Pedro Andrade, Talkdesk | The AI ROI in Contact Center Summit

AI is rapidly becoming part of the customer experience environment, but many organizations are still using it to automate isolated interactions or assist individual agents. Now, talk desk is taking a broader approach through customer experience automation or CXA, which uses multiple AI agents, enterprise data, and cross-system orchestration to resolve customer needs from beginning to end. Hey, welcome everyone to the CX Summit. I'm Bob La Liberte, principal analyst, joined by Zas Caravala, principal analyst and founder of ZK Research. Welcome Zas. >> Bob, thanks. It's uh this event's been great. So >> yeah, looking forward to another great session here. And joining us to explain talk desk strategy and its latest innovations and and how organizations can really translate AI investments into measurable business results is Pedro Andre, VP of AI at Talkesk. Pedro, welcome to the CX Summit. >> I thank you so much, Bob. It's a great pleasure being here with you. >> Absolutely. So, this is going to be a fun session and I wanted to kick things off and and obviously I mentioned this in the in the opening, but you've been positioning yourselves as this customer experience automation company rather than, you know, simply just a CCast provider. What does CXA mean and how is it different from the AI capabilities enterprises may already have in their contact centers? >> Yeah, right. It's a great question to start. So Bob um customer experience automation u CXI CXA is um what we defined as a operating model okay so it it's not a a category of uh technology um it's um it's the system that coordinates um an hybrid workforce of AI and human employees um then connecting the systems connecting knowledge And also connecting the workflows so you know when a customer calls in, reaches out to a brand uh they get what they need resolve it end to end independent of how complex the situation is or how many systems you need to get and put together to resolve that issue. Um so for customers uh this is not acquiring a technology just alone. It's a it's an operating model shift. Okay. It's not a it's not a tool purchase. Um majority um of the enterprises um already have some sort of uh type technology about either generative AI or or scripted AI. And so talk desk launched recently um a survey on on that with a get very interesting data. So around 74% of the enterprises already have some sort of generative AI. Um that's okay that that mean that gets the work started. You you can have some answers to your to your questions. It can eventually draft responses for agents. You can follow a script. But the problem is that it that situation alone without having a this operating model is that it stops the moment where you need um a judgment or an action that needs to cross multiple systems. Okay. So basically then you need uh to have a human to take the data from place A to place B so things continue working. So that's the change. CXA is a layer that keeps that work moving um across different systems and so you get your work uh done across systems, knowledge, humans and machines. >> Yeah, thanks you know Pedro. Um thanks for the update on on what CXA is and it's a great pivot for Talk Desk. Now Talk is though by and large one of the leaders in the context industry, right? and uh you scored very well on all the the different rankings and so talk about the relationship of CXA to the context in our platform. Is it a replacement force on top? How do you how does talk to us think about the the relationship of the two? >> All right. So different from other um situations or other companies, other providers that um we sell technology infused in the contact center. We hear that that term um very very often about infused AI actually talk when when we decided many years ago um on that we want to move from a human picking up an interaction a phone call or an SMS or an email and try to do that manually. Um we we understood that other contact centers also have the same the same problem and so instead of building a platform in fully uh infused on the contact center we built CXA as a platform as a different offer. So talk desk has two offers CXA and Cass and the CXA platform can be offered on top of other contact centers beyond talk desk contact center. Um because the reality uh Zus is that majority of the seats are still on prem and those customers that are still on prem uh for multiple reasons contract reasons because moving into the contact center to the cloud is still a big project they they can't afford to miss the opportunity of AI. So CXA works as well as a as a standalone on top of other contact centers to help relieve that pressure for the customers that are still on on prem that cannot uh get the latest and the greatest of AI on their onrem system but to be able to again to use CXI to relief that pain and gain start gaining those productivity um gains on top of an existent contact center. Okay, thanks for that explanation and I think the ability to to work with the legacy ones actually you know it really helps customers modernize. Now um you mentioned some of the research that uh talk desk has done. We actually looked through it in preparing for this and we saw an interesting data point that 98% of companies have deployed AI somewhere and that's probably going to be 100% pretty soon um somewhere in the customer journey but only 15% combine agentic AI with cross departmental orchestration. So why do you think that gap is so big right now? Well, the reality is that um adoption is easy. The orchestration is the hardest part. So um adoption is easy. You you can you can get to a uh to a provider, buy a chatbot, put it to run and it's easy. um the problem uh that you are mentioning and the reason why only 50% use um a genetic cross department and orchestrating um this journeys um is specifically for two reasons. Number one is that agentic orchestration requires AI to maintain um a context across all those