28,000 Customer Care Reps. One AI That Makes Them All Better. (Verizon)

If you’ve ever called a major carrier to dispute a charge, you know the rhythm. The rep picks up, sounds confident for about eight seconds, and then goes quiet while they search through something. You hear typing. Sometimes a lot of typing. Eventually they come back with an answer that’s either right, or confidently wrong in a way that requires a second call to fix.

Verizon had 28,000 people in that situation every day. Not because they hired bad reps, but because the knowledge those reps needed was scattered across systems that were never designed to be searched quickly under pressure, by someone also trying to hold a conversation. Experienced reps would learned to navigate it over years. Newer ones were doing their best. The gap between those two groups showed up in customer satisfaction scores, first-call resolution rates, and a lot of escalations that didn’t need to happen.

The goal wasn’t to replace anyone. It was to make all 28,000 of them as effective as the best ones. That turns out to be a harder and more interesting problem than simple automation.

The Personal Research Assistant

The solution Verizon built with Google Cloud is called the “Personal Research Assistant”. It runs on Vertex AI and Gemini, and it works the way a good colleague would: listens to the conversation, anticipates what the customer probably needs next, and surfaces the answer before the rep has to go digging. No tab-switching. No hold music while someone searches. The rep asks a natural language question and gets an immediate, contextual response inside the workflow they’re already in.

Verizon also built a second tool called “Problem Solver” for more complex troubleshooting cases. This one was aimed squarely at newer reps. Instead of putting a customer on hold while a junior agent works through a billing anomaly or a network issue, Problem Solver walks through the problem with them in real time. Senior knowledge, available to everyone, instantly.

From July 2024 to January 2025, six months, all 28,000 frontline workers were live. That timeline is worth pausing on. Most contact center AI deployments at this scale take years, involve multiple vendors, and still end up as partial rollouts covering maybe a third of the team. Verizon did the whole thing in half a year.

What 95% Actually Means

The headline number from Verizon’s April 2025 announcement is 95% comprehensive answerability across customer inquiries. In plain terms: nearly every customer who calls about a billing question, a technical issue, or a plan upgrade reaches someone with the right answer on the spot. The hold music is largely optional now.

The downstream effects follow logically. Faster resolution. Better first-call close rates. Less time spent on post-call documentation. Newer reps handling complex cases from week one instead of month six.

And then there’s the number nobody expected: a 40% increase in sales through the customer service channel. Contact center AI is almost always pitched as a cost story. Fewer escalations, lower average handle time, headcount efficiency over time. Verizon’s experience is a useful corrective to that framing. When reps can answer questions about plans, upgrades, and features confidently and quickly, they sell more. The AI isn’t just removing friction. It’s creating opportunity. That distinction matters when you’re building the business case internally.

Beyond the Contact Center

Once the internal deployment was running, Verizon did what most companies say they’ll do and few actually pull off: they kept going. The same AI infrastructure that helps reps find answers now also powers the customer-facing side of the “My Verizon” app, where account holders can ask questions in plain language instead of tapping through nested menus. Same underlying stack, consistent experience, no seams showing.

From there it expanded further. Customers with complex situations got access to dedicated experts backed by the same AI tools. Small business customers got an AI-powered messaging assistant. What started as an internal productivity project quietly became the connective tissue of a much broader customer experience strategy. That expansion didn’t happen because someone wrote a roadmap for it. It happened because the first thing worked, and working things tend to grow.

What ISVs Should Take From This

Verizon isn’t a software company, but the story is directly useful for ISVs building in contact center, workforce enablement, or customer experience. The deployment answers the question that comes up in nearly every enterprise sales conversation: has anyone actually done this at scale, in production, with real people? Yes. 28,000 of them, six months to full deployment, 95% answerability, 40% sales lift. The numbers are public, sourced from a press release, and backed by a brand nobody will ask you to explain.

The architecture pattern is worth internalizing as a product principle too: real-time AI assistance surfaced inside the rep’s existing workflow, not as a new tool they have to remember to open. That distinction is the difference between adoption and shelfware.

The other thing Verizon’s trajectory illustrates is how these deployments tend to expand. You get in with a rep-assist tool, it works, and the conversation naturally moves to the customer-facing layer, then analytics, then automation. Verizon went from an internal pilot to a multi-product AI platform in roughly a year. That’s not a fluke. It’s the playbook.

Want to go deeper?