How Verizon’s Business Intelligence Team Built a Single Customer View From 250 AI Models
- Kothai Sriramachandran, Senior Director of Business Intelligence at Verizon Business Group and Natalie Sartain, Director of Business Intelligence at Verizon Business Group
- AI and Automation, Campaign Optimization, Content Strategy and Marketing, Customer Data Platform (CDP), Customer Experience (CX), Data Management and Analytics, Digital Transformation, Lead Gen / Lead Mgmt, Managing Org Transformation, Managing Your Career, Martech Stacks, Multi-Channel Journeys, Other Martech, Platform and Data Integration, Sales and Marketing Orchestration, Strategy and Trends
When Verizon Business Group won Forrester’s B2B Program of the Year Award for its Customer Experience Index, or CXI, the recognition validated something the team had suspected for a while: Predicting customer behavior is not the same as understanding the customer. To learn how the program came together, and what other marketers can take from it, the Martechify team spoke with Kothai Sriramachandran, Senior Director of Business Intelligence, and Natalie Sartain, Director of Business Intelligence, both at Verizon Business Group.
Kothai leads the team that built CXI from the ground up, a group that now runs a variety of AI and machine learning propensity models focused on growth, retention, and customer experience. Natalie leads strategy for the customer experience and revenue operations organization, working closely with frontline teams to put those models into practice. Together, Kothai and Natalie offered a candid look at what it actually took to move from fragmented data to a unified customer strategy, and it wasn’t the modeling.
Kothai leads the team that built CXI from the ground up, a group that now runs a variety of AI and machine learning propensity models focused on growth, retention, and customer experience. Natalie leads strategy for the customer experience and revenue operations organization, working closely with frontline teams to put those models into practice. Together, Kothai and Natalie offered a candid look at what it actually took to move from fragmented data to a unified customer strategy, and it wasn’t the modeling.
Watch the full interview with Verizon Business Group’s Kothai Sriramachandran and Natalie Sartain to hear how they unified 250+ AI models into a single, award-winning Customer Experience Index. Learn how breaking down silos and balancing speed with governance turned predictive data into a truly unified customer strategy.
The elephant problem
Verizon’s business group did not lack data or predictive power. Before CXI existed, Kothai’s team had already built more than 250 AI and machine learning propensity models, each predicting a different piece of customer behavior, from churn risk to product add likelihood. On paper, this looked like a major win. In practice, the models worked in isolation, each one owned by a different function like sales, marketing, customer experience, network, or service, with no clear way to see the full picture of a single customer.
Kothai used a colorful metaphor when she described what Verizon customers experienced, “It looked like a wild buffalo with a trunk and two tusks,” she said. “And after seeing that, we realized that maybe the customer wanted an elephant.” Each team was technically doing its job well, but the fragmented approach meant customers experienced Verizon as a series of disconnected interactions rather than one coherent relationship.
Kothai also pushed back on how most people inside telecom frame competition, a mindset she considers outdated. “Most people inside corporate telecom still frame competition as Verizon versus AT&T versus T-Mobile,” she said. “But that is not how your customers think.” A customer, she explained, doesn’t benchmark their carrier against another carrier. They benchmark it against every digital interaction they have in a day: “They expect Amazon’s level of support. They expect Apple’s seamless onboarding. They expect Uber’s real-time visibility. That is what the customers are looking for.”
That gap between accurate pieces of insight and a coherent whole is exactly what CXI was built to close. As Kothai explained, the industry standard Net Promoter Score wasn’t cutting it because it only captures a snapshot, twice a year, from a fraction of the base. “It is reactive, it is survey based, and it’s done probably twice a year,” she said. “It’s not going to encompass all of your customers. It’s not representative of your entire base.” CXI was designed to be proactive instead, using the 250-plus propensity models and a layer of nuanced micro-segmentation as its foundation, treating every customer according to their own behaviors and attributes rather than a blanket score applied to everyone.
The result does more than classify a customer as a promoter, passive, or detractor. It explains why a customer falls into that category, and what specific action would move them toward a better one. And for customers who are already promoters, the model’s job shifts from selling more to selling smart. As Kothai put it, when a customer already has a great experience, “don’t sell anything and everything. Just understand what the customer needs and be very precise about what you’re selling.”
Kothai used a colorful metaphor when she described what Verizon customers experienced, “It looked like a wild buffalo with a trunk and two tusks,” she said. “And after seeing that, we realized that maybe the customer wanted an elephant.” Each team was technically doing its job well, but the fragmented approach meant customers experienced Verizon as a series of disconnected interactions rather than one coherent relationship.
