Autonomous CVM in Telco: the What and the How
Why Listen
Silvia Gomez Dominguez brings 20 years of CVM and customer experience expertise from Optus, Vodafone, KPMG, and Westpac. She co-authored the Customer Value Management Body of Knowledge and founded the Women in Customer community.
Autonomous CVM is on every telco’s roadmap. But most teams stall between buzzword and build. The path from definition to reality is shorter than you think. So what are the 5 moves to make autonomous CVM real?
5 Tips to Build Autonomous CVM in Your Telco
1. Define the Outcomes, Not the Process
AI optimises objectives. Hand it a process map and you’ll automate yesterday. Hand it an outcome with guardrails and AI will find paths your team never considered. Tip: for every CVM use case, write the outcome and three constraints – then let AI find the steps.
2. Layer Your AI – Predictive, Generative, Agentic
Autonomous CVM isn’t one technology. It’s three layers working together.
- Predictive = scoring propensity and risk
- Generative = mining verbatims, writing content
- Agentic = executing actions, talking to other agents
Tip: map your top 5 CVM use cases to predictive, generative, or agentic – pick tools per layer, not per vendor.
3. Stand Up an Orchestrator Agent
The unlock isn’t more agents. It’s one orchestrator that talks to all of them – data, marketing, digital, contact centre, finance, legal – end-to-end across the customer journey. Tip: list the 5 systems your CVM team queries weekly and design one orchestrator to query all five.
4. Pair CVM With the AI Lead
CVM as a solo strategic function is closing. Whoever owns AI in your org is now the gatekeeper of your priorities. Don’t lobby – partner. Tip: book a recurring biweekly with whoever owns AI strategy. If no one does yet, pitch the role into existence.
5. Run Risk as a Living System
AI is non-deterministic – the same prompt today may behave differently next week. Launch-time audits are useless. Risk monitoring becomes always-on. Tip: set up red-amber-green monitoring on every AI use case, with humans in the loop for any decision touching money or churn.
Episode Highlights
- Autonomous CVM optimises objectives, not process steps
- Three AI layers in CVM: predictive, generative, agentic
- Orchestrator agents talk end-to-end across the enterprise
- CVM head must partner with AI head
- Hybrid operating model beats centralised or decentralised
- AI agents filter brands at the awareness layer
- Customers may be less happy with objectively better choices
- Non-deterministic AI needs ongoing risk monitoring
- Compliance is the foundation of customer centricity
- Replicating today’s playbook with AI wastes its power
Instead of trying to use AI to conduct the steps of your current processes, rethink the processes using the new technology. – Silvia Gomez Dominguez
Recommendations for Growth as a Professional
Start with the why before any AI use case, define what you’re trying to achieve in CVM. Set the KPIs that align with that why, and let AI find the how. Don’t replicate today’s processes – rethink them with the new technology. And learn from outside CVM – from legal, admin, anyone already using AI day-to-day.
Final Words
Autonomous CVM is no longer a future concept – it’s a build challenge. The teams that learn to set objectives, layer their AI, partner with the AI lead and treat risk as a living system will run circles around the ones still stuck in pilots. Start small – but start now.
TRANSCRIPT
[00:00:00] Silvia: Probably one of the biggest mistakes would be just to replicate what we do today with AI. If I want to grow, let’s say a customer value 2% every year, how can I do it differently? And instead of trying to use AI to conduct the steps of your current processes, rethink the processes using the new technology.
[00:00:24] Exacaster: Welcome to CVM Stories, the podcast on customer value management. Together, we explore how companies can be more successful and the customers happier through the use of latest customer value management techniques. Learn key commercial and analytical insights from telecoms, retail, finance and other industries that drive CVM forward.
[00:00:42] Egidijus: Hello, I’m your host, Egidijus. Today our guest is Silvia Gomez Dominguez, a customer experience and customer value management leader working over two decades across telecoms, financial services and consulting industries at companies like Optus, Vodafone and KPMG. She is also the co-author of the Customer Value Management Body of Knowledge Book, and the founder of a brand new community called Women in Customer. Today we are going to talk about something that is going to change CVM forever, Agentic and Autonomous Customer Value Management. What happens when AI agents start talking to AI agents on behalf of our customers? How do our CVM processes change and where should we start investing today when so much about tomorrow is still unknown? So let’s dive in.
[00:01:48] Egidijus: So welcome back Silvia. It’s the second time when we are in our podcast. Thank you for taking your time to come and share your insights with us. And, um, before we dive into our today’s topic about authentic CVM and all the genetic impact into customer value management, I would like you to briefly remind our audience about your perspective and your career, how you end up in CVM.
