How Enterprises Are Moving from AI POCs to Impact (Full Transcript)

[00:00:00] Speaker 1: I Everyone how are we doing today? Good Pretty nice. It's amazing to see London being our global headquarters so many nice and good faces familiar faces here Thank you again for coming. We have a unique session right now because what we wanted to do is like, we see that like AI is getting adopted in the real world and who best to actually bring three like leaders in different industries to actually tell us more about that. This is not about the Level Labs, we're in Level Labs event, it's more about like what is actually happening in the world and that's what for me is like the most interesting piece of it. Juan, welcome, thank you for coming. Juan leads AI, generative AI, across BCGX. He's fantastic, we've been working together for over a year. We also have Terrell, who leads AI across Naturgy, 18 million customers globally, 24 countries, one of the largest utility companies across Europe. Oscar, thank you very much for coming as well. Oscar heads generative AI for deployment for Connecta, the sixth largest BPO in the world. Thank you guys again for coming. For me, it's important to start, like on Monday we announced our collaboration strategic partnership with BCG. It's been a fantastic honor to actually work together very closely over the past months. And it's one of the key things, Matti was talking earlier about the fact of going deeper and helping the world deploy AI and agentic systems, and that's what we think like collaborating Between both both companies is super key because you have the expertise across a number of industries Big small customers all of that stuff. We have the technology. So there's really good synergies in there Thank you so much for picking us and helping us Juan, what are the trends that you see in the industry?

[00:02:09] Speaker 2: Well, you know, first of all, I think we are entering into amazing times You know people are really and organizations are adopting gen AI and agentic AI We are seeing numbers about 40% in terms of adoption, but companies are doing basically three things. Okay companies are Implementing final users applications like, you know transcribers Summarizers for the day-to-day of the employees The second thing that we are seeing which is a heavy trend and 80% of the companies doing AI are doing is reshaping an entire business unit transforming the customer service Transforming the work the way their engineers work transforming the way their do digital marketing taking a process reshaping it completely and Identifying it and the third trend that we are seeing very preliminary, but also also with huge potential is the reinvention, you know, companies that are trying to find new revenue streams, even creating greenfield autonomous organizations from scratch. So these are the three things that are happening in organizations today. Huge penetration in reshaping to capture quick value.

[00:03:27] Speaker 1: I really like it. Oscar, like, what are you seeing, like, you lead one of the biggest, like, deployment for one of the biggest, like, BPOs in the world, like, what are the terms that you're What are the industries where you think highest adoption tell us a bit more?

[00:03:40] Speaker 3: So we're actually seeing quite a lot of transformation across the board industry wise. We're seeing a lot of work on Industries that are process heavy that are that have complex activities that they need to go through activities around Banking and In an insurance for topics around back office processes also a lot of engagement with customers on the front end and retail, quick service restaurants for order taking or for some of the processes that are heavy and repeated processes are also something we are observing a lot. From a region perspective, we saw a surprising uptick in Latin America as one of the hotbeds of innovation. It's been a lot of pressure from the customers coming to us first saying, hey, I see that this is happening, I want it now, with all the complexities also from a language standpoint. A lot of the models started very heavy in English with a little bit more of complexity on LATAM languages. So that's been an interesting evolution. And we are seeing a lot of growth also in Europe and in the rest of the regions as we go, with a lot of pressure on the compliance side. So we have a lot of ad hoc processes and ad hoc products that we can deploy. The reality is that enterprises need an enterprise ready solution to have in place, and that allows them to manage, orchestrate, and control whatever is deployed in front of their customers and what they can have.

[00:05:05] Speaker 1: It's so funny. Actually, I was thinking, and we're going to go off script all the time. But I was actually thinking, there is multiple velocities in deployment across the world. I would love to hear your thoughts, Juan, as well. Fundamentally, because what we're seeing is Latin America, India, specifically some of the countries in Latin America, India, Spain, UK, Germany, they're adopting Nordics, are adopting generative AI much quicker than many other countries, right? So you end up having the two tiers. Is it something that you're also seeing, Juan?

