As one of the world’s largest communication technology companies, Verizon serves over 140 million retail connections and operates a leading 5G and fiber-optic network across the U.S. With a business spanning consumer, enterprise, and public sectors, Verizon’s competitive edge increasingly rests on how it manages and activates its data.

In the second installment of our three-part conversation series, Kalyani Sekar, Chief Data Officer at Verizon, joins Yali Sassoon, Co-founder and CTO of Snowplow, to explore how Verizon is aligning its behavioral data strategy with evolving business goals while building the real-time and AI-powered infrastructure required to deliver seamless, proactive customer experiences.

Part one of this series delved into Verizon’s enterprise-wide approach to embedding data quality, governance, and observability at scale. In this follow-up, Sekar goes deeper into the transformation of Verizon’s data landscape – from real-time network insights at the edge to AI-powered customer recommendations, sales nudges, and call center pairings.

She shares how behavioral data layered over core operational data helps Verizon move from reactive to predictive to proactive, and why success in AI isn’t just about models, but managing GPU demand, infrastructure, and FinOps with equal rigor.

Edited Excerpts

Q: How are you aligning Verizon’s behavioral data strategy with its evolving business goals and focus on customer experience?

The customer is at the heart of everything we do. When I put the customer at the center, everything else like employee data, network data, device data, and distribution data, forms a circle around that. The key is how we bring the data together to extract behavioral patterns because combining core data with behavioral data provides a 360-degree view of every entity in the organization.

Once we have that complete view of how an entity is interacting with Verizon, it enables us to apply both predictive and prescriptive AI capabilities. When those capabilities work well together, they allow us to become proactive. And when we’re proactive, we can anticipate customer needs and act ahead of time, so the customer doesn’t even experience an issue with Verizon.

All of this becomes possible through behavioral data that is layered on top of the core data. This behavioral data comes from the entity itself and, in some cases, from third-party data sources.

By developing a strong understanding of both our own managed behavioral data and third-party behavioral inputs, we’re able to create a holistic view of the entity. That’s what empowers us to move from predictive to prescriptive to truly proactive.

Q: You’ve mentioned behavioral data, how are you using data in real time, especially given the volume, speed, and variety of data at Verizon?

We’ve been leveraging real-time data for quite some time. Real-time data is critical, especially on the network side. We bring in real-time data from various touchpoints and channels.

Let me give you a few examples. Customers interact with us through multiple channels like IVR, digital platforms, or the call center. We also send a lot of information to customers, whether it’s a bill or other communications.

Bringing all of that data together in real time and mapping the customer journey gives us a clear view into how the customer is experiencing Verizon. With that insight, we’re able to accurately predict and prescribe the next best action or offer – from a commercial standpoint.

On the network side, the devices in the field emit massive volumes of data every subsecond. We process that data right at the edge without even bringing it all into a centralized warehouse because it would introduce latency. And when someone’s making a call, even a tiny glitch is unacceptable. So we analyze the data in real time at the edge. As we ingest it, we immediately measure key KPIs to understand how the network is performing. Whenever we detect a glitch or anomaly, we act instantly and that’s what ensures a seamless experience when customers make a call or browse the internet.

Our goal is to continuously bring in real-time data, detect anomalies quickly, and ensure our customers enjoy a flawless experience on our network.

Q: Turning our attention to AI, how do you see AI transforming Verizon’s approach to data infrastructure and customer experience? And are there particular initiatives or ideas that you’re especially excited about?

It is really transforming the way we manage infrastructure. Let me start from the machine learning days, back when we were working with structured data and statistical models. Even then, as we transitioned into deep learning, the infrastructure needs went up tremendously. That shift made us pay a lot more attention to how we forecast storage and compute demand regularly. It wasn’t like before.

In the earlier days of machine learning, it was enough to plan hardware every quarter or six months. But with deep learning, it became a daily, even minute-by-minute concern. As we moved closer to real-time applications, every minute started to matter.

Along with forecasting compute and storage, we had to start thinking about auto-provisioning and building self-healing capabilities into the infrastructure. All of that became essential and that’s when hyperscalers also started embedding these capabilities into their offerings.

Then came GenAI, and with it, a huge demand for GPUs. But it wasn’t just about the GPUs, it was about how to manage them. The demand far exceeded supply, and that made it critical to manage GPUs economically. Our team really focused on how to effectively plan and optimize GPU usage continuously. Infrastructure management itself became almost as complex as managing AI.

As we expanded into hyperscalers and the cloud, managing costs, FinOps, became just as important as managing infrastructure. Otherwise, costs would skyrocket. So that became another priority. Infrastructure and FinOps are really two sides of the same coin. It’s a continuous process of evaluation.

We also started introducing AI techniques to detect anomalies and trigger proactive responses for managing infrastructure. So yes, operating in a multi-cloud world and performing AI at the edge became central. Across all these shifts, managing infrastructure while keeping a strong focus on FinOps was absolutely critical for us.

Q: Looking ahead, how is AI starting to impact the business now that you have well-governed, high-quality data and the right infrastructure to deploy models effectively?

AI is becoming an integral part of business processes. Before GenAI, deep learning, and real-time capabilities emerged, AI was mostly seen as a nice-to-have. It was primarily used to support operational decisions but it wasn’t expected to make decisions on the fly.

Now, with advances in agentic AI and deep learning, we’re entering a space where AI is embedded into systems and integrated into day-to-day business operations. AI is starting to recommend products that customers may need – not just in digital channels but also in physical retail environments and partner stores. It’s beginning to generate scripts or nudges for sales teams, guiding how pitches should be delivered. In call centers, AI is helping route calls to the right agents to ensure the best customer experience. It also provides coaching after the interaction.

Across the board, AI is becoming an augmentation of the individual, almost like a personal assistant that supports people in their roles. I see a clear shift: AI is no longer a fancy, good-to-have tool. It’s becoming a core part of how businesses operate and how people work.

CDO Magazine appreciates Kalyani Sekar for sharing her insights with our global community.