- Verizon rolled out Claude Code to its full 33,000-person technology team in January to push toward Level 4 network autonomy
- A tool called WatchTower cut simulated outage remediation in midtown Manhattan from hours to seconds
- CTO Yago Tenorio says Verizon made 70 million automated configuration changes last year
Verizon is nearing Level 3 autonomy and is on the road to Level 4, thanks in part to a boost from Claude Code, CTO Yago Tenorio told Fierce.
Verizon's goal is to build a network "to serve as the preferred connectivity foundation for the world's most responsive AI solutions," Tenorio wrote in a blog. "We are moving away from traditional hardware constraints to turn our infrastructure into a software-defined, self-organizing reasoning engine that optimizes itself in real time intentionally."
He added in the post the operator is "pushing hard toward Level 4 autonomy in critical segments of our core network."
Turning everyone into a software developer
Claude Code is important to Verizon's network transformation. In January, the operator rolled out Claude Code to its entire 33,000-person technology team, Tenorio said. Previously, only a smaller group of 500 application developers had access to Claude Code.
"We decided that it would be a good thing if we could use Claude Code to turn everyone into a software developer," Tenorio told Fierce. With Claude Code, Verizon's technology team is developing tools for automation, engineering, operations, financials, capital deployment, modeling, forecasting and capacity planning. Code developed with Claude goes through the same quality control and safety processes as other code used by Verizon.
To truly achieve an autonomous system to run the network, Verizon needs a "common layer of good, real-time quality data" and a "common repository of skills, tools and agents that is shared across the organization," Tenorio said.
Seeing what the network can't see
"We migrated to device telemetry based not just on what the network sees, but what the network doesn't see as well," he said. A customer with no signal in a Manhattan basement was previously invisible to the network, which had no record the connection attempt happened. Device-level data closes that gap — and if one phone can't report a dead zone in real time, another recovering signal in the same spot will. "We don't care about the identity, we delete that," Tenorio said. "We care about the characterization of a network in a space."
But Tenorio would not be pinned down on a timeframe on achieving Level 3 and Level 4 autonomy. "I don't care a lot about labels, I care about process outcome, I care about making an impact for our customers," he said.
Tenorio showed Fierce a tool called WatchTower ("I don't like the name," he noted), demonstrating simulated remediation of a network outage in midtown Manhattan in 90 seconds, down from five to eight hours last year.
"And that's how peace in Gotham was restored," the demonstration concludes. ("I love this part," Tenorio said.)
A crowded race toward autonomy
Verizon is entering a race for autonomy in which a lot of players are already running.
T-Mobile has claimed "Level 4.5" autonomy for its new AI-powered Dynamic CX platform, tied to World Cup stadium readiness. Meanwhile, a TM Forum report found the industry is moving past years of "tinkering at the edges," with China Mobile citing a 30%-plus reduction in fault and complaint repair times after reaching Level 4 in its network operations centers. But most operators are far behind: Accenture found that 79% of telco networks remain at level 0 or 1 automation, with only 22% expecting to reach level 4 by 2030.
AWS, meanwhile, has argued that telcos won't see real ROI until they stop automating one process at a time and connect those efforts into domain-level autonomy.
To achieve its AI goals, Verizon has changed how it approaches network architecture.
A network was previously static. "Engineers calculated its raw capacity, deployed fixed hardware infrastructure and reacted to issues after the fact through flashing terminal alarms and reactive troubleshooting dashboards," Tenorio wrote in his blog post.
Verizon's network was previously automated using closed-loop Python scripts, Tenorio told Fierce. Now, "AI is becoming the operational control system of the network itself," he wrote on his blog.
Automation isn't autonomy
"There is a sharp distinction between running automated scripts on a rigid schedule and true, decentralized autonomy," Tenorio wrote. "Autonomy means building an intelligent system that can reason about an unexpected, complex situation it was never explicitly programmed for, isolate the hidden root cause, and execute an architectural solution completely out of the human loop."
Nonetheless, automated scripts have gone a long way to optimizing Verizon's network. Last year, closed-loop automation executed more than 70 million network configuration changes autonomously. Tenorio noted in his blog that such changes would have required thousands of technicians and labor hours if implemented by hand.
Those changes were programmatic — rule-based scripts rather than reasoning agents — Tenorio clarified in the interview. "There were no tokens in the loop this year," he said. Now, AI increasingly writes that code, but the scripts themselves don't reason. He cast the scripted layer as the foundation for what comes next: "If it wasn't for that, I don't think we would be where we are today talking about putting tokens in the loop."
The road to 6G
Verizon is doubling down on generative AI to reach Level 4 autonomy. Teams are using natural language for network optimization and operations. Engineers are no longer writing custom code, but instead defining high-level architectural structures — and the AI does the rest.
Looking ahead to 6G in 2029 or 2030, "this intelligent infrastructure will achieve full physical, contextual awareness," Tenorio wrote. Future wearable devices will be able to sense and interact with the real-world environment in real time, he said.
To achieve those goals, open standards like O-RAN and "deep ecosystem interoperability" are essential, Tenorio wrote. Verizon is already "running multi-vendor AI orchestration applications simultaneously across live production platforms," he said.