[00:05:38] Speaker 2: Yes, yes, we are seeing, and also, the things that they are doing are different. For example, in Asia, we are seeing a huge innovation in terms of automation. They want to manage the customer interface, you know? Huge adoption of customer interface management. In United States, focus is mostly in two main dimensions. Acceleration of their engineering departments, we see heavily that. And the second thing, it's augmenting the capabilities of their professionals. Like real-time co-pilots, solutions that support the telephonic agents or the chat agents. So geographies are doing this because they have different efficiencies and also different cost structure. It's very different, the cost per call in India, which is near 25 cents, to the cost of a call in US, which is near $3. So that also changes the equation of where you will invest the next AI dollar.

[00:06:35] Speaker 4: Yeah, 100%. Tarell, how are you deploying agents across the internet? Yeah, I mean, we've been having a kind of a, we think it's a unique approach because we started this process back in 2021, where AI wasn't even heard of. And we basically redefined our entire IT stack stack, and we started a transformation process with people first, because we wanted to be the best customer service utility in Europe, and that's still our goal, mainly because that's where we think our differentiator in the market lies. And so this has allowed us two things. First, is to have people on board from the beginning, because they know that we're in this transformation project and in this journey, and it's either we do it or we're out of the picture, right? And second is that our systems are AI ready from the beginning. So then what we did is we started doing a lot of pilots and POCs to make sure that we understood what we could achieve with the technology, and we know the right partners to go along on this journey, because that's also pretty important. So right now, we are on our beta, and we're almost ready to go into production and put this solution in front of our customer, which is something that we're really excited about.

[00:08:14] Speaker 1: One thing that we see a lot, and I would love to get your take, right, is the fact that there's a very, like, C-level wants to deploy generative AI across a number of areas, right? And you're saying you're now going live, you've tested and so on, but there's two sides of this equation. One is what we want to do as enterprises, as businesses. The other one is the team that we have supporting us as a business, right? How is the team transitioning? How are you helping the team transition so that they all are engaged with this big transformation that is happening in the business versus against it?

[00:08:49] Speaker 4: Yeah, I think it's really important to have C-level sponsorship on these projects. And when I mean sponsorship, it's not just because they want to have a press release and say that we are working with AI. It's because we really understand the technology, or the C-level executives understand the technology and where the value lies. That's pretty important. But also, to get everybody on board from the beginning was really important because people tend to see AI as a threat at the beginning, right? And the way that they lowered that nervousness around AI, I think it worked when you're working with it, you understand what it's capable of and what your relationship with it will be in the future. And that makes you a believer in the technology. And that's what our journey has been in the last 12 plus months. Makes perfect sense. You brought us a demo, right? Yeah, I did. Well, we're working with BCG and with you guys on, as I said, our beta test. I just did a very quick call because we don't have that much time. And our calls tend to be pretty long. But I have a demo that I want to show on how the interaction with this voice session will be in the very near future for our customers. Excellent. Let's play it.

[00:10:24] Speaker 5: I'm Eva, your virtual assistant at Naturgy, and I'm here to make everything easier for you. By the way, we're recording this call so we can improve every day. And if you want to see the privacy policy, it's on naturgy.es. Now, let's go there. Tell me, are you a client?

[00:10:42] Speaker 6: No, I'm not a client. Look, I'd like you to give me your information on how to install solar panels in my house.

[00:10:50] Speaker 5: Before we continue, can you tell me your name?

[00:10:51] Speaker 6: Sure, I'm Terrel.

[00:10:58] Speaker 5: Delighted, Terrel. I'll tell you. First, a study is made of your house to see if it's viable. Then you get a budget and if you accept it, the contract is signed and Naturgy takes care of permits, installation and legalization. If you want, you can also hire a special fee for self-consumption with compensation of exceeds and lower prices at night. Would you like me to explain that part to you or do you prefer to focus only on the installation?

[00:11:25] Speaker 2: No, that's all. Thank you very much. Next day, we had a human-like voice, the one that you are hearing in Spanish.

