Verizon’s customer service and sales function runs through a roughly 28,000-person team that fields support calls and, increasingly, sells products and plans within those same interactions (Reuters, April 2025). By early 2025, generative AI was already capable of looking up account details, drafting responses and resolving routine issues faster than a human could search Verizon’s internal systems by hand. At the same time, Verizon was losing subscribers after a round of 2025 price increases, intensifying pressure to cut costs in exactly the part of the business — large, headcount-heavy customer service — that looks easiest to automate (CX Network, August 2025). Other companies in the same position had already chosen to automate that function outright: Klarna, for one, replaced roughly 700 customer-service agents with an OpenAI-powered chatbot in 2022–2024 (Reuters, April 2025). Verizon’s leadership had to decide whether to follow that path or build something else.

Two paths were available. Path A: replace the team — is the route Verizon’s own leadership explicitly named and rejected when describing its strategy: “Its efforts contrast with some companies, like Swedish payments group Klarna, which have opted to use AI to replace customer service staffers,” Reuters reported, and Sowmyanarayan Sampath, CEO of Verizon’s consumer group, said there are “much easier ways” to reduce costs than firing the people who deal with customers (Reuters, April 2025). Path B: the route Verizon chose — put a Google Gemini-based assistant on every representative’s screen to handle the lookup-and-paperwork side of a call, freeing the person on the line to spend more time on the conversation itself. Sampath described the shift in his own words: “We are doing reskilling in real time from customer care agents to selling agents” (Reuters, April 2025). Verizon’s leadership made the same case publicly a few months later, when its own research into customer sentiment came back: “The future of CX isn’t about AI replacing humans, but about using AI to make human interactions better,” said Daniel Lawson, SVP of Global Solutions at Verizon Business, in the company’s CX Annual Insights report (Verizon, August 2025).

Verizon began deploying its Google Gemini-based assistant to customer service representatives in July 2024 and ran it at full scale across the 28,000-person team by January 2025 (Reuters, April 2025). The system was built by feeding nearly 15,000 internal documents into a customized version of Google’s Gemini large language model, giving representatives an on-screen tool that looks up account details, suggests likely solutions and drafts the next response so they no longer have to search manually through internal systems while a customer waits on the line (Reuters, April 2025). A second, customer-facing layer followed in June 2025: a “Customer Champion” team built around the same Gemini models, intended to resolve issues in a single contact rather than the multi-call loop common to telecom support, plus an upgraded My Verizon app, expanded live-agent hours and 24/7 live chat (Verizon, June 2025; Yahoo/CNET, June 2025). In a CNET interview, Sampath gave a concrete example of the kind of case the AI is designed to resolve before a human ever needs to get involved: “You send your phone back. It gets lost in transit. We know it’s lost in transit. We don’t need the customer to call. We use an agentic framework to automatically solve the case and then give them credit” (Android Headlines, June 2025). The assistant is also designed to hand a customer to a human representative whenever the request is complex or the customer asks for one, rather than trying to resolve every case on its own (Android Headlines, June 2025).

The system was tested in public on January 14, 2026, when a nationwide outage affected more than a million Verizon customers (Inc., January 2026). Rather than routing the flood of angry replies on the Verizon Support X account through a bot, the company kept named human agents — Sean, Kristen, Jeremy, Sarah, Latoya and others — visibly answering customers one by one, even when the replies were necessarily repetitive (Inc., January 2026).

By Verizon’s own description, this is primarily a reskilling case: Sampath’s own language was “reskilling in real time from customer care agents to selling agents,” a genuine change in what the job is for, not just how well it’s done (Reuters, April 2025). It sits close to the upskilling definition too, though: the same 28,000 people kept their jobs, their team and, for most of them, their job title, and what actually changed was the mix of tasks within that role once AI took over the lookup work. The honest summary is that this case has elements of both — a role whose core purpose shifted (reskilling) without anyone having to change desks (the structural marker of upskilling) — and Verizon’s own description leans toward calling it reskilling.

The more interesting HI signal is the contrast Verizon’s own leadership chose to draw, unprompted, with a competitor. Most companies in this report only get compared to an AI-first alternative by outside analysts; Sampath did it himself, naming Klarna’s chatbot-replacement strategy as the road not taken and stating plainly that there were “much easier ways” to cut costs than removing people from customer service (Reuters, April 2025). That a telecom CEO would say this in the same breath as reporting a 40% sales increase suggests the HI framing was not just a defensive talking point but the actual rationale: AI freed capacity, and the company chose to spend that capacity on people doing more, not fewer people doing the same.

The January 2026 outage is the case’s clearest piece of evidence, because it is the one moment Verizon did not control for effect. Under real pressure, with more than a million angry customers, the company did not lean on a bot to absorb the volume — it put named humans in front of every reply, and its own survey data explains why: customers are categorically more satisfied with people than with AI in exactly this kind of moment (Verizon, August 2025; Inc., January 2026).

What makes the case worth watching rather than closed is the shift in tone since Dan Schulman became CEO. Where Sampath drew a bright line against replacing people, Schulman has spoken openly about AI displacing “a large percentage” of customer service roles and has already begun experiments in which agents are doing the work people used to do, not just assisting them (CX Dive, June 2026). Whether Verizon’s HI architecture — AI handles the rote work, humans own the relationship and the moments that matter — holds as a permanent design choice or proves to be a transitional phase on the way to deeper automation is, as of mid-2026, still an open question rather than a settled result.