKER_02: Bye now.
• 20:48
  SPEAKER_01: Thank you, sir.
• 20:50
  SPEAKER_00: AI requires more than a conversation, it takes action, research, and collective wisdom.
• 20:56
  SPEAKER_00: If today's episode resonated with you, we'd love to hear your insights.
• 20:59
  SPEAKER_00: Join the conversation and help us shape the future together.
• 21:03
  SPEAKER_00: We'll be back with more stories, strategies, and real world solutions that are making a difference for everyone.
• 21:09
  SPEAKER_00: In the meantime, be sure to subscribe, rate, write a review, and share it with someone you think would benefit from it too.
• 21:15
  SPEAKER_00: Thanks for listening, and we'll see you on the next episode.