Digital transformation laid the groundwork. Now, AI transformation has become a top business priority for 4 in 5 companies. 

Why? 

Because the technology delivers superior customer experiences, optimizes operations, and drives innovation at scale. 

But despite extensive digital transformation across industries, many legacy organizations are struggling with AI transformation. Deeply entrenched, outdated systems persist, locking data and business logic into silos and making AI adoption extremely difficult.

However, one legacy organization that has bucked this trend is Verizon. The 40-year old telecommunications company has made significant headway in its AI transformation. 

In this article, we’ll share insights from Kalyani Sekar, Chief Data Officer at Verizon. Kalyani recently sat down with Snowplow’s CTO and Co-Founder, Yali Sassoon in a video interview for CDO Magazine. Here, we’ll provide you with Verizon’s AI transformation blueprint that Kalyani outlines in the interview. And at the end of the post, you’ll have a clearer idea of how you can tackle your own AI transformation journey. 

The Foundation: Data Quality as the Cornerstone of AI Transformation

Hopefully, this initial learning from Verizon is not new to you. But let’s dissect it a little. 

Kalyani explains: 

"Data quality is foundational to Verizon's long-term goals. As we roll out more and more analytical use cases — from predictive to prescriptive, generative, and agentic — it's important to ensure that these AI capabilities are built on data that is trustworthy and of the highest quality."
Kalyani Sekar, Chief Data Officer, Verizon

Surprisingly, many organizations push ahead with AI transformation without thoroughly considering data quality. In fact, research from Qlik estimates that 81% of companies still struggle with AI data quality. Consequently, the ROI of AI investments and business stability are at risk. 

So, what has Verizon done to ensure data quality? 

Shifting Left: Quality at the Source

One of the most significant insights from Verizon's approach involves "shifting left" with data quality practices. 

“Shifting left” basically means that you implement quality controls at the point of collection. This is as opposed to the common mistake of addressing quality issues after the data reaches the warehouse. 

Kalyani elaborates on this point: 

”We realized that data quality is only as good as the data coming from the source. We really need to shift left—when we talk about data quality, we have to start thinking about it right from the source.”
Kalyani Sekar, Chief Data Officer, Verizon

This approach becomes even more critical when dealing with the multi-modal data sets that modern AI technologies require. By multi-modal data sets, we mean different types of data like structured tables, images, videos, text, and PDFs. 

As Verizon discovered, generative AI and deep learning technologies require organizations to manage these diverse data types in real time. 

‍

‍

Real-Time AI: Powering Seamless Customer Experiences

Verizon’s AI transformation goes way beyond traditional data analytics. It’s about using real-time intelligence to keep the company’s network running smoothly. 

Verizon collects a lot of data every second. This data comes from its network equipment, like cell towers and routers. The company’s AI systems scan this data to find anything unusual, before automatically making changes to improve performance. 

Kalyani highlights this point: 

“Our network devices are constantly sending out data — multiple times every second.  We collect that data in real time, right where it’s happening at the ‘edge’(meaning close to where the data is generated). As soon as we get it, we measure key performance indicators (KPIs), which are just metrics that tell us how well the network is doing. If something looks off — like a glitch or abnormal behavior — the system takes immediate action to fix it. That’s how we make sure customers don’t feel any disruption.”
Kalyani Sekar, Chief Data Officer, Verizon

This learning reinforces the importance of feeding AI systems with high-quality, real-time data. It enables companies to go from waiting for problems to happen and fixing them (reactive), to using AI to spot early warning signs and fix issues before they affect customers (proactive). 

Proactive operations = a smoother service, faster performance, and fewer interruptions. 

‍

‍

Behavioral Data: Creating 360-Degree Entity Views

Verizon’s AI transformation doesn’t just stop at real-time operational data. 

Kalyani also stressed the need to integrate AI with behavioral data. This helps Verizon to create a more comprehensive view of its customers, employees, and network assets. 

“Bringing together core data and behavioral data truly creates a 360-degree view of every entity in the organization. Once we have a complete view of what that entity is doing in relation to Verizon, it really enables the predictive and prescriptive capabilities of AI across the organization.”
Kalyani Sekar, Chief Data Officer, Verizon

What this teaches us is that AI systems require a holistic approach to data. By combining multiple real-time data sets, you can take your AI from simple automation to intelligent orchestration. 

