...
Itential Platform Pricing Explore flexible plans and options for your team
Itential logo
Case Study

How Lumen Is Turning Proven Automation Into Faster AI Agent Development

Lumen’s combined Workflow Development & Orchestration teams delivered $22.8 million in value through Q3 2026. Within that program, its work with Itential saved 1.09 million hours in 2025 and supports more than 8 million Autopilot Enterprise jobs a month.

Now Lumen is building on that foundation to help more engineers understand existing automation, develop AI agents, and put their network expertise to work through natural language.

$22.8M
Value Delivered
Across Lumen’s combined Workflow Development & Orchestration teams, Q1–Q3 2026
1.09M
Hours Saved
across Lumen’s Itential deployment in 2025
8M+
Autopilot Enterprise Jobs
processed per month
The AI Foundation

Years of Automation Gave AI Somewhere to Act

Lumen’s automation journey spans service assurance, enterprise IT, Network as a Service, and managed services. Teams turned recurring tasks into reusable workflows, with checks, permissions, approvals, and verification built in.

The deployment saved 1.09 million hours in 2025 and another 802,000 hours through July 2026. Those workflows are now capabilities engineers can reuse and agents can call, within the controls defined for each operation.

Acceleration
395 Production Workflows

Recorded in the October 2026 debrief, with approximately 120 workflows separately identified as tools exposed to AskGreg.

Reach
16 Business Units & 90+ Connected Systems

Supported through Autopilot Enterprise, with peak throughput reported at 52,000 jobs an hour.

Quality
21% Fewer Construction-Related Customer-Impacting Incidents

Reported for service assurance in 2025, alongside the time savings delivered through automation.

Proof in Practice

IP Disconnects Went From an Hour to Five Minutes

The automated process includes checks before and after execution, plus IP address reclamation. Lumen’s deployment overview attributes these 2024 savings to this single use case.

5 Mins
Per IP Disconnect (Down from 75)
Before and after automation
48K
Hours Saved in 2024
From the IP disconnect use case
$667K
Saved in 2024
From the IP disconnect use case
The Impact

Measurable Value Creates Room for More Innovation

Lumen’s combined Workflow Development & Orchestration teams reported $22.8 million in value, 298,000 efficiency hours, and 234 completed projects through Q3 2026, reaching 119% of their annual efficiency-hours goal.

That value breaks down to $13.2 million OpEx, $7.9 million CapEx, and $1.7 million netex. Another 74 projects are in flight and 297 are in the funnel. AskGreg has provided 77,000 answers to 1,700 distinct users.

These results cover the combined teams’ work, not Itential alone, and the program and deployment figures cover different scopes and periods, so they aren’t additive. Together they set up the next phase: more measurable improvement, with more engineers contributing.

When you’re operating infrastructure at Lumen’s scale, the question was never whether AI could help – it was whether we could trust it in production. Itential FlowAI answered that. Our teams were building production-ready agents in minutes, within the same governance and access controls we already rely on. As we build the next digital backbone for AI, this is the next evolution in our journey with Itential – and it’s redefining how we operate networks at scale.
Image of Greg Freeman
Greg Freeman
VP Network & Customer Transformation, Lumen
The Opportunity

7 Inherited Workflows & A New Way to Understand Them

For one Lumen engineer, the opportunity started with work already on his desk. After a reorganization, he inherited seven workflows of varying complexity. Before he could extend them, he had to understand them.

The workflow-to-spec skill in Itential Builder Skills gave him a starting point. He pointed Claude Code at an existing workflow and reverse-engineered it into a Markdown specification that both a person and an agent can read.

With that spec, he can work through the logic with a copilot, ask questions, and plan changes. The knowledge captured in one engineer’s workflow becomes the next engineer’s starting point.

The Agent Approach

Network Expertise Became the Starting Point for AI Agents

The same approach carried into hands-on FlowAI sessions, the first FlowAI experience for most participants. After a short introduction, engineers explored tasks they already knew: checking device state, reconciling inventory, investigating drift, and evaluating configuration changes.

One participant combined a collector, an auditor, and a risk analyzer into a three-agent router assessment. Another agent produced a network map and findings table.

The participation and activity records show how broadly engineers explored:

33
of 41 registered users
built or ran an agent across the workshop records, approximately 80% participation.
31
Engineers
ran agents during the busiest 45-minute lab window.
40
Distinct Agents
produced 149 sessions and 435 tool calls.
9
Engineers
ran more than one agent during that window.

The 40 agents covered seven use-case categories, led by interface and device status (13 agents), inventory discovery and topology (9), configuration changes (7), and drift and source of truth audits (5). Engineers also explored business process workflows, notifications, and reporting.

Their operational knowledge shaped what they asked, which evidence mattered, and how they judged results. Review stayed part of the work: each session was assessed against its task, and proposed changes still needed the appropriate checks and controls.

The Outcome

From Individual Agents to a Coordinated Team

In a separate demonstration, John Capobianco used Claude Code and Itential Builder Skills to create a six-agent team for human-approved loopback provisioning in about 15 minutes.

A coordinator invoked specialists for planning, implementation, testing, communications, and gate review as tools within the platform. The planner built a specific plan from live device evidence. The implementer could apply only the frozen plan, after Jev decision-gate and human approval in Work Center. The tester confirmed the result and that existing interfaces were unchanged.

Because each specialist has one job, a capability like communications can be reused in another project. Moving from lab to production still requires acceptance testing and operational validation. Spec-Driven Development documents that path, from requirements through as-built reconciliation.

6 Agent Team for Human Approved Loopback Provisioning
The Future

A Larger Role for Engineers in Agentic Operations

Lumen’s next chapter builds on what its teams already have: reusable automation, operational controls, and deep network expertise. FlowAI gives engineers another way to reason through a task using established capabilities. Builder Skills help them understand existing work and build new assets in natural language.

Execution stays deliberate. Reasoning helps interpret context and investigate unfamiliar conditions, while deterministic workflows handle known, repeatable tasks. Teams combine both as they define how an operation should run.

The inherited-workflow example and the hands-on sessions point the same way: Lumen extending its automation foundation through the people who know its infrastructure best.

Build on the Automation You Already Trust

Explore how FlowAI connects AI reasoning to governed infrastructure execution, and bring your own operational use cases to the VibeOps Forum.

Keep Learning

The Latest in Agentic Operations

Get Started

Agentic infrastructure operations starts here.

See how Itential connects AI reasoning to governed execution across your entire infrastructure.