Kothai also pushed back on how most people inside telecom frame competition, a mindset she considers outdated. “Most people inside corporate telecom still frame competition as Verizon versus AT&T versus T-Mobile,” she said. “But that is not how your customers think.” A customer, she explained, doesn’t benchmark their carrier against another carrier. They benchmark it against every digital interaction they have in a day: “They expect Amazon’s level of support. They expect Apple’s seamless onboarding. They expect Uber’s real-time visibility. That is what the customers are looking for.”
That gap between accurate pieces of insight and a coherent whole is exactly what CXI was built to close. As Kothai explained, the industry standard Net Promoter Score wasn’t cutting it because it only captures a snapshot, twice a year, from a fraction of the base. “It is reactive, it is survey based, and it’s done probably twice a year,” she said. “It’s not going to encompass all of your customers. It’s not representative of your entire base.” CXI was designed to be proactive instead, using the 250-plus propensity models and a layer of nuanced micro-segmentation as its foundation, treating every customer according to their own behaviors and attributes rather than a blanket score applied to everyone.
The result does more than classify a customer as a promoter, passive, or detractor. It explains why a customer falls into that category, and what specific action would move them toward a better one. And for customers who are already promoters, the model’s job shifts from selling more to selling smart. As Kothai put it, when a customer already has a great experience, “don’t sell anything and everything. Just understand what the customer needs and be very precise about what you’re selling.”
Getting an organization to agree on one starting point
Building the model was hard. Getting the rest of Verizon to work from it was harder. Once the team set out to unify these models into CXI, the bigger challenge wasn’t building the technology, it was getting buy-in across teams that each had their own goals and metrics to hit. Sales, product, service, and network teams were all optimizing for different numbers, and none of those numbers were wrong on their own, they just didn’t add up to a single customer experience.
Kothai described the dynamic directly: “Teams have their own priorities, right? If you talk about sales teams, they have their own gross-add goals. If you go to product teams, they would say, ‘My top products that I need to focus on this year are Verizon Business Complete and FWA managed services.’ If you go to service teams, they will say, ‘My priority is reducing average handle time. My priority is reducing the number of chronic callers who call us over and over again.’ If you go and talk to network teams, they will say, ‘I need to resolve my tickets faster. I need to make sure that I do not get so many outages anymore.’”
She was also candid about her own team’s role in creating this challenge in the first place. Building standalone models by function, she acknowledged, had reinforced the silos to begin with, which made the pitch for a single, unified model harder rather than easier, since every team already had a model of its own solving its own narrow problem. The fix wasn’t a new model so much as a new mindset. “Doesn’t matter whether it’s marketing, doesn’t matter whether it’s product pricing,” Kothai said. “It all has to start with the customer.”
Importantly, CXI didn’t ask any team to give up its own goals. Kothai described it as giving each function a package built just for them. Product teams still got to prioritize their own top products, but CXI told them which of those products actually fit a given customer. Sales teams still had their own quotas to hit, but CXI told them where to start so they could hit those numbers without over-pitching a customer who wasn’t ready. Service teams still needed to bring down handle time and repeat calls, but CXI helped route the right customers to the right queue so agents could do that without treating every caller identically. That reframing, giving every team a reason to see the shared model as an accelerant rather than a threat to its own goals, was central to getting adoption across the business.
Kothai described the dynamic directly: “Teams have their own priorities, right? If you talk about sales teams, they have their own gross-add goals. If you go to product teams, they would say, ‘My top products that I need to focus on this year are Verizon Business Complete and FWA managed services.’ If you go to service teams, they will say, ‘My priority is reducing average handle time. My priority is reducing the number of chronic callers who call us over and over again.’ If you go and talk to network teams, they will say, ‘I need to resolve my tickets faster. I need to make sure that I do not get so many outages anymore.’”
She was also candid about her own team’s role in creating this challenge in the first place. Building standalone models by function, she acknowledged, had reinforced the silos to begin with, which made the pitch for a single, unified model harder rather than easier, since every team already had a model of its own solving its own narrow problem. The fix wasn’t a new model so much as a new mindset. “Doesn’t matter whether it’s marketing, doesn’t matter whether it’s product pricing,” Kothai said. “It all has to start with the customer.”
Importantly, CXI didn’t ask any team to give up its own goals. Kothai described it as giving each function a package built just for them. Product teams still got to prioritize their own top products, but CXI told them which of those products actually fit a given customer. Sales teams still had their own quotas to hit, but CXI told them where to start so they could hit those numbers without over-pitching a customer who wasn’t ready. Service teams still needed to bring down handle time and repeat calls, but CXI helped route the right customers to the right queue so agents could do that without treating every caller identically. That reframing, giving every team a reason to see the shared model as an accelerant rather than a threat to its own goals, was central to getting adoption across the business.