[00:02:24] Silvia: Yes. Thank you so much for having me here for the second time. Yeah. It’s good to remind the audience about my experience and why I’m, I am so interested in, in CVM and customer value management overall. And I have spent approximately 20 years of experience just at the intersection of the customer and the technology, that human side of the customer and that technical side of the tools, the data and AI. And it’s been mostly to design customer strategies that fuel growth and retention. I had the pleasure and privilege to work with the big four. Management consulting firms spend a lot of time in telco, with Vodafone, with Optus, and also banking like Westpac, for example. I am super excited about being here today because AI is one of the hottest topics at the moment. And when we co-author the CVM Body of Knowledge, we didn’t have that big explosion of AI and AI in the day to day of the workplace. So I think today it could be a great opportunity to talk about that. And also, just one of the latest achievements that I’m very proud of is I recently launched in International Women’s Day, the women in customer community, because I believe there is this need in across the board to get women and men equally represented in the leadership roles when it comes to bringing customer centricity, customer experience, customer value management to life. So yeah, super excited with this side project of mine that is having a lot of traction in the early days.
[00:04:20] Egidijus: Uh, I heard that it was a super successful launch. Could you, uh, tell, is this community available only in Australia or is it like an international one?
[00:04:32] Silvia: Yeah, it’s available internationally. There are different ways of engagement. And yeah, you’re not the first one asking me this question. Um, because one of the beauties of social media, LinkedIn, and having worked in global brands is that you connect with people across the world. You can join just for, let’s say, the newsletter. There are three types of initiatives. The one related to connect, which is mostly newsletter events and roundtables. The second bucket is about win, which is providing like templates, frameworks, or even coaches that help you with your day to day at work. But also, if you are looking into what’s your next role and can help you to prepare for opportunities. And the third pillar is about give back. So the community has a number of leaders across the board. Like there are different levels of experience. There are very senior people from the C-suite to very junior people that probably are in their first few years of experience. And there are also like recruiters that have joined or coaches that have joined as well. So it really was an unexpected success. I knew from my network that people were interested, but there are members that I didn’t know before and I have met through this community. So throughout connect, win and Give Back, we are starting to get some traction. And one of my objectives is to scale up. So that is ongoing. Having big momentum by the end of the year.
[00:06:15] Egidijus: So, we will wish you the biggest success in helping the community to grow. Now let’s dive straight into the hottest topic. So last time when we were talking with you, Silvia, we were talking about a customer love concept, which is like focusing on customer perspective and aligning this perspective with business needs, how we grow it. But now things are changing Now, more and more people or customers will start interacting with us as businesses, not directly, but through their personal agents, through their open claws or whatever tools they will use. And the question is, how does it change everything? Is customer love still kind of a good framework? How do they interact? You know what is changing. Could you walk us through this?
[00:07:19] Silvia: Yes, absolutely. So I think you’re absolutely right. The way that the the world operates today is different and much more AI and AI driven than a couple of years ago when we had our first podcast together. And I think in that assumption you made, we have somehow like press the forward button as well, a few years in the sense of making sure that all the risk and regulatory considerations have been taken in place, and that there are agents that are risk safe, right, for the businesses as well as for the customers. So if we imagine we are in that situation and only the safe agents are in place. If I think about how as a consumer, I will use it, and probably this is how other people will use it as well. I think we need to consider like the traditional three layers or communication layers from a brand perspective. And when we are talking about buying a product, we are talking about the lowest level in the funnel, like just conversion. But before we get there, we need to think about the why of the brand, right? Like for example, if we think about Apple, why they sell is more important to customers than the specific product. So those customers connect with the brand in the. Purpose in the values.
[00:08:51] Silvia: For example, if I want to buy a new phone or I want to get a new mortgage, I could use my agent, my agent in a way that I do not ask them to buy me the best mortgage or the best phone. Looking just at objective criteria like price, for example. But I call Subselect or ask questions such as what are the most sustainable brands that cell phones? Or what are the most risk safe financial services that can offer me a mortgage? So in that high level of the brand, like when you drive that awareness, the consumer can press select and if you like minimum criteria before making a choice. And that minimum criteria is, as I said, aligned to the values. It can be aligned to previous purchases like look at my purchase history and out of all the banks that I have engaged with, or all the telcos that I have engaged with Pre-select today, what is the best phone, for example? So that awareness level is what the consumer needs to choose, whether how important that is. There will be people that are just looking to get the cheapest, or there are people that they want to buy from specific brands that they feel more connected to. So they can do that in that awareness layer.