systems. A solution is not just get my balance where it's only in one place. You are going to resolve a problem of uh a situation with your um with with one of your transactions that may require multiple uh interactions across multiple systems to get that problem solved. So majority of these steps are steps that require um an interaction between a a human. Someone is going to approve something. It's a world journey that that embeds multiple systems, multiple um humans involve it, machines involve it. And today that situation is fragmented that those workflows are fragmented. they are not um linked and connected to each other. So orchestration is really the hardest part. Currently Z you know this right we have still have companies that are information in silos when and when you ask them to to solve this end to end you will need to orchestrate across these multiple systems. There are those this is a um the the reason why um this requires the connections between multiple systems. Uh gets a blocker when um you reach customers that are still working with systems in isolation and then expect humans to connect the dots and to transport the data from one place to the other. Um and reality is that for example six in that report 64% of the organizations they already running some sort of uh AI agents for specific functions like for example I don't know uh billing or for example identity right it may sound that is coordination but the reality is that this end off between those agents still requires someone to go there and move the data from one place to the other >> um and I told you there are two reasons for that gap So for the 98 to the 15. Um the second reason is governance. Okay. Governance um connects and and then connecting to that report. Um it is more related to um an inter an enterprise readiness. Is this more related to an enterprise readiness than actually just an AI capability? And if you take a look at the the four uh top reasons why um people um to justify that difference of adoption is because number one compliance taking 15% of all the reasons um to for that adoption cross the predamental and this orchestration. Second with 48% of responses is about security. disconnected systems takes third place with 45% of the responses uh pointing that as a as the issue and with 44% are legacy infrastructure um systems that don't talk systems that are legacy still on prem um and so that was the kind of the four reasons z to you answer your question is you need to orchestrate that but uh beyond the orchestration comp governance including compliance security disconnect ED system and silos is being the top barriers and that differentiates those customers that are part of the 98% that deploy something versus the ones that are able to actually um harnessing uh those agents into uh one um experience that goes cross departmental and cross system. >> All right, thanks for your thoughts on that. >> Yeah, Pedro, I'm I'm I'm interested in this as well because it sounds like there's, you know, some organizations are a little bit more mature. Clearly you've seen a lot of this activity these organizations what works what doesn't work how are you at talk desk helping customers integrate those AI agents with all those systems of records right whether it be CRM or ERP or others and helping them to to get over this bottleneck get over these challenges that they're happening accelerate that time to value >> right so over time um talk desk gen created um a gazillion sets of um integrations over time because automation Bob is not it's not new right we we already have automation on IPR press one press two and you still get your balance you still file your claim um and the fundamental piece that was required to do that level of automation is that one way or the other some cases needed to put some of these places the solutions in place some of the some issues use right sometimes is that um building an integration requires a wall project that takes month with talk desk specifically on industries that we relieve that pain uh because all those integrations that you require to have an integration let's say with epic right that requires integrations with specific healthcare protocols or with majority of the banking systems if you are that talk desk offers that out of the box you have those integration out of that and allows you to with minimum configurations and setup allows you to have these agentic systems ready to connect to those data sources with quality uh reli with um security that is that is fundamental. So over years we created those connections, we created that relationship with those partners, with those providers uh that now we are reusing um especially then on on our verticals to accelerate the adoption and the time the time to value. >> Excellent. Yeah, that that makes a lot of sense and especially for those specific verticals where you've got those tight integrations to be able to accelerate that. And it leads me to another question I have for you because we know there's a lot of right the CX leaders other business leaders they're all under a lot of pressure to show that AI is creating value and in the in this space right it's about you know just not call deflection and maybe even headcount reduction when you're thinking about CXA what business outcomes should organizations expect and are there new measurements excuse me new measurements and new metrics that they should use to measure Oh absolutely yes and if I had $1 every time that I get this question I could retire myself today. >> So Bob majority of times I get this question is is going AI to help me to reduce my uh my ad count. People are just so so crazy about this and nothing wrong about that. But let me share some additional data uh that we also got in that survey. And I think that that kind of a needs to be used to change the way you need to think when you look at adopting AI. If you are just thinking about reducing count and and and do these savings, you are missing part of the