Blog | Self-healing networks and the race to level four: what Verizon and Google are building together

At DTW Ignite 2026 in Copenhagen, TM Forum CEO Nik Willetts sat down with Muninder Singh Sambi, VP Product Management and Supply Chain at Google Distributed Cloud, and Abhitabh Kushwaha, AVP AI Networks at Verizon, to talk about one of the most ambitious engineering programs in the industry right now: using temporal digital twins turn AI ambition into AN reality.

The problem no dashboard can solve

Verizon's starting point is brutally honest. Network operations, as they exist today, are reactive. Engineers wait for tickets, then investigate. Dashboards show you what happened, not what's about to happen. And with the explosion of wireless complexity — Wi-Fi, prepaid, postpaid, and now satellite all handing off to each other — manual operations simply can't keep up.

"We are moving into a crisis of complexity," said Abhitabh. "Therefore, we have to have an AI-first strategy for network operations."

That's not a future aspiration for Verizon. It's a company mandate. And the scale of what they're building to deliver on it is significant: a full replica of their network, connecting RAN, transport, and core, built in partnership with Google. Not a static model, but a living, real-time digital twin that evolves as the network evolves.

Why temporal matters

Most digital twins are a snapshot in time. That's useful — but it's not enough. Networks are living, breathing entities. Upgrades happen constantly. Infrastructure gets added. New cores come online. A snapshot tells you where you were, not where you are.

The temporal digital twin Verizon and Google are building changes that. It captures the network as it evolves — within seconds — creating a single source of data foundation that isn't siloed between operational and relational databases. As Muninder explained, traditional data copy approaches don't scale, introduce latency, and become prohibitively costly. When you're trying to achieve Level 4 autonomy, the ability to detect and fix network issues in seconds is non-negotiable.

What Google is bringing

The technical backbone of this work is Google's Graph Neural Network (GNN) — a capability built on graph technology developed by Google researchers. What GNN brings is the ability to map the massive, constantly evolving web of network dependencies in real time. As infrastructure gets added or updated, the model keeps pace. That's what makes self-healing possible at Verizon's scale.

Alongside the digital twin work, Google launched its Gemini Enterprise Agent platform, including an autonomous Guardian agent — a set of agents that CSPs can deploy for domain-specific or cross-domain network operations. The design is deliberate: these agents are built to align with TM Forum's standards architecture, not to create new proprietary systems on top of the complexity operators already carry.

As Muninder put it plainly: "The future has to be standardized."

Open, interoperable, and built for the industry

What makes this work genuinely significant beyond Verizon is the commitment to openness. The framework being developed — built on TM Forum standards including the digital twin guidance of IG 1488 and the autonomous networks validation framework — is designed so any CSP can follow it. Verizon and Google are also co-creating a Business Intent Extension Model (BIEM) and contributing it back to the industry.

Inside the Catalyst project itself, three agents running on Google Cloud are interoperating with agents running on Ericsson's on-premises cloud. Multi-cloud, multi-vendor, fully agentic — and doing it in production code, not a slide deck.

The numbers that make the case

The expected business impact is hard to ignore. Verizon is projecting up to 35% cost savings in operations and maintenance — with knock-on improvements to customer experience as the self-healing network catches and resolves issues before customers feel them.

But Abhitabh was clear that the bigger shift is what it does for people, not just processes. Engineers who are currently spending their time as data miners get to become decision makers instead — guiding agents, setting priorities, applying judgment where human ingenuity matters. That's the workforce multiplier effect that agentic AI makes real.

Trust is the gate

None of this scales without trust. And trust doesn't come from the technology alone — it comes from humans gaining confidence through experience. At Verizon, around 163,000 employees have access to Gemini Enterprise. Of those who have engaged with it, more than 80% report increased confidence after using it.

The path to full autonomy runs through humans in the loop first. Certain actions get trusted, then delegated. Confidence builds. Autonomy expands. As Abhitabh put it: "The world's not going to move to autonomous overnight."

That's not a limitation. That's the plan — and it's the right one.

The Race to 2030 is a race to do this well, not just fast. What Verizon and Google are building together, in partnership with TM Forum, is proof that "well" and "fast" aren't mutually exclusive.

Watch the discussion in full: The Industry’s First Temporal Digital Twin: Turning AI Ambition into Autonomous Network Reality