[00:10:20] Silvia: The second layer is about why we are different, right? It’s like the preference consideration. And again, in these two levels, I think the consumer can again say I want sustainable banks, but that are very focused on do not rip off customers, for example, or that give me a number of Additional values that help me to preselect the the number of options. And then when we get to the last layer, the conversion, then yes, the agents can play themselves, right? They have all the criteria, like for example, in a phone, how much does it cost? For how long is it going to last? What’s the memory that I need? All those objective characteristics play a critical role in the final selection. But if I summarize the awareness and the. The preference selection is more a conversation that each customer will have with the agent. And then obviously they learn. So it’s not something that you need to be doing every single time, but that helps you to pre-select the options that you want to be presented. And then after that, the from the available options, you can rely on your agents delegate with some controls so that they purchase the phone or the or the home loan or whatever that is.
[00:11:45] Egidijus: So if I interpret correctly of what you say, it’s that the brand trust will become even more important because this will guide a major decisions, you know, kind of to become the customer for the first time and etc. because kind of, uh, um, if the brand trust is low, even if I delegate a decision for agent for my open agent to find or limit the options for me, it’s like, it can definitely recognize that the brand is, let’s say not too much trustworthy. And I can fail really early in the funnel.
[00:12:37] Silvia: Basically That’s right. You wouldn’t get those options even presented to you if if you do not meet the minimum criteria, like for example, a very good prompt could be, I want to get a home mortgage. I don’t want to bank with any bank that has been in the news because of, I don’t know, trust issues or integrity issues or risk breaches issues in the last five years. So then the options that your agent will present to you will filter out those financial services organizations that have those issues. And as I said, the AI, by definition, learns a every, every time you use it. So it starts to help you to provide shorter prompts because they somehow those agents, those personal agents get embedded. What’s important to you? You could call it value. You could call it a very minimum, but restrict your options when they present the the purchase option so they can just buy for you, like even flights, for example.
[00:13:51] Egidijus: Let’s go one, um, uh, slightly down to the funnel. So for example, I personally, I am super easy to upsell like the person when I walk into Apple store, you know, I never purchased the device I need, I always purchase the device that looks the most amazing one. And then I also purchased some add on, uh, things that, uh, I might need or I might not need, you know, kind of. So all this, um, you know, experience when, uh, I am upsold it’s a very kind of easy process, you know, I walk into the store And and and it happens. Now, if I ask my agent to choose a phone that I need, it will never purchase, uh, the, the best option because I am not so good at photography and I will not, uh, become a photographer if I take the phone with the best camera and so on. So it doesn’t have, let’s say, these images that pop up when you are making the decision of what to purchase. How does this process also change in this gigantic world?
[00:15:12] Silvia: Yes, I think that’s a great question as well. And I think that the prompting and the if you like, conversation that we will have with the agents change as well, um, I don’t think it’s going to be in the same way that we do it today. Like we buy maybe they have the fanciest phone, if that’s a way to put it. For whatever reason, maybe because we believe we will need it in the future, even if we don’t even know how to use all the features today. But then the conversation it will. It won’t be about I want the fanciest font at all. It will be more about. I use the phone on the weekends to take photos and show me how to take the best quality photo for this moment. Like. And the phone will look at. Through the agent. We look at the light, the location, a portrait of distance so you won’t need to be as expert in terms of maximizing the features in the same way that today you can tell your city or even in in Google apps or any platform almost today. You don’t go now and schedule a meeting. You ask Siri or your agent to set up time for this meeting with this person, and even they send an email or reminder.
[00:16:44] Silvia: You will talk to your agent so that it takes the photos for you. In the same way, when you are looking at buying it, you will ask or talk to them about what’s important to you. And although sometimes it’s difficult to put in words why you buy the latest device, or why you buy the all the accessories that come with it, I think there are different segments of people. Some of them will do that. Some of them will go with the cheapest. That lasts probably longer because they don’t want to spend that much. But all those, let’s say less speakable, a decisioning or rationale that happens in our head. We will be able somehow to have those conversations with the agents in a very easy way. And not only that, but also we will rate the experience of that purchase, right? So for example, if in a normal situation today, you will go to the store and will get the the latest Apple phone as well as all the accessories and in the future or with the agent, you get not the latest, but one maybe from last year or two years ago. And the accessories, you really don’t need them. And in this case, maybe not the fancy. The fanciest one. You don’t need that.