Intelligent orchestration means that artificial intelligence can predict what customers need. It can suggest the best next steps and make decisions on its own within set limits. 

‍

‍

Infrastructure Evolution: From Siloed to Distributed AI Systems

Amongst the companies we speak to, this is one of their biggest headaches. And it’s also one of the most significant blockers to successful AI transformation. 

This issue is what drove Verizon to completely rethink its data architecture and infrastructure management. 

Previously, Verizon had multiple isolated systems that served different business units. Now, it has what Kalyani describes as a:

"Scalable distributed ecosystem that serves the enterprise where we truly see cross-functional data collaboration happening live."
Kalyani Sekar, Chief Data Officer, Verizon

In simple terms, this means Verizon no longer has separate data systems that don’t talk to each other. Now, it has connected systems. Teams from all over the organization can access and work with shared data in real time. 

Managing Modern Data Types

Your infrastructure should connect your data systems. It must also handle different data types like vectors, graphs, and knowledge bases. 

This aligns with what we mentioned before around multi-modal data sets. Kalyani highlights the importance of this point: 

“With the advent of deep learning and generative and agentic AI, there’s a huge need to work with multi-modal data sets. Managing just structured data is the old way of doing things. Now, managing data that includes images, videos, text, and PDFs — that’s the new norm."
Kalyani Sekar, Chief Data Officer, Verizon

Of course, this shift will create a new challenge for your data engineering teams. If they haven’t already, they need to build data platforms capable of storing, processing, and managing diverse data sets in sub-seconds, all while maintaining enterprise-grade reliability. 

The GPU Challenge

When embarking on your AI transformation, Kalyani speaks of the need to efficiently manage GPUs (Graphics Processing Units). 

Engineers originally designed GPUs for video games and graphics. But now, they’ve become essential for AI. This is because they can perform thousands of calculations simultaneously–exactly what AI systems need for training (teaching AI models) and inference (having trained AI models make predictions and decisions). 

As Kalyani notes: 

"The GPU demand was much more than what the supply was. When we look into the value and everything, it is important for us to economically manage GPUs."
Kalyani Sekar, Chief Data Officer, Verizon

GPUs are expensive. They’re in high demand, and you want to make sure you’re using them efficiently rather than sitting idle. 

You therefore need to think about when and how you use these powerful processors. 

Kalyani suggests using automated systems. These systems assign GPU resources to the most important AI tasks. They also turn off resources when not needed to save costs. 

Governance: The Ethical Foundation of AI Transformation

Successful AI transformation programs require robust governance frameworks. This is to ensure your AI systems operate ethically and within regulatory constraints. 

Verizon's approach centers on transparency and accountability. It achieves this through what Kalyani refers to as an “AI registry system”. 

Here's how the system works in practice: before anyone can build a new AI model or modify an existing one, they must complete a detailed proposal that non-technical business leaders can understand.

Kalyani explains: 

"Every time somebody wants to build a new model or somebody wants to modify the existing model, they need to clearly articulate what they want to do in plain business language." 
Kalyani Sekar, Chief Data Officer, Verizon

The proposal must answer four key questions:

1. What business problem are you trying to solve? (described in plain business terms, not technical jargon)
2. What data will you use and how will you transform it? (ensuring data usage is appropriate and legal)
3. How will this AI system make decisions or recommendations? (understanding the logic and potential impacts)
4. Who will use this system and how will it affect customers? (considering real-world consequences)

Once submitted, this proposal goes through a review process involving privacy experts, legal teams, and business stakeholders who evaluate whether the AI system aligns with company values and regulatory requirements.

Navigating Regulatory Complexity

If your organization is in a regulated industry, your AI transformation must follow many compliance rules. As Kalyani notes: 

"Verizon is a regulated organization. We are governed by lots of local, state, federal, and international bodies on a regular basis. Every time when there is a new regulation that comes up, translating that into what it means for Verizon and then translating that to what it means for data is another big art by itself."
Kalyani Sekar, Chief Data Officer, Verizon

Kalyani's main advice is to create a cross-functional governance team. This team should include members from legal, privacy, IT, and business. They can quickly understand new regulations and turn them into clear AI policies. 

By following this advice, you can reduce the chances of delays in your AI projects. This is especially important for avoiding compliance issues that may be discovered late in development. 