story. Okay. So I'm going to give another example for um so let me give some data and some examples here. Um in in our report we we created four categories uh of different customers depending on their maturity level okay or their maturity to adopt uh to adopt AI. The top ones are what we call CXA leaders. Um, compared to the one the ones that are right be right right before uh these ones right below these ones. Um, they are getting uh much more um results in terms of NPS games than the ones that we call agentic scalers. So we have the agent scales and uh on the top the agent the CXA leaders um in terms of N in terms of NPS the difference is crazy. It's about four times. So you have 5% uh more better NPS scores uh for the ones that are agentic scalers. So they have some level of maturity. They are running a gentic AI in their contact center. They are deploying it in production. um but they are just not like the leaders that they are harnessing those agents to resolve end to end. The difference between those two is four times 22% for the CXA leaders in terms of NPS gains versus five 5% more on um for the agentic scalers um in that in that KPI. So what is interesting is that the cost per contact also improves but not dramatically as uh as NPS for example. So for the CXA leaders you get 57% um improvement on cost per contact versus 48% which is a moderate reduction in uh in terms of your uh between the C agitic scalers versus the CXA leaders. Um so that suggests that savings alone don't tell the whole story. they they they understate the the value that you can get with you can get with AI. So the other way then to look at that is to look at um retention and revenue >> and let me give you some date some some numbers. So for example, the customers that are using CXA to run predictive sh modeling, they are getting uh way better results if they are on the top tier of this adoption. 51% better results in a productive sh modeling versus the 28% that um the the the group of customers the agentic scalers right after um they are they are they are seeing also personalized recommendation. All right. 44% of the companies that are running personalized recommendation, they are seeing gains in um um if they are part of the uh CXA leaders versus 19%. So again, it's almost double um of the of the gains if you are increasing your maturity of your um your your AI um your AI adoption. And again there's um using it as a as an harnessing system um that connects those Asians humans systems and and knowledge. So in some um reality is that if you took a look at all the spectrum of companies that are adopting AI um it's true that only 5% can say that oh I have a clear way of measuring uh measuring impact but the reality is that 46% of the all CXA leaders they have the impact they have measured and it is a good impact across metrics that are just not cost-saving. They are red um reduction of uh churn and um increase uh increase their revenue through personalized for for example personalizer recommendations. >> Now Peter, I'm glad you're actually focused on a lot of the revenue generating type of metrics, right? we see a lot of cost cutting and and I'm curious within the customer base what's the typical time frame for these customers to start seeing meaningful ROI because I think a lot of companies really aren't they want to invest but they're not sure of when they'll start seeing the upside. >> Well, that answer is less of a pattern. Why? It's going to depend a lot of um where your current pains are. Um you may have a solution that can you can spin and put to run in a few a few weeks, maybe two weeks, you can put it to run uh to three weeks. And if you for example have a problem of um optimizing your assignment of the right people. Imagine that you are you are a company you are an insurance company. You you need you are in a season of renewal of policies. The biggest problem that you have is that you need to guarantee for a specific scenario of a customer. You need to have the right person to do that because it's sales. It requires a touch a personal touch. The AI job here is to connect the right people. So rout for example intelligent decision routing is one of typical use cases that may affect uh those uh a customer um in an insurance company. It can you we we you can you can put them to run in uh in a few weeks. um understand the business rules, understanding their business, put that into an in an energetic system and make guarantee that the user is not um forced to press one, press two, press five and hopefully it gets to someone that is going to help to um to to renew their policy. Or if you case you are doing outbound you need to make sure that you you you connect the right people depending on the right uh in the right customer profile. So other scenarios uh may take a little bit more time depending on what is the pressure the pressure point. Um you can run for through uh two weeks to four weeks. Um if you if if it goes more than two months maybe you are trying to bowl the ocean. there's so much uh to do that you should break the problems in smaller pieces and and maybe you are not attacking the right painoint there and understand so understanding first the what is your journey where are the friction of your journey is the first thing that we when we interact with a customer doing um consult um consulted service trying to understand where we can help them that's is the first thing that we Instead of answering yes, we can do whatever the customer asks, we do the questions first. We try to diagnose the pain points, the friction points and trying to see where we can automate. Sometimes it's not a voice system. It's not a a bot. Sometimes it's an operation on the back office that is breaking the whole experience that you have. This is where we we start. So you basically we are in a to answer it directly your question Z is this is a matter of weeks not a multimonth project something is going to be really wrong if you need to spend multi multimal month