[00:12:00] Speaker 4: That's one of my... A person from my team at NaturG, it's her voice, and we did it just to test, as Juan said, the voice calling capabilities of 11 Labs, which is pretty impressive. And as... I mean, it was really exciting. I didn't do it because I didn't want to make a very long conversation, but this demo is also grounded on our system, so it allows us to do things on behalf of the customer. And because of how our systems are built, everything that the customer can do in his private area on the web page, the voice agent will be able to handle as well. So that will be pretty exciting to deploy this and achieve, We think a very, very impressive customer service.

[00:12:54] Speaker 1: I think like, so like Antropic put it really well, in Davos, where like last year it was all about proof of concepts for the big enterprises, right? This year is about a proof of impact, right? It's like, how are you generating actual value for both your customers and for the organization? So I think there's been like a massive shift like towards like showing it, because everyone is testing it, but like who has it in production? how much actual ROI you're getting on it, right? What is the biggest issue, Juan, that you see with organizations trying to deploy agentic systems?

[00:13:30] Speaker 2: Well, there are three things that organizations are challenging today. The first one is focus. They don't have focus. Every time that I land in an organization, I see that they are doing 120 POCs, but impacting zero dollars in the P&L. You don't need to do 120 things. You need to do 10 that impact the P&L. The second thing is, you know, people think that this is a tech project. This is a complete business project. Tech, and as we like to say, 10% of the effort in a project and the impact relies on the intelligence of the model. 20% on the IT infrastructure, platform, et cetera. But 70% of the effort and impact will come from changing processes, changing people, and making this capillaries along the organization. So this is the first challenge, focus. The second challenge that I see, sorry, the third challenge that I see is not focusing value and focusing technology. you shouldn't just think about which is the best LLM, which is, that's secondary. What you should pursue is value, because today technology is very mature. We hear that voice, for the ones that understood Spanish, that woman took a breath in the cold. She basically laughed a little bit. So we are in a level, in terms of technology, that it's already sufficient to start bringing heavy impact. Let's take that as a priority instead of just improving one extra point in accuracy. That's my view.

[00:15:22] Speaker 1: I fully agree. Oscar, what are the biggest challenges that you're seeing when deploying? Because you're in this really interesting space in the middle between companies and technology overall, right? So what are some of the challenges that you're seeing right now?

[00:15:35] Speaker 3: So taking on Juan's point, I think the element around focusing on the value and how you are transforming the process of the experience that you are creating is the first big hurdle. Technology, it's been proven, works. You have significant ways to implement and put in place solutions that really feel natural, that really are driving the conversation. Ensuring that you are focusing on the right part of the processes, on solving for the friction points that a customer might have. And then addressing all the different silos that an organization has to actually get you there is where usually a lot of the time is taken. And then also creating that feedback loop or that intelligence to listen, understand, share an intimacy with our clients to explain, hey, we are observing this. This is impacting your business. Let's focus on how do we change it this way. And iterate much faster than what we were used to iterating back in the day with a traditional service where you have to train the agents. You have to spend time getting them ready to follow a new script. And then you can accelerate that feedback loop and training capability. And that's an area that requires some change in the behaviors of our teams and the client teams. But it's really the point that can make it or break it.

[00:16:52] Speaker 1: And do you see some interests that are much more forward-looking towards this technology stack that we're all building?

[00:16:59] Speaker 3: We've seen quite a lot of interest in, as I was mentioning before, on industries that are heavy on some of the processes in back office, or industries where there is a lot of interaction to provide basic information to customers, or where they need to take several steps to complete the activities. Tara was mentioning on the utilities industry all the processes that you can do in your own personal page. That's a perfect example of right now being 100% reactive. It has a lot of opportunity to transform into being more accessible and be more proactive in how you engage that area.

[00:17:39] Speaker 1: That makes sense. And I'm actually quite curious, like, how did you pick your stack, your tech stack? Because there's plenty of providers out there, there's like 11 labs, we power pretty much 90% more than 90% of anything that is AI voice in the market, but there's plenty of solutions out there, our agents and a bunch of other ones. How did you pick your solution, your stack and everything?