[00:18:10] Silvia: You do that other one. Then when you start using it, even in that moment, maybe you are not as happy. But if you, let’s say, okay, I trust the agent. I want to be objective. When you are using it, you probably are not as happy and you will probably replace it even sooner, okay. Or may start calling the contact centre because I don’t know. It’s not working as you expected. And so I genuinely believe that this new technology will enable to address the customer needs and deliver those customer needs in a more efficient way, more focused on objectives, not on the steps or the process. And just because of how it learns through feedback and iteration, it will become better and better. And it will be very personalized because the feedback, the decisioning criteria will be different to each individual. For some people, they will probably write them very well. If they are saving maybe $500 or euros, and other people will rate them very highly if they get the best photos with the lowest effort. So yeah, I think it’s going to be better. It will take a little bit of time, but I think it will definitely address all the customer needs in a better way than what we do today.
[00:19:35] Egidijus: Mhm. So I would like to put one more touch from the customer perspective. And then I would like to to move to the business perspective as well. So, um, many of our organizations, uh, they spend a lot of energy and, and sometimes they even differentiate themselves, uh, through the help of the customer to experience the products. So for example, from the telecom perspective, if you sell a new device, uh, you want to help customers to adopt this device fully so that they are like training programs, uh, help out sessions when you just, uh, Come to the store, they set things up for you. They kind of teach you to do some tips and tricks, which was kind of the the way to differentiate and to, to build the trust with the consumer. Now a lot of, of this is moving to, you know, ChatGPT, Gemini or, uh, cloud or whatever tool you use. And here is like, uh, there is an intermediary between customer and some agent that helps you out with all these questions, you know? So, uh, how, how does then, uh, kind of this switch moves the business perspective. So if customers take more and more ownership of, you know, learning about, uh, the products and tools, uh, to AI agents, how does it changes the perspective of the organization, how they enable customers to do it better, you know?
[00:21:28] Silvia: Yes, I think there are different models to be successful in doing that. And also the different big brands in the AI space. Like for example, if I think about Nvidia or Apple, they are integrating a lot, a lot of that AI technology in their devices, whether those are the chips or the iPhone devices. So to some extent, you could argue that for those players that are integrating the intelligence in the device, you could argue that that could be part of the training. Anyways. So those workshops and I myself attended to a few of them on the year. I bought my first MacBook, for example, to get like my way many years ago to understand how it was different to Microsoft, for example, you could design those workshops to help the customers to interact with the AI agent, to customize the AI agent to what they need in the day to day. So I think there is still an opportunity to help and different customers in that process and use that as a difference or a differentiator. Then there are other brands, maybe also in the technology space that are less about including the AI technology per se in the device. Like for example, the Microsoft, um, um, laptop versus. And instead of doing that, they are investing more in the agents workflow or the applications on top. But equally, I think you could organize sessions, customer engagement, customer support to help the consumer to use those applications and agents for the purpose that they need.
[00:23:26] Silvia: And so how how to select and how to decide what’s the best model. I think in these days where we have so much technology evolving so fast and with new findings happening every every week, I think probably one of the biggest mistakes would be just to replicate what we do today with AI. And I think probably more successful organizations will be those that rethink their, the, their processes. And those could be like the core processes of a bank, a telco, a retailer, and also not only the core processes, but also the support processes like finance and other things legal, for example, and then make very conscious decisions about which processes they want to reinvent using AI and how in terms of maybe more than how first, like what is the outcome or the impact that is going to have? Is it going to be about growth? Is it going to be about efficiencies? Um, and then once they are clear on those goals, it’s a matter of establishing the, the plan and the resources, the different operating model to continue to differentiate and growth using AI more as an, as an enabler that is embedded in the, in the new operating model.
[00:25:04] Egidijus: So if we would kind of continue this path and let’s think, how do these processes change in customer value management area. So, uh, in customer value management, we think about programs like acquiring customers and on boarding them, retaining, upselling, cross-selling, etc. and we kind of combine them with user journeys. We combined with marketing communication, which some of this is outbound, and then some of it is like inbound, uh, in our web or mobile app, etc.. And how do you see the agentic processes here? Is it like, uh, do we have outbound campaigns, inbound campaigns, and then agentic campaigns? Or is it like something else? You know, so what should be our thought process about it?
