Building Autonomous Network Operations: From Vision to Measurable Business Impact

At TM Forum DTW Ignite 2026, industry leaders gathered to discuss one of telecom's most important transformations: the shift from network automation to truly autonomous operations. In the session, Building Autonomous Network Operations, Microsoft's Thomas Steagall joined AT&T's Mark Austin, VP of Data Science and AI, and Microsoft's Jamie Gaudette, VP of Networking and Infrastructure, to share real-world lessons from deploying AI and agents in production network environments. Rather than focusing on theoretical architectures, the discussion centered on measurable business outcomes, operational governance, and the organizational changes required to scale autonomy.

One theme stood above the rest: autonomous networks are no longer a future aspiration. They are delivering measurable value today.

TM Forum DTW Ignite 2026 | Building Autonomous Network Operations

AT&T's Agentic AI Journey Starts with Data

According to Mark Austin, the foundation for autonomous operations remains unchanged regardless of whether organizations are implementing traditional AI, machine learning, or generative AI: it starts with data. AT&T's approach has evolved into a four-layer architecture consisting of a Data Layer, Knowledge Layer, Agentic Layer, and User Experience Layer. This layered model enables AI agents to move beyond analytics and actively participate in operational workflows.

The scale of AI adoption at AT&T is remarkable. The company's Ask AT&T platform now hosts 150 Generative AI solutions in production, processing approximately 40+ billion tokens every day, with generative AI representing roughly 85% of the organization's work.

However, the most compelling story is not the technology itself—it's the operational outcomes.

Turning 136 Billion Signals into Customer Outcomes

AT&T processes approximately 136 billion network signals per day to monitor fiber and wireless network performance. Managing this volume of data manually would be impossible. Instead, AT&T is using AI agents to transform network operations from reactive troubleshooting into proactive problem prevention.

One example is Agentic Root Cause (ARC), an AI-powered capability that analyzes charts, trends, telemetry, and operational data to identify the root cause of network issues. Tasks that previously required four hours of engineering investigation can now be completed in approximately one minute.

The operational and financial impact is significant:

• 136 billion network signals processed daily for incident management and monitoring.
• Four-hour troubleshooting processes reduced to approximately one minute.
• More than 3 million truck dispatches avoided through precise fault identification in broadband and fiber networks.
• Approximately 30x ROI achieved over the life of the program.

Perhaps the most impressive outcome is that these operational improvements have enabled AT&T to offer a customer-facing service guarantee. The company now guarantees that customers will experience no more than 20 minutes of downtime on fiber services or 60 minutes on wireless services, or AT&T pays compensation. According to Austin, the guarantee is possible because of the organization's ability to identify outages faster and remediate issues proactively.

"We've gotten so good at this in catching these outages and fixing things ahead of time, we now issue a customer guarantee. We guarantee you're no longer going to be out more than 20 minutes on fiber or 60 minutes on wireless, otherwise we pay." — Mark Austin, VP Data Science & AI, AT&T

For telecom operators pursuing autonomous networks, this is the ultimate proof point: AI is no longer just reducing operational cost—it is directly improving customer experience and supporting new commercial commitments.

Governance Matters: AT&T's RODEO Framework

While excitement around autonomous agents continues to grow, AT&T emphasized that governance must advance alongside autonomy.

Austin introduced AT&T's RODEO framework, a simple but memorable approach for governing AI agents at enterprise scale:

• Register agents to ensure they are identifiable and managed.
• Observe their behavior and actions.
• Discover rogue or unauthorized agents.
• Enforce operational and governance policies.
• On Guard through continuous monitoring and oversight.

The RODEO model reflects an important lesson for telecom operators: trust is built through visibility, control, and continuous evaluation. AT&T's approach is to start with small proofs of concept, rigorously evaluate outcomes, and gradually expand agent autonomy as confidence grows.

As autonomous networks move from experimentation to production, governance frameworks such as RODEO may become just as important as the AI technologies themselves.

Microsoft's Experience: Scaling Networks Without Scaling Headcount

While the AT&T story demonstrates business value, Microsoft's experience provides evidence that similar principles apply at hyperscale.

Jamie Gaudette described Microsoft's challenge of growing infrastructure 10x over the next four years without increasing operational headcount, while managing one of the world's largest backbone networks, which recently surpassed one exabit per second of traffic.

Microsoft discovered that many operational bottlenecks were not technical configuration tasks, but rather the coordination of people, field services, carriers, and repair activities. AI agents proved particularly effective in these workflow-heavy scenarios.

One example is Miles, an AI agent responsible for coordinating the resolution of outside plant fiber cuts. The agent recently handled 200 fiber cut incidents in a single day and now performs these coordination activities 95% of the time without human intervention.

"Build your system in a way where they can work across your whole organization seamlessly, and you can get out of the way, because these things are going to surprise you with what they can do." — Jamie Gaudette, VP Networking & Infrastructure, Microsoft

Importantly, Microsoft maintains strict controls around network changes. Agents can recommend and request changes, but code-based change management systems remain the final authority for executing actions on the network. This "human-on-the-loop" model provides safety and accountability while still enabling substantial operational gains.

Autonomous Networks Require Organizational Transformation

One of the more surprising insights from the discussion was that autonomy is driving organizational change as much as technological change.

Gaudette explained that Microsoft reorganized its operations teams around the concept of humans managing digital workers. Subject matter experts increasingly supervise teams of agents, update skill files, and refine operational knowledge rather than performing repetitive processes themselves. The result is that employees remain central to operations, but their focus shifts toward governance, knowledge management, and continuous improvement.

The vision presented by both organizations is not one of replacing people. Instead, it is about enabling human experts to oversee increasingly capable digital workforces.

The Road Ahead

Looking forward, both leaders expect the next wave of innovation to come from expanding agentic workflows across the enterprise. Austin described applying agents beyond network operations into functions such as finance, HR, and broader business operations. Gaudette emphasized the growing importance of knowledge systems and skills frameworks that allow agents to collaborate across organizational boundaries and optimize end-to-end processes.

The message from DTW Ignite 2026 was clear: autonomous networks are no longer defined by experimentation. The leaders who are creating competitive advantage are those connecting AI directly to operational outcomes.

AT&T's results—150 AI applications, 40+ billion tokens processed daily, 136 billion network signals analyzed, four-hour investigations reduced to one minute, 3 million avoided dispatches, approximately 30x ROI, and customer-backed service guarantees—demonstrate what happens when agentic AI moves from pilot to production.

For telecom operators pursuing TM Forum Level 4 and Level 5 autonomy, the lesson is straightforward: start with a strong data foundation, build knowledge and governance layers, introduce agents into targeted workflows, and measure outcomes relentlessly. Autonomous networks are not a future vision—they are becoming an operational reality today.