doing a a project in the CX space >> yeah now I suspected your answer would be it depends which it sounds like it is but I'm glad you time bound that within a couple months because I think that gives uh businesses some sort of frame of reference to work with now I want to shift gears a little bit uh here pedro into the workforce and talk just recently introduced your CXA operations uh center uh to manage both AI and human agents and I know workforce is really you know a hot topic right now and so when when you look ahead with what CXA is what's that operating model that CX organizations will work with is AI agents assume more responsibility but we still have to rely on our humans. Yeah, it's a great point and is one of the most one of the strongest beliefs that we have at talk desk. Um we believe in uh the hybrid workforce and so the CX operation manager is a role that emerges from that from this when you have um when you have imagine this imagine a convoy belt you are going to place machines and humans as the work passes through the the convoy belt this harnessing is the most important thing it's the that is what optimiz izes your operation. So the place where that shift is from having people that before they were doing the job now they are monitoring. So the the supervision for example shifts from building scripts into behavior monitoring. How is your harnessing uh your agent your machinery your hybrid workforce working? Are you seeing problems when um an AI ends off to a human? Are we losing something here? And what about when a human engages with an AI to complete part of the job? Are we losing something? So that behavior uh that behavior monitoring, it's important and exactly what it means. It means basically three things. guarantee that your nonhuman workforce is ready to go before you launch them in uh into production. What is the quality of this skill that you are about to launch as part of your team? It's like recruiting exactly the same thing. Do you do interviews? Well, in AI world, you are doing evaluations and that's a job for the CX operation manager. After you hire that um agent, you not human agent but uh AI agent. The second thing that you are going to do is exactly what you do with human agents. You are doing evaluations of performance. You are going to do observability. You are going to understand how is it performing. are what is the errors uh what are the error rate that they are doing. So and do you do course correction, you do training, you do um an optimization of an instruction. Uh that is again the job for the operation manager is not a technical skill but is a beh is a behavior changing or a behavior monitoring skill. And final finally is the last the last piece of this skill of this emerging role is understanding how your hybrid team is operating and how this is affecting your uh business KPIs. It shifts now you don't measure just alone ever gend you are going to measure how much time it gets to from opening of a problem until it gets closed. No, because average time just measures the time of an agent. But what about the rest of the process or what about the other systems? What are the people on the back office? They are not counted traditionally in those KPIs. But now that you have an hybrid workforce, you need to measure the efficiency of the whole process from the beginning until the end. Even if it takes 300 interactions to get solved. And Pedro, this is really fascinating for me because, you know, developing these new skills and in a lot of cases I often look at AI and refer to it as the time to comfort with the technology and so forth. And now you're talking about a kind almost a cultural shift of these supervisors having to manage these hybrid environments, learning new skills. So, how is TalkE helping them make that shift? Are you actually offering some guidance classes, things like that to help them accelerate and understand this is going to be their new role and these are the new skills that they're going to need? >> You know, Bob, I at the beginning I thought that it will be much much difficult. Um, reality shows that when you talk about bringing AI into the contact center, a first reaction that you get is scare. People get scared. People get fear. Um, that's fear is the first reaction is this is going to mess up big time because I see ship ship messing big time as well. Um so the question is how do I guarantee that this is not going to create a problem for me? So the the answer to your question comes very natural. It's a need. People didn't even know that they need a role in the contact center that is the CXI operation manager. The reality is when you present it as that is the response to their fears. So the except the adoption and the reorganization internally to remove people from previous roles into this role comes very natural. Um they from day zero they are involved in the the the the designing of the solution. So Bob, this is not like a traditional uh SAS uh sale where you you install the product and then here's the the the video, here's the training team. Actually, the customers are involved from the very beginning in co-development. So we work with them and as we work, we present these tools. We present here's what here's the the agent that we're just deploying. Here's the quality provided by this evolve report. And after the launch we are keep monitoring and they have access to all of that data. So for them it's it's a fundamental part of their journeys. We you don't need to have a specific specific action on that because it comes very natural. They customer are involved in those tools and in this role as operation manager from the very beginning. >> Excellent. No that's great. And I also this next question I wanted to ask you is kind of a follow-up. We've talked about it a little bit earlier on and you've talked about the value of it, but I know you've done a lot of specialized capabilities for verticals, whether it be healthcare, financial, right, insurance, retail, etc. Why