YOU ARE AT:AI InfrastructureBookmarks: Network automation — six levels, five blockers, four questions

Bookmarks: Network automation — six levels, five blockers, four questions

Editor’s note: I’m in the habit of bookmarking on LinkedIn, books, magazines, movies, newspapers, and records, things I think are insightful and interesting. What I’m not in the habit of doing is ever revisiting those insightful, interesting bits of commentary and doing anything with them that would benefit anyone other than myself. This weekly column is an effort to correct that.

Compute, storage, and curated data innovation are converging in a way that unlocks a new class of AI-enabled autonomy that will only be supercharged once agent-to-agent collaboration becomes routine. The center of gravity is shifting from human-in-the-loop to human-in-command; agents act toward goals while people set intent and oversee outcomes. Add the rise of lightweight, task-specific “ephemeral” agents, and you get horizontal scale and more complex reasoning without monolithic systems.

On paper, communications service providers (CSPs) are advantaged in that they’re data-rich, operationally proven, sovereignty-minded, and positioned to compress time-to-value by augmenting people, process, and platforms. Done right, this is the path to efficiency, but also the path to growth. That said, TM Forum has found that more than 50% of AI experimentation doesn’t make it through to production. The business cases are hard to parse, and organizations generally lack the controls, guardrails, and trust needed to truly become AI-first. 

How to cross that chasm was the big theme at TM Forum’s DTW Ignite event held each summer in Copenhagen, Denmark. Answering that question starts with using a pragmatic maturity ladder to identify your current state and target your future state, confront five recurring blockers head-on, and adopt four hygienic questions as a decision-making framework. Oh, and keep in mind that the technological complexity represented by the shift to AI-first network automation is likely subordinate to the change management program that will make or break success. 

Where we are and where we’re going

TM Forum, through its Autonomous Networks (AN) Project provides a useful six-level model ranging from Level 0 (manual operations and maintenance) to Level 5 (cognitive, closed-loop autonomy across services and domains) for self-assessment and strategy development. TM Forum describes each level: 

  • Level 5 — Fully autonomous network: The system has closed loop automation capabilities across multiple services, multiple domains (including partners’ domains) and the entire lifecycle via cognitive self-adaptation
  • Level 4 — Highly autonomous network: In a more complicated cross-domain environment, the system enables decision-making based on predictive analysis or active closed-loop management of service-driven and customer experience-driven networks via AI modeling and continuous learning.
  • Level 3 — Conditional autonomous network: The system senses real-time environmental changes and in certain network domains will optimize and adjust itself to the external environment to enable closed-loop management via dynamically programmable policies.
  • Level 2 — Partial autonomous network: The system enables closed-loop operations and maintenance for specific units under certain external environments via statically configured rules.
  • Level 1 — Assisted operations and maintenance: The system executes a specific, repetitive subtask based on pre-configuration, which can be recorded online and tracked, in order to increase execution efficiency.
  • Level 0 — Manual operations and maintenance: The system delivers assisted monitoring capabilities, but all dynamic tasks must be executed manually.

The majority of CSPs engaged in self-reporting through TM Forum place themselves around Level 2, however there are a growing number of examples of Level 4, including China Mobile, China Telecom, TDC Net and Telefonica Vivo, and others. More on that later. As deeply inter-related business and operational imperatives push CSPs to continue climbing the ladder, TM Forum’s Guy Lupo, mission lead for AI and Data Innovation, laid out five blockers that halt the all-important transition from experimentation to production. 

  1. Readiness: getting a coherent cross-functional program up and running. 
  2. Financial justification: proving ROI in a dynamic, nascent space. 
  3. Solution scoping and architectural fit: integrating the new with the old. 
  4. Operational acceptance: to derive value from automation, people have to trust automation. 
  5. Engineering accelerators: missing platforms, tooling, and talent are a fundamental impediment to time-to-value. 

“We’re tripping on all of them,” Lupo said of the blockers. He outlined four “hygienic-level questions” CSPs need to answer in order to make decisions and keep moving. 

  1. How do you get the right data to the right agent? “Don’t boil the data ocean,” he said. Rather, focus on the minimum viable data for the minimum viable agent. 
  2. How do you get the right tool to the right agent? Expose the necessary capabilities but with appropriate guardrails and security protocols in place. 
  3. How do you get the right model to the right agent? This is about matching capability and context to tasks.
  4. How do you monitor agent-to-agent and agent-to-person interactions? 

Answering these questions and addressing the blockers then, Lupo said, “That is where we open the floodgates to all these use cases to start moving from experimentation to production.” And remember, AI and data are horizontal plays but CSPs don’t typically have horizontal operations; getting rid of data and operational silos is crucial to success in the AI era. 

And ultimately, Lupo said, leadership “is at the heart of it. AI, unlike functional areas of a business, is everyone’s. So it’s the business of the CEO…Without the focus of the CEO to make sure it happens across the business, it tends to become silos.” His advice is to “take the buzzword bingo out of the boardroom and actually put financials instead.” 