[00:17:58] Speaker 3: So our main guiding principle is being open and interconnectable, because at the end, the reality of an organization like ours is we are working with the technology that a lot of clients already have. So the theoretical, I can go and implement whatever I want into a client environment is nice to have in a lab, it doesn't really work in real life. And then the focus is looking at best in breed solutions for each one of the elements and that modularity of how do we put everything together to deliver the value that we are looking that we are looking for, starting from the process and the value we're trying to create, not so much the technology. Technology has to be an enabler, is critical, but the focus is what solves best the problem that we are trying to get into.

[00:18:40] Speaker 1: Okay, that makes sense. You put us the demo as well.

[00:18:43] Speaker 3: I did as well, yes. What you'll see is a brief demo of a financial customer in a financial client in Latin America. The focus in there was to transform one of the processes that is taxing for human agents, which is cash collection, which is basically calling people and saying, especially the first 30 days post delinquency, you have not paid for your credit card. We should negotiate a payment plan. And what we built or what we worked on building there is an agent together with BCG that would help us turn that not only into a notification, but making sure that there is a negotiation that we can get to execute. And it actually is performing amazingly well. We have managed more than 350,000 calls. And the reality is that it is performing at the level of the best agents that we have working on those services, focusing on additional business that the customer, the client in this case, couldn't address before implementing this.

[00:19:39] Speaker 1: That's excellent. Well, we'll talk about ROI and performance and everything in a second, but let's see the demo.

[00:19:48] Speaker 7: Awesome. I would be able to pay 50% tomorrow and the other 50% the day after tomorrow.

[00:19:59] Speaker 8: I understand Ana Maria, so you would be able to pay 50% of the debt tomorrow, February 6, 2026, and the other 50% on February 7, 2026? That's right. Perfect. Perfect, William. With the $200,000 bond, we would distribute $14,881 to the mortgage credit and $185,119 to the free investment credit. Do you agree with this commitment? Yes, I agree.

[00:20:37] Speaker 2: Nice. Let me say a fun fact about this, I think it's an interesting project, right, which is when we were making it, I mean, you know, collections. Collections is a hot topic, you know, not easy calls. And you would assume, hey, people are going to hear a robotic voice, which is not the case at all. And by the way, 11 Labs was not only the best Colombian voice, but also one of the only ones that exist in the market, okay, without a cloning, okay. fact is that the NPS of the experience of the depter improved. So you would expect a much lower what we call RPC which is real connection because a body is calling. That's not the case and Oscar can give more details here but when you look at the funnel and as Oscar mentioned the body is performing equally to a human.

[00:21:37] Speaker 1: We see it a lot in Brazil, for instance, where it ends the conversation, people end the conversation talking to agents saying, vaya con Dios, right? It's the equivalent of, like, have a good day, but in a very specific way for Brazil. And for me, I'm always amazed, because fundamentally, I'm like, really, people forgot that they were talking to an agent, and they're thinking they're talking to a human, because that's the natural thing. I believe like once you get that you've made it right you've made it in the sense like People have exactly the same experience or better that they were having before and tell her how are you thinking about the ROI? Success metrics all of that stuff.

[00:22:15] Speaker 4: Yeah for us as I said the main success metric is to really have The best the best customer experience and the happiest customers in a utility Which is which is a big ask because nobody loves its utility company So, that's our driving force behind the project that we're doing. And we are really excited about it because if you think about, we have over 800 agents in four different locations across two continents serving our customers, which do a terrific job. But it's really hard to have a homogeneous experience for our customers depending on who are they talking to. So we feel that if we can handle most of our conversations through an AI agent, there are our more day-to-day things that we treat our customers. And we can give them a very homogeneous experience, which is, I think, really impressive what you can achieve with AI. And leave our agents to handle the most important and most complex parts of the operations with the customers, then we will achieve a much better customer experience from what you get today. And as Maddy was saying at the opening, reduce wait times, have 24-7 support, multilingual support for our customers, and so on and so forth. So that's how we are looking at ROI. We're not in a very financial type of approach it right now.

[00:23:58] Speaker 1: Juan, are you seeing similar metrics, KPIs, success, all of that stuff across different organizations, different industries, or does it vary a lot based on the industry?