do you believe that vertical specialization is going to be essential for that successful agentic AI adoption? >> Yeah. So it it's fundamental um because uh when you look at industries um their level of maturity um differs a lot um from from from one to the other. Retail is the most uh mature vertical um where for example around 24% um of the all the the retail organizations reach uh the top tier of maturity um for example comparing um to what the average in terms of maturity is of 15%. So this this tells you that um your um adoption um and the way you adopt you adopt is going to be very close to your level of maturity of those industries. The way to accelerate that maturity is by bringing um pre-built solutions that are preconnected uh to their systems. Don't expect to bring an empty platform and expect the customer to connect the dots, connect the systems and transform the organization on uh alone that that that is not going to work. So verticalization is important because it brings resolves part of the pain which is bringing the systems together, bringing the knowledge together and for that that is not just about integrations, APIs, it's not about instructions. This is also about having people on our side, specialists in each of those verticals that we can talk about. We know your market. We know how you operate. This is how um this um the this orchestration should work. This is what we have been seeing in other uh companies within your market. This is what works. This is what doesn't work. this is what it worth to invest. That differentiation for us is fundamental is and we invest a lot in technology and in people that knows uh those knows those verticals. So the whole goal is to have people process and product that helps customers to reach higher levels of maturity. >> Make sense? >> Yeah, absolutely does. Yeah. Yeah. And I wanted to uh you know finish up talking about the the the way customers can think about CXA as a way of transforming their organization. So I'm one of the the interesting things about CXA is it's extended AI beyond just using it for inbound services and answering calls quicker, right? and you do things like proactive interactions, you know, such as cart recovery, loan pre-qualification, collections, customer outreach, things that we historically didn't think of as a part of the service organization. And so when you when you think about that vision, how do you think this changes the organization? Does the contact center become revenue generating or does it merge with you know the CX organization and and talk about that? Yeah. So absolutely it it is a change um it is a change and is happening now especially because um we and the customers are seeing this transformation not at the lenses of only automate one use case but automate the whole journey. And guess what majority of the journeys are not just inbound. Inbound is kind of the last piece of a journey is when everything broke people call in. So the when you speak start talking about CXA automation orchestration you start uncovering those journeys. Come on let's talk about that journey. Where does it start? Where what what is what is the trigger of this? Oh people are calling because um they want to schedule uh their um to get their car serviceed. Hold on a second. Why is that? You know when the car gets needs to get serviced. You know when um an u an AC equipment needs to get serviced. Why don't you when is that part of the journey? And that's when starts you you know you start you you almost hear uh the gears changing moving in in the customer's brain when they start thinking oh yeah the reality is that is in another place somewhere in the organization. All right, let's bring them in. And then that that's when you start automating the the whole journey. So basically instead and instead of just waiting the call to want someone to get their car or the AC services, basically you have an AI agent that automatically verifies periodically what are the customers today that I need to contact that are going to get their car service or their AC service in the next x amount of time. And then they start outbounding those uh those messages or phone calls depending on the strategy and then they may not pick you pick up the call. They may call you back like 10 minutes later. But because you start the journey, the customer is already on that journey. So when you pick up the call, you know what this is all about because the context is shared across these multiple agents that are taking care of that. So it come it it comes naturally when you start looking at at use cases that are not isolated and that's where that pains me when majority of the AI thinking is about oh I have this issue I'm going to put a chatbot here it's going to answer those questions no man there's there's a reason why that is happening look at the whole spectrum look a whole journey and put your journey all in a paper and now start thinking about the automating the wall journey instead of just having a onepoint solution that takes you nowhere. >> Yeah, that's a that's a great explanation. Thank you so much and this has been an awesome discussion. Unfortunately, we are running out of time. So, Pedro, thank you so much for joining us. >> It was a great pleasure. Z and Bob, thank you so much for inviting me. >> Yeah, thanks Pedro. >> Yeah, absolutely. So clearly the key takeaway is that that next phase of AI and customer experience won't be defined simply by how many interactions it can automate. It's really going to depend on whether AI can securely and reliably coordinate people, data, and enterprise systems to resolve customer needs and deliver measurable business outcomes. Now, Talk Deck's evolution towards customer experience automation reflects that broader shift from isolated bots and co-pilots toward governed multi- aent systems capable of supporting the entire customer journey. Zas, thanks again for co-hosting and thank you to everyone for watching this segment of the CX Summit.