The bottom line here is that breakthroughs in network automation will be technical but also, and potentially more importantly, organizational. With a clear maturity target, a plan to neutralize those five blockers, and a disciplined approach to decision-making, CSPs can move from experimentation to production, and convert autonomy into tangible positive business outcomes. 

Here’s the full interview with Lupo from DTW Ignite.

Swisscom and Netcracker take on the blockers to drive network automation domain by domain

Before getting into the Swisscom and Netcracker’s award-winning work, I’ll point out that Lupo was largely discussing agentic AI strategy and implementation; however, his points are still very relevant in a pre-agentic context because the same careful considerations are necessary to build a foundation for the introduction of new technologies.

Recall the five blockers and change management imperative described by Lupo. Swisscom, through extensive work with Netcracker, won TM Forum’s 2025 excellence award for pioneering autonomous network operations. This collaboration, as described in detail in an excellent interview with Appledore Research Consulting Analyst Robert Curran, demonstrates how to address the blockers, how to answer the hygienic questions, and how autonomy is won through combining technological investment with organizational redesign. 

Rudolf Strijkers, enterprise and security architect with Swisscom, took on the blockers. 

On readiness: “We have to rethink completely the way that we work,” he said, describing how a separation of operations from development made room for inefficiencies, and how interdependencies between groups slowed delivery. “We really had a fundamental transformation. We had to think of how process, how people and how IT should actually be working together.”

On financial justification: “There we had to learn a lot because we don’t do these kind of transformations just to reach a beautiful architecture. We need to do this to reach business goals.” He stressed that lifecycle events are the opportunity to push transformation forward and recommended that this not be taken for granted. “Financing can only be done over these lifecycle events…Use them and use them very wisely.” 

On solution scoping and architectural fit: Circa 2014, “We were basically stuck…We didn’t consider the IT complexity that was involved…When you have a very complex IT infrastructure, you cannot just draw it on a whiteboard. You need other mechanisms…We needed to understand how we manage complex architectures…This was about aligning and coordination and about finding the right architectural frameworks in which we can work together.” 

On operational acceptance: “This is by looking at the backlogs of the teams. If there are a lot of tasks going back and forth…then you don’t have autonomy…Being autonomous means that we can do innovations in the domains without affecting other domains. We can introduce new elements like AI.” 

On engineering accelerators: “It’s super important to have this architectural framework ready because when these [lifecycle] events come, you need to be able to use them to move towards these operational domains…Autonomy means you need to be able to implement your business logic without having to wait for other teams.” 

Here it’s worth zooming in on how Swisscom thinks about “domains.” Strijkers explained that the first step in defining a domain is identifying where and why one team may need input or output from another to achieve a goal. Then within the identified domain, it’s necessary to essentially take inventory of the relevant IT systems which are then grouped together into a CI/CD pipeline. Swisscom started this process in the network domain and came out the other side with 20 operational domains within the network mega-domain. Today, he said, “We have let’s say [an] almost complete picture…And of many of those domains we also achieved organizational alignment, and that means not only there is an architectural blueprint available for the domain, it also means it’s mapped to an actual organization.” 

The ultimate end of network automation is agility which means organizational decomposition

Netcracker’s Sue White, head of strategy and portfolio marketing, gave some color on the work with Swisscom and the sheer ambition of the project. “There are different approaches of course. Some operators are looking to modernize their IT but maybe not decompose into these domains. But the problem then is you’re never really going to get that agility.” Organizational and technological interdependencies that aren’t fully understood result in one person waiting for another person to send an email so that first person can push a button (my reductive description, not hers). But back to White: “I think the other aspect is when we work with some operators, [they are] not really looking beyond that first domain…That piece is something we did differently with Swisscom…to align on how to get from the current network to the future network not just for one domain but for the entire group…I think that brought a lot of benefits later on.”

As Strijkers mentioned, no CSP is undertaking transformation for the sake of architectural elegance. It’s about making more money while spending less money, and delivering on core business objectives. That means being able to measure and quantify outcomes. In Swisscom’s case, they’ve reduced incident after change by 25% in the IP transport domain; they’ve reduced mean time to repair by 50%; and they can roll out new features within two weeks. These are the kinds of KPIs that set a flywheel spinning and keep financing coming to get it spinning faster. 

In terms of what he’s learned that other operators may keep in mind, Strikers channeled advice “that’s maybe thousands of years old: know your landscapes.” With that knowledge, develop architectural frameworks to give you the orientation needed for planning, organization, design and decoupling. And, “There is organizational alignment necessary to make this really successful…Consequently, there must be a very close buy-in of the CEO and the leadership.” 

ABOUT AUTHOR

Sean Kinney, Editor in Chief
Sean Kinney, Editor in Chief
Sean focuses on multiple subject areas including 5G, Open RAN, hybrid cloud, edge computing, and Industry 4.0. He also hosts Arden Media's podcast Will 5G Change the World? Prior to his work at RCR, Sean studied journalism and literature at the University of Mississippi then spent six years based in Key West, Florida, working as a reporter for the Miami Herald Media Company. He currently lives in Fayetteville, Arkansas.