[00:26:03] Silvia: Yes. Um, I think it’s a great question and probably we can align on some definitions for, um, in particular the audience, because every person might have a slight different understanding of the AI or a genetic or generative AI. And so the way I understand it, understand the whole thing is AI as a very global term that includes any artificial intelligence model or method. Traditionally, before this big ChatGPT and probably here before the 2020 223, I think everyone referred to AI more from a predictive side of things, mostly like those propensity models that will help you to understand what’s the best next product or action or offer to send to a customer, or equally for retention. It will help you to identify the highest risk customer cohorts through segmentation and classification methods, even neuronal networks at the time. But no one talk about them in the way they do today. And then the second very important packet that in customer value management can be used extensively, is the generative AI side of things in customer value management, it is super helpful to understand from customer verbatims what is happening. Instead of reading all the verbatims and then spending a lot of time trying to make sense of those verbatims, synthesize them, and create trends. With generative AI, you can have all those analyses done like super quick, right in a few seconds, and then you can look at the trends.
[00:27:54] Silvia: So it’s much more informative and it truly helps you to, to make them more actionable. And then if I go to the, um, I also, sorry, I forgot in the generative side of things, the other very important benefit from a customer value management perspective is in the content creation. So it’s not just that the LLM can help you with the text. You have great video or visual content creator tools that can help you to produce those campaigns much faster. And then the last one that I think is important to talk about is the agentic side of things. So in the Identic definition is not just about predicting. It’s not just about giving me some insights or even language that sounds good. It’s also about executing actions. So how you execute those actions can be in different stages of the value chain or the process. You could have agents that help you from the data stage. Like for example, I identify some problems with the data. I want to eliminate the outliers, or I want to make some replacements or convert a continuum function like h into um h, not per se as continuum. It could be just years, but let’s assume it’s a continuous function into a different categories. So you can have an agent that does that for you again super quick versus spending a lot of time, um, coding that, um, so that you create the attributes that will help you with the, the campaign.
[00:29:45] Silvia: And you can also use an agent to construct the campaign itself, like creating the segments, creating the different logic, uh, from the previous predictive analytics and rules that you get from that to help you streamline massively the funnel of building a new campaign with control groups and all of that. And you can also have agents that send the campaigns with specific triggers as well. And so I think that is a good explanation of how to use the different layers of AI from a CVM perspective for outbound. And then if you look at the inbound component, it’s probably one of the most famous as well these days in terms of having the, the front line service agents or support agents helping customers with identic tools as well. And so in that situation, the, the activity or the process could vary as well. It could be from a service inquiry, like, I want to move from this plan to a cheaper plan, for example, in a phone company, a phone company, or it could be a sales conversation as well. I want to buy another service for my children, for example. And so overall, it’s not that you have identical and non-identical campaigns. It’s about how to use the different elements of AI for the best purpose across the, the, the whole process.
[00:31:23] Exacaster: If you are interested in customer value management, check out our customer Value Management body of knowledge. Cvm blog is a comprehensive guide for CVM professionals offering tips, tools, and best practices to help you in your job. Visit cvs.com for more.
[00:31:41] Egidijus: I would like to link this to how a general way of work will change for CVM. Historically, you know, the CVM team sits in between of everything between data, between, uh, campaign configuration pricing, inbound channels, uh, outbound, uh, channels like sales teams, etc. and it’s kind of a customer value manager would be working as a coordinator, you know, of like, please, I need new customer attribute added to the data warehouse or I need a new predictive model, you know, and the tasks would land to other teams, uh, uh, in the organization. And now how does it, how does this thing change? Because if I can ask that the attribute myself, uh, in the data warehouse, you know, how does it change my role?
[00:32:42] Silvia: I think with the new ways of working and AI being so central to everything that we do, and I think while CVM is critical in terms of that orchestration of the customer interactions and engaging across the board. I think the the role of the I don’t know if it is called the head of AI or the chief of AI. It’s also super important, and I think CVM will need to team up with this AI chief or VP to make sure that as part of those AI priority use cases or strategic priorities, overall, CVM is a central part of it. But how specifically it works, it will depend on the organization and the regulation in the specific country as well. And overall, what is the appetite? There are multiple things that you can consider in terms of appetites or flavors. Some organizations might be more growth focused or risk focused or a combination of both. And if I think about those two terms, just for simplicity, I think there could be a situation where the focus is growth. Um, and then in those scenarios, maybe it’s more like, um, a decentralized model where the different parts of the business have autonomy and knowledge to use AI in the way they want for the specific purposes of the departments, like for example, um, in an environment where growth is the priority, maybe in, if I talk about telco in the home space, the team is maximizing opportunities using AI to win more customers using internet, but in parallel in the postpaid space, the same thing might be happening. I find that this growth mindset and with AI, where each of the individual teams are kind of upskilling themselves individually and independently and building somehow like the ecosystem in different spaces, like the home space for this example, or the postpaid space for this example.