The Cultural Shift: AI as Business Integration

AI transformation involves changing how your organization thinks about artificial intelligence. 

Currently, many organizations view AI as a separate technology layer. But successful transformations integrate AI tools directly into business processes and human workflows. 

Kalyani observes this cultural shift: 

“AI becomes part of the systems and gets integrated into day-to-day business. It starts recommending products customers should buy, generating scripts for sales pitches, and even taking calls in the call center—pairing customers with the right person to deliver the best experience."
Kalyani Sekar, Chief Data Officer, Verizon

AI as a Personal Assistant

Verizon aims to create AI systems that act as personal assistants for all employees. Kalyani explains: 

"For every individual working in an organization, AI becomes an augmentation of themselves. AI becomes almost like a personal assistant for that individual" 
Kalyani Sekar, Chief Data Officer, Verizon

Kalyani believes that AI should not replace your people. Instead, give them AI tools that make them better at their jobs.

Give your sales teams AI-generated talking points. Use AI in your call centers to match customers with the right specialist. Let AI handle content creation for your marketing teams while they focus on strategy.

A good way to tackle this, is to start by identifying repetitive tasks that eat your employees’ time. Next, use AI to take care of those tasks. This way, your team can focus on solving problems and building relationships.

‍

‍

Overcoming Common Challenges

Before starting your AI transformation program, Kalyani talks about some challenges Verizon faced. She also explains how the team solved these issues. 

Data Synchronization

Keeping your data synchronized across distributed systems can be a nightmare. But this practice is absolutely essential for AI. As Kalyani explains:

"If the data is not in sync, we always get into a wrong interpretation of data." 
Kalyani Sekar, Chief Data Officer, Verizon

As you deploy more AI systems that consume and act on data in real time, this synchronization challenge becomes more complex. 

Key learning: Establish clear data governance processes and single sources of truth for your critical business metrics before scaling your AI systems.

Quality Definition Alignment

It’s likely you’ll have different teams in your organization who each have varying definitions of data quality. This can create friction for your AI transformation program, as Kalyani explains: 

"The definition of quality at the source is very different from the definition of quality at the consumption."
Kalyani Sekar, Chief Data Officer, Verizon

Therefore, you must ensure consistent quality standards across your entire data pipeline.

Key learning: Align your teams on quality definitions early—what IT considers "clean data" may be different from what marketing needs for AI-driven campaigns.

Scaling AI: From Pilots to Production

When moving from small AI projects to larger AI adoption, you must focus on change management and team alignment. 

Successful AI transformation programs, including that of Verizon's, share several characteristics: 

• Cross functional collaboration between technical teams and business stakeholders
• Investment in AI tools and platforms that democratize access to AI capabilities
• Focus on deploying AI solutions that solve specific business problems
• Continuous measurement and optimization of AI system performance
• Strong governance frameworks that ensure responsible AI implementation

The Economics of AI Infrastructure

You also need to consider your infrastructure economics. Verizon found that the quick shift from traditional machine learning to generative AI has greatly raised infrastructure needs. This is especially true for GPU resources and cloud computing costs.

Kalyani advises organizations to implement FinOps practices to manage cloud economics effectively. This means regularly bringing your finance, operations, and engineering teams together to track, control, and optimize your cloud spending. 

As Kalyani warns: 

"We need to treat FinOps with the same importance as infrastructure—otherwise, costs can quickly spiral out of control."
Kalyani Sekar, Chief Data Officer, Verizon

Key learning: Monitor your AI infrastructure costs closely and implement automated resource management to prevent budget overruns.

Looking Forward: Continuous Evolution

AI transformation is not a destination. It’s an ongoing process for adapting to technological change. 

If your company is just starting out on its AI transformation journey, make use of Verizon’s experience which provides a clear blueprint:

1. Start with data quality 
2. Invest in governance
3. Focus on real business problems, 
4. Prepare for AI to be an integral part of how work gets done across your organization

The question is no longer whether to pursue AI transformation. It's about how fast and effectively your organization can create the foundations needed to scale AI solutions. These solutions should provide clear business value while keeping the trust and transparency that customers and stakeholders expect. Good luck! 

If you’d like to learn more about Verizon’s story, check out the full three-part interview here.