[00:24:09] Speaker 2: It varies a lot based on the domain use case and industry, and also the strategy that the company has. It's very different to have a cost optimization strategy than to have a growth strategy, and depending on the market, you will decide where you want to invest your buckets. Today, most of the projects are focusing in cost optimization. Having said that, the business case becomes much more positive when you focus in top line impact, of course. A lot of people think that agentic is about, let's cut here, cut there. In reality, the way you should think about this is imagine scaling your organization without any limitation. Imagine contacting any, you know, every person living in London today with no limitation. So that's not cost efficiency. Yes, it's productivity, but it's about top-line impact. It's about scale. Agentic will bring into companies the scale that they didn't have and the speed that they didn't have because we were focusing in human capacity, you know?

[00:25:21] Speaker 1: Like, we were talking backstage about agentic organizations, fundamentally, right? And you were telling us, Juan, like what you're seeing right now in the market, like, give us like 30 seconds.

[00:25:31] Speaker 2: Yes. So this is related to the first topic that I was saying, which is some companies are really creating a green field where autonomous organizations are being created. And we are not very far away from seeing a first autonomous telco, or a first autonomous sales distribution channel for a utility, or even an entire autonomous BPO, right? The limitation is not in the tech. The limitation today is mostly in regulation. For example, what Terrell was saying, the experience is delimited by regulation. But from a technology perspective, you are fully capable today to create this type of organizations, and we will see it very soon.

[00:26:16] Speaker 1: Oscar, what's the future of BPOs? Probably what everyone is thinking here, how are you seeing all of this? And how are you getting ahead of it? And what are the trends? And all of that stuff.

[00:26:28] Speaker 3: So it definitely changes the positioning that we have as an organization. We traditionally were working in an industry that was based on providing seats to address processes that companies wanted to outsource. that moved more into outcome-based activities, so focused on, hey, how much of that can you deliver and I'll pay you for this? And it is focusing up the chain on how do I get to the outcomes of the business, to Juan's point, more focused on what is the top line and what is the benefit for the customer as we evolve. We see it as an increasingly tech-focused activity, so we need to ensure that we not only operate the activities that are happening, but we also are leveraging the best inbreed that we can find. And also maintaining that hybrid mode of working we still have. And for the next years, we foresee that there are parts of the processes, parts of the activities that we will want humans to be part of. So we want humans to be managing, addressing, engaging with customers in certain situations where you want to have that closeness and that empathy. whilst there are many cases where you can automate and remove that part of noise from the activity by making the whole process hybrid. And that orchestration, that positioning on, how do I manage the right approach? How do I make sure that the customers of my client are getting the right support in the right place at the right time through the right channel? Optimizing the channel is a critical part of where our industry will be evolving.

[00:28:01] Speaker 1: Because we go multichannel across the board, right? I mean, I hate when a company is forced you to actually interact with them in one single channel, and that is a specific channel, right? And I think that's one of the big benefits of a genetic system, is that essentially you end up plugging in multiple channels and be where the customer is, versus be only where you want to be. Very briefly, Tero, what's next for your deployments? Anything that you can share, like you don't have to reveal any major stuff.

[00:28:27] Speaker 4: Yeah, so our focus is to deploy this on second quarter this year and we're, as I said, we're really excited on putting this in front of our customer and seeing their reaction. And also, I think that we're starting to look ahead on what's coming next to us because we were really focused on this project for the last 12 months and we think that there is a lot of other places in the organization where agentic AI will help us deliver better experiences for our customers.

[00:29:06] Speaker 1: Will you focus more on the next ones on internal optimization of the organization? Will you focus much more on customer optimization?

[00:29:13] Speaker 4: I think both, Carlos, and I mean, the possibilities are almost endless right now. But we want to have focus and impact on what we're doing. So those are the two main criterias that we're following on how to approach our next projects. That's excellent.

[00:29:36] Speaker 1: Well, I mean, for me, it's been great having you here, guys, seeing the impact that your organizations and the deployments are having across the world. What we wanted to do fundamentally is just share a different perspective of what is actually happening. Because sometimes what we see is we only are in our own vertical, our own industry. And I think it's important to actually see what other people are actually doing. So I really, really thank you guys for coming here on stage and being with all of us. It was an absolute pleasure. Thank you, everyone.