[00:35:20] Silvia: I think that can work okay for experimentation and for early days, just to help really understand what works, what doesn’t, and where do I invest more time or efforts? I think it has some problems as well, in particular around duplication and not learning from each other could be very fast, but maybe not as efficient in the longer run. However, if I go to the other extreme right, like organizations that basically they want Consistency. They want minimizing the risk and they do not want different teams working in isolation. And those organizations might choose to implement a very risk and control oriented model, which is probably very, very centralized. Let’s say one center of excellence that has to review every single use case that wants to use AI and for what purposes, and then has dedicated people, a small group of people that are dedicated to implementing AI case use cases across the business. So that could be for customer value management or for legal or for post pay anything. And so I think that model has benefits in particular around that consistency. Learning from each use case and creating that intelligence across the organization. At the same time, if you take it to the very extreme, it can be very slow and demotivating as well for people that want to launch their use case because they really see the benefit, but they are like experiencing bottlenecks from other initiatives that got there before them, and they kind of move through the bottleneck.
[00:37:25] Silvia: And so I think that probably the most successful model will be in the in the middle, in between the only growth mindset and only risk mindset and somehow finding, um, probably a hybrid center of excellence where you have a central hub that defines like the AI ethical principles that defines the ways of working the methods that somehow provides a lot of training across the organization so that then the individual, if you like, squats or can operate and implement AI initiatives in a consistent way, sharing knowledge, and then you have a lot of common operational rhythm, like, for example, quarterly planning that helps across the organization to understand which initiatives to focus on based on the growth, the impact, whether they can be reused across different teams. It’s not just one particular siloed initiative. And so that that hybrid model provides guidance, provides guardrails, but also helps to accelerate the delivery across the organization. And so I think if we were to operate in that hybrid model. Coming back to the beginning of your question, like, how is this now working for the CVM team and the head of CVM? So I think the head of CVM will need to be very close to the head of AI and also find a way that the CVM initiatives are prioritized through that center of excellence, hybrid center of excellence to get the resources, the training, and the impact.
[00:39:20] Egidijus: So I have a provocative question for you. Yeah. Uh, I am one of those, you know, uh, super AI adopters. Um, what we see in our organization is, uh, that, uh, the biggest impact of AI we can get the biggest impact of from adopting AI in the situations when we have, let’s say, large, extremely cross-functional projects because kind of in large, extremely cross-functional projects. The biggest cost is not doing the work itself. It’s usually handovers, alignments, and kind of all the things that fall into between two chairs. And it’s like figuring all these things out. And I see parallels in CVM, like when we do this CVM trans reports and we interview, uh, telecoms and there are a kind of everything. What sits directly inside the CVM team is really fast. Can I create a new campaign? Yes, you can very fast. But as soon as we are talking about entering new data points, adding new integrations, configuring maybe new price plans or, or whatever, which is, uh, more cross-functional work when you need to align a couple of, um, uh, aspects of our organization. These get, uh, kind of the timing becomes months, half a year, sometimes years, you know, and in our case, we see situations that when we apply AI across, you know, large cross-functional projects, we can go from like a year of delivery towards maybe weeks, uh, which is like, uh, extreme, uh, speed up. Uh, do you see this extreme speed up for the CPM processes in this age of AI? So for example, let’s say CVM team gets access to all necessary systems. And then if they need to do integration, they ask their agent to do an integration. Then to add the data point, then, you know, to configure a new price plan if necessary, you know. Do you see the kind of this potential to speed up drastically some of the aspects of the CVM work?
[00:41:55] Silvia: I think that there are different components to consider when answering this question. I think that if we focus for a moment, just in the identic AI component, we can imagine, and actually this is happening already organizations that are focusing on specific use cases, like, for example, customer support, and they might have started using Identic AI with proofs of concepts or pilots. And then once they have become more confident and customer experience is good, the effectiveness and efficiencies are better than not using it, then they are implemented and part of the day to day. So there have been efficiencies in how they operate in that particular use case. However, there is only one use case, as I said. For example, in customer support, then this organization could have another use case around legal and marketing communications and how to use a the agents to review the communications according to the legal considerations and the brand tone of voice and propose changes, or even implement those changes within guardrails. That would be another use case. However, what I think is starting to happen the more and more, and I would imagine a world where this is functioning in full esteem is like an orchestration layer, like the whole interface. If you think about the all the IT terms, but an agent that orchestrates and talks to all the agents across the enterprise like this Orchestrator will be talking to the agent in the data space, the data warehouse, or any kind of structure or unstructured data, and ask questions to get the right data for the activity that they want to conduct end to end, from the data to the customer. Then they will talk to the CVM agent, for example, right, to understand how that data is being used throughout campaigns through different customer segments. What’s the success of that campaign versus the control group? Then they can talk to the agent in digital and so on and so forth. So I believe with this agent side is taking us, is towards that conductor of agents and orchestrator that can operate through the organization end to end within the respective guardrails.