Orchestration, not the algorithm, was the hard part
Once the organization agreed to rally around the customer as the starting point, the real engineering challenge began. Kothai was direct in saying that “the orchestration and implementation part was the difficult part, not the model itself.”
That meant sales teams needed CXI’s insights inside their own front-end application, since that’s what the system reps actually use every day. Service teams needed the scores and logic running on the back end to route customers to the right support queue in real time. Marketing teams needed the same underlying intelligence translated into audience targeting for specific, already-prioritized campaigns. Every one of those integrations had to be built and automated separately, even though they all pulled from the same underlying model and the same customer record. As Kothai put it, because it was all starting from the same point, the various pieces ended up “acting together” and “in synergy,” but getting there required custom plumbing into nearly every team’s existing tools rather than a single shared dashboard.
Natalie brought the sales and marketing leadership perspective to that rollout, having spent time in Verizon’s Public Sector sales leadership and regional marketing before moving into business intelligence. She described what CXI unlocks for frontline employees this way: “It allows them to pause and really listen and understand and empathize… allowing them to expedite the resolution as well as hear the customer out.”
She also framed CXI as more than a targeting tool; it’s also a way to build “actionable trust.” “Yes, the intelligence is there. It’s a matter of how we apply it,” Natalie said, describing how teams across the organization absorb the model’s insights and apply them to outreach, triggers, and campaigns so that every customer touchpoint feels informed rather than generic. She also pointed to training built alongside Verizon’s learning and development team, meant to help employees understand not just that the model exists, but how to actually use it in the moment, whether that’s an inbound call, a scheduled check-in, or a proactive outreach.
That meant sales teams needed CXI’s insights inside their own front-end application, since that’s what the system reps actually use every day. Service teams needed the scores and logic running on the back end to route customers to the right support queue in real time. Marketing teams needed the same underlying intelligence translated into audience targeting for specific, already-prioritized campaigns. Every one of those integrations had to be built and automated separately, even though they all pulled from the same underlying model and the same customer record. As Kothai put it, because it was all starting from the same point, the various pieces ended up “acting together” and “in synergy,” but getting there required custom plumbing into nearly every team’s existing tools rather than a single shared dashboard.
Natalie brought the sales and marketing leadership perspective to that rollout, having spent time in Verizon’s Public Sector sales leadership and regional marketing before moving into business intelligence. She described what CXI unlocks for frontline employees this way: “It allows them to pause and really listen and understand and empathize… allowing them to expedite the resolution as well as hear the customer out.”
She also framed CXI as more than a targeting tool; it’s also a way to build “actionable trust.” “Yes, the intelligence is there. It’s a matter of how we apply it,” Natalie said, describing how teams across the organization absorb the model’s insights and apply them to outreach, triggers, and campaigns so that every customer touchpoint feels informed rather than generic. She also pointed to training built alongside Verizon’s learning and development team, meant to help employees understand not just that the model exists, but how to actually use it in the moment, whether that’s an inbound call, a scheduled check-in, or a proactive outreach.
Real examples: Six months of runway and a data breach report
For Verizon’s business customers, especially mid-market and large enterprise accounts, the model’s early-warning capability matters in a very concrete way. Kothai explained that the earlier generation of standalone models could predict churn 30 days out, seven days out, or even as an imminent event, but for larger business customers, that kind of short runway isn’t much use. “Is seven days enough time for you to turn things around? Is one day enough for you to turn things around, improve the customer experience, and save the customer?” she asked. For consumers, that might mean a last-ditch effort. But with businesses, you need real time to work with account teams.
CXI closes that gap by flagging a business heading toward becoming a detractor as much as six months before it churns. “Six months was enough time for people to bring things together and make sure they work with the customers directly,” Kothai said, adding that for large accounts, the stakes are especially high, since losing one enterprise customer can mean losing thousands of lines at once. That lead time lets dedicated account reps focus first on resolving whatever is actually frustrating the customer, and only then start exploring what else that business might need. As Kothai explained, Verizon’s role with business customers goes well beyond connectivity. “Our factory floors and refrigerators are smarter than we are now,” she said, pointing to a broader shift where Verizon Business increasingly functions as a technology partner rather than just a network provider, offering private 5G networks, IoT products, Total Mobile Protection for large lines of devices, and unified communications through One Talk, alongside its more traditional fiber, fixed wireless, and mobility products.
Natalie offered a specific example of that advisory relationship in action, describing how Verizon’s Data Breach Investigation Report, or DBIR, has become an entry point for deeper conversations with small and mid-size businesses that might assume cybersecurity risk doesn’t apply to them yet. “Because we can be a better listener, we might be solutioning other things that help them react or respond, or proactively engage with their customers where they’re bringing delight as well,” she said. She noted that some of those outcomes have since become public success stories of smaller businesses that were able to think beyond the scope of the original conversation entirely because a rep had the context to recognize an opportunity the customer hadn’t yet flagged themselves.