[00:44:39] Egidijus: Do I understand correctly that In this world where every organization function has an agent or two, maybe three. Uh, can we imagine the autonomous car world where, you know, I come to the office on Monday morning, take my coffee and just observe what what my CVM agent is about to do, you know, uh, maybe launch a new, uh, onboarding campaign, uh, test or maybe proposing to add a new price plan that would help to do the upsell and cross-sell better. Uh, do you see this coming?
[00:45:22] Silvia: Um, do I see it coming? That’s a great question. Do I think is it possible? I think it’s possible, yes. Um, I’m not sure it will happen though. Um, maybe it happens within a specific um guardrails because what this technology is known for is for maximizing the objectives. So if you set up an agent to maximize, let’s say, growth, it will do whatever it needs to do to get that growth KPI maximize. That might not be the best business or customer outcome because it might be detrimental to other KPIs, right? So I think maybe what happens is there will be much more autonomy. I think it’s super important that we embed the risk frameworks in the different countries, in the different industries properly. There are some very basic questions that we should be asking the AI agent or orchestrator conductor in terms of what information is AI using for what purpose? A what is the impact if something goes wrong? Do we have the controls in place to identify if something goes wrong before. Before it escalates and becomes even a more problematic issue? Can we stop it at any time if we want to? Like, what’s the role of human in the loop? And so I think from a disability perspective and a technology perspective, I completely see that being possible.
[00:47:09] Silvia: However, I think we haven’t yet understood all the implications and all the considerations. So we need to be careful in how we implement it. And if there are use cases that have no impact to the customer, to the economy, to the industry, we can be much more flexible and futuristic if we want. If AI is deciding, for example, whether to And give a loan to an individual or not. I think we need to have definitely human in the loop, and we need to make sure that from a regulation perspective, we are being with AI non-biased, transparent, and we are not benefiting some specific customer cohorts and being detrimental to others. So there are so, so many considerations that it’s, it’s actually a great time to be in business in terms of reinventing the operating model, the risk frameworks, and making sure that those risk frameworks are evolving over time because AI is non-deterministic.
[00:48:15] Egidijus: Customer value managers typically are quite senior roles, and you get kind of a you become a member of customer value manager when you send the wrong message for the wrong customer base, uh, kind of, or you send a bad message to the whole customer base, you know, because everyone has the story. And and once this happened, happened when you send the bad message, no. Or you spam a single customer with 100,000 messages, you grow as a customer value manager. You know, you come from the first level to the second level after this. Yes. After when you get this car, something changes. It’s like you become way more paranoid about making mistakes. Et cetera. Et cetera. So I’m wondering how how these cars will be implemented into the, you know, the agentic world. Because when you talk about the guardrails, it’s like, these are the, the scars that we as professionals have. Um, but it’s like we, we will need to transfer those into, into this autonomous thinking or autonomous, uh, machines that will help us out in the future. From the practical perspective, what would be, let’s say first 3 or 4 areas where you would build the biggest guardrails for for AI agents?
[00:49:48] Silvia: Yeah, I think it’s a great question. And it’s also a reflection of how the world overall is changing, because at the moment, you can progress really fast across all the steps in the funnel. Like for example, the data, the analytics models, the campaign build, and therefore, what’s the impact to the customer versus before? Maybe you were progressing in a more a waterfall way. So you could assess, for example, the risk and issues from a data perspective before you will build the models, before you will build the campaigns, before you will see what’s the response from the customer. And so if I think about the risk at the moment, the. The one first consideration is being conscious that. Again, as I said just a moment ago, that AI and AI is non-deterministic. So the traditional models for risk have been based on an assessment that happens in a in a particular moment in time, for example, when you launch or are about to launch, and then that risk assessment continue to be valid, like forever, unless something changes like an upgrade or a technology, something changes. But in AI, that risk assessment is valid only in the moment where you did the iteration, then you need to establish those risk mechanisms ongoing because the response, the result that you are going to get from the agent Might be different the next time.
[00:51:31] Silvia: So ensuring that we have those risk mechanisms that continue to monitor and reassess and reevaluate the risks from an ongoing perspective. It’s super important. Establishing a specific alerts within the guardrails, right? Like maybe it’s all green, but something is starting to become amber. Let’s review. Maybe there is more human oversight for those particular situations. Because once you understand what is the baiting from an outcome perspective, then you can identify how to fix it, whether it’s more data related or more related, or so on and so forth. And so that’s one very important component. And the other component is from a regulatory perspective, and again, this might be different across different industries and countries. But from our perspective there are some common ones. You touched on the spam act or from a overcommunicating perspective and making sure that we respect customers wishes from. I don’t want to get marketing campaigns, and even if they want to get marketing campaigns so that they have a very easy option out to opt out. So making sure that those links processes are always in place, whether that is via identity campaigns or not, and working fully and easily for customers, that’s super important.
[00:53:06] Silvia: And also from a privacy perspective, making sure that we do not share information that is private information in across the organization or the channels. And so making sure that all those security and privacy components are in place. Is is critical. It’s like the super bare minimum. It’s a non-negotiable. And I know that at times compliance might not feel like very customer centric. But I really believe it’s like the bare minimum. Like if you are not compliant, you cannot say your customer centric. It’s impossible. No way. Yeah. And so yeah, I think those are some of the ones that I will absolutely establish. And then one important one is like the, the decision making process, like making sure that the right governance is in place, that there is seniority across the forums that will be to be that will need to be established to approve the use cases, the oversight, and also understanding when each use case moves from a POC to, if you like, embedded in the ways of working for one individual use case versus the orchestration that combines multiple agents talking to each other and making decisions.
[00:54:28] Egidijus: So the future is very bright. But, seems far as you know, uh, what would be your recommendation for the first actionable step in the CVM area towards this, uh, bold vision?
[00:54:50] Silvia: Yes. And so I think it would be similar to defining a new strategy in the sense of, uh, why are we doing this? The first question, like, what’s the purpose of what we are trying to achieve in CVM? I know like, in very simple terms, it’s about growing, retaining and making sure that customers also, uh, get more value, right. But why are we doing this? Setting up a specific KPIs that help establish the direction of this new, if you like AI native AI, CVM, right. And the second one is normally will be the. How do we get there? But in this particular case, I’m going to skip that because AI, as I said, is a is design or workspace with objectives. It’s less about the process and the how. It’s more about the objective. So instead of talking about the how, I will go to the what, which is that outcome that we want to deliver. So what are those KPIs, those targets that you want to nail that are aligned with the why we exist? So once you know the why, once you know the KPIs and objectives, then the big job is to change the how. Because today in CVM, we can operate following processes that have been there for a while.
[00:56:21] Silvia: Using tools that didn’t have AI embedded before or not. Definitely the AI. So I think we need to reflect on the current processes and how we can reinvent them differently in a way that uses the new technology, the new AI agents, and that is a focus on on achieving the goal differently in a more efficient way. Like if I want to grow, let’s say a customer value 2% every year, how can I do it differently? And instead of trying to use AI to conduct the steps of your current processes, rethink the processes using the new technology. How to do that? Obviously, workshops with people that have been in CVM for a while, but also outside of CVM people that have used AI for the day to day life from an admin perspective. From a legal perspective, there are so many tools that each of them can help you to get a different perspective and learning. And out of that, if you like incubation process, I’m sure there will be lots of ideas on how to deliver that goal much faster and using AI in the in the process.
[00:57:45] Egidijus: I think it’s a really, really good advice. I think we also see organizations maturing towards the way they think about adopting, especially a genetic processes, because many start with small scattered use cases, and then they navigate towards more the use cases that cover their core value proposition aspects. Because it’s like you start with something small scattered around the company, but the real value comes when you move to these more strategic ones which help to deliver your that your key value proposition from the organization. So yes, I think here it’s a really nice part to wrap up today’s conversation. So I wish every company to find their place for or their way to adopt AI. Thank you, Silvia, for this really insightful conversation.
[00:58:57] Silvia: Thank you so much for having me. And future will tell.
[00:59:01] Exacaster: CVM Stories is produced by Exacaster. We help companies take their customer value management to the next level. To stay updated on our latest episodes, subscribe to the podcast or sign up for an email newsletter at exacaster.com/cvmstories.