CXI closes that gap by flagging a business heading toward becoming a detractor as much as six months before it churns. “Six months was enough time for people to bring things together and make sure they work with the customers directly,” Kothai said, adding that for large accounts, the stakes are especially high, since losing one enterprise customer can mean losing thousands of lines at once. That lead time lets dedicated account reps focus first on resolving whatever is actually frustrating the customer, and only then start exploring what else that business might need. As Kothai explained, Verizon’s role with business customers goes well beyond connectivity. “Our factory floors and refrigerators are smarter than we are now,” she said, pointing to a broader shift where Verizon Business increasingly functions as a technology partner rather than just a network provider, offering private 5G networks, IoT products, Total Mobile Protection for large lines of devices, and unified communications through One Talk, alongside its more traditional fiber, fixed wireless, and mobility products.
Natalie offered a specific example of that advisory relationship in action, describing how Verizon’s Data Breach Investigation Report, or DBIR, has become an entry point for deeper conversations with small and mid-size businesses that might assume cybersecurity risk doesn’t apply to them yet. “Because we can be a better listener, we might be solutioning other things that help them react or respond, or proactively engage with their customers where they’re bringing delight as well,” she said. She noted that some of those outcomes have since become public success stories of smaller businesses that were able to think beyond the scope of the original conversation entirely because a rep had the context to recognize an opportunity the customer hadn’t yet flagged themselves.
Balancing speed with governance
The final challenge came after CXI was built and adopted across the business: how to move fast on real-time customer insight without compromising Verizon’s data governance and compliance standards. This is a tension almost any marketing organization working with AI and customer data will recognize.
Kothai explained that every model goes through legal review, and that marketing use cases must also account for compliance rules like CPNI, or Customer Proprietary Network Information, along with do-not-contact preferences. “There’s a lot of governance in play. There’s a lot of compliance in play,” she said.
At the same time, she pushed back firmly on the instinct to over-engineer approval processes in the name of caution. Companies, she said, often fall into “the ruts of organizational norms,” building “boundaries, putting up rules and endless review processes that make it impossible to move fast.” What frontline teams really need, in her view, is enough agency to act the moment they identify a frustrated customer, rather than waiting on approval from multiple layers of management. “You have to give your teams the extreme agency to see a frustrated customer and instantly pivot to solving their problems without needing to get approval from three different managers.”
Natalie’s work on the policy side was central to solving that tension, simplifying internal processes so that governance and compliance could coexist with the speed and responsiveness CXI was designed to enable. The two of them described the outcome less as a tradeoff between speed and safety and more as a design problem, building the rules in tightly enough at the model and data layer that the humans acting on the output don’t have to stop and ask permission every time a customer needs help.
Kothai explained that every model goes through legal review, and that marketing use cases must also account for compliance rules like CPNI, or Customer Proprietary Network Information, along with do-not-contact preferences. “There’s a lot of governance in play. There’s a lot of compliance in play,” she said.
At the same time, she pushed back firmly on the instinct to over-engineer approval processes in the name of caution. Companies, she said, often fall into “the ruts of organizational norms,” building “boundaries, putting up rules and endless review processes that make it impossible to move fast.” What frontline teams really need, in her view, is enough agency to act the moment they identify a frustrated customer, rather than waiting on approval from multiple layers of management. “You have to give your teams the extreme agency to see a frustrated customer and instantly pivot to solving their problems without needing to get approval from three different managers.”
Natalie’s work on the policy side was central to solving that tension, simplifying internal processes so that governance and compliance could coexist with the speed and responsiveness CXI was designed to enable. The two of them described the outcome less as a tradeoff between speed and safety and more as a design problem, building the rules in tightly enough at the model and data layer that the humans acting on the output don’t have to stop and ask permission every time a customer needs help.
The bigger picture
What stands out about Verizon’s CXI story isn’t the model count or the award. It’s the admission, from the team’s own senior leaders, that 250 well-built models were not enough on their own. Kothai and Natalie’s experience suggests that the hardest problems in customer experience work are rarely technical. They are organizational: Getting every function to agree that the customer, not the department’s quarterly goal, is the starting point, and then building the unglamorous plumbing that gets one model’s insight into a dozen different teams’ daily workflows.
As Kothai summed it up, whether you are in marketing, product, or pricing, “It all has to start with the customer.” For any organization sitting on its own pile of disconnected models and dashboards, that is the elephant in the room worth naming first.
As Kothai summed it up, whether you are in marketing, product, or pricing, “It all has to start with the customer.” For any organization sitting on its own pile of disconnected models and dashboards, that is the elephant in the room worth naming first.
Related resources
Want more like this resource? Find helpful links below.
