What problem does it solve?
Telecom networks have become too complex to run by hand. A 5G network mixes several radio layers, cloud native core functions, transport, many vendors and new services such as network slices with their own performance promises. Vodafone notes that manual tuning and step by step automation scripts are no longer enough as networks grow more complex, and Deutsche Telekom puts it plainly: traditional rule based automation falls short in addressing real time challenges.
The result is that engineers spend their time on repeated manual checks, tuning and firefighting. When a large event fills an area or a server fails, the right fix is often known, but applying it depends on someone noticing and acting across several vendor tools that do not share data, a barrier Telstra describes in its own network. Vodafone, Google Cloud and TM Forum describe the goal as a shift from manual, reactive operations to intent based autonomy, where operators state the outcome and the network works out and executes the actions.
How does it work?
- Express the intent. Operators set target outcomes (availability, latency, throughput, energy) per service, area or customer, instead of individual configuration steps.
- Observe. Agents continuously collect performance, alarm, inventory and external data, such as public event listings, and build a live view of each service across domains.
- Analyse and decide. Specialised agents detect deviations or predict them, diagnose the likely cause, and select actions, often testing them first in a digital twin.
- Act within guardrails. Low risk, reversible actions (reallocating resources, adjusting parameters, moving workloads to healthy hardware) run automatically; major changes go to an engineer for approval.
- Document and learn. Every action and its outcome is logged, explained and used to improve future decisions, and the closed loop checks that the intent is met again.
- Audience
- Back office
- Autonomy
- Supervised agent
- Adoption
- Emerging
- Channels
- API and system to system, Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Interactions handled | Not pooled | 100 to 2.5 million | 2 | 2 organization |
| Cycle time reduction | Too few to pool | at least 95% | 1 | 1 organization |
Value drivers: Speed and cycle time, Customer experience, Lower cost to serve, Employee productivity.
Indicative value
A mobile operator with 300 staff in network operations and optimisation
USD 1.2 million to USD 5.4 million
Operations capacity released per year
How this is calculated
Formula: staff * costPerFte * automatedShare. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Staff in network operations and optimisation staff, full time employees | 300 | 300 | The reference operator. |
| Fully loaded cost per employee costPerFte, USD per year | 80,000 | 120,000 | Editorial assumption. Replace with your own cost. |
| Share of routine operations work taken over by closed loop automation automatedShare, fraction of working time | 0.05 | 0.15 | Conservative. Deutsche Telekom reports a more than 95% reduction in the time to manage major events, but that covers one task type, not all operations work. |
What it leaves out: Staff capacity only. It leaves out faster recovery from incidents, better service level performance, revenue from assured services such as network slices, energy savings, and the substantial investment in data, cloud platforms and integration that autonomy requires.
Market estimates (analyst estimates, not deployments)
- Vodafone, Google Cloud and TM Forum cite an STL Partners estimate that the potential upside of autonomous networks is circa USD 800 million per operator annually. Vodafone, Google Cloud and TM Forum Unveil Framework for Self-Optimising Autonomous Networks (2026)
Who already uses it?
6 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Telstra
Australia · Telecommunications · 2026
In a proof of concept in a live telco cloud environment, Telstra showed an agentic AI capability that detected an unplanned infrastructure outage and resolved it autonomously by moving critical network applications to healthy hardware in minutes rather than hours. Using the Model Context Protocol with retrieval augmented generation, AI agents connect data from multiple vendor platforms into one view and give teams context aware recommendations to speed up fault resolution. Telstra presents it as groundwork for self healing, self optimising operations, not yet as a production service.
No outcome disclosed.
Deutsche Telekom
Germany · Telecommunications · 2025
Deutsche Telekom's RAN Guardian Agent, built with Gemini models on Google Cloud, went live in its German mobile network in November 2025. It is a multi agent system: one agent finds upcoming public events from public sources, another assesses whether nearby cells can carry the expected traffic and monitors them live, and a third executes corrective actions such as reallocating resources or adjusting configuration, documenting every action. It is being extended to the Czech Republic and Croatia. In February 2026 Deutsche Telekom announced MINDR, which applies the same approach end to end across radio, transport and core domains, with first production releases planned for later in 2026.
- Cycle time reduction: at least 95%, live operations, major events
"And in live operations it has reduced the time needed to manage major events from hours to around a minute, a more than 95% improvement."
Claimed by: organization - Interactions handled: at least 100, first month after launch, Christmas market events
"Since its launch in November 2025, RAN Guardian Agent has autonomously triggered over 100 remediation actions at Christmas market events during its first month."
Claimed by: organization
Telstra
Australia · Telecommunications · 2025
Telstra's SmartFix system is embedded in its network operations and automatically fixes many issues before customers notice a problem. Telstra reports the number of proactive actions it performed in FY25 and says they prevented nearly 1 million support calls. The blog post is part of Telstra's description of its wider AI program and gives no detail on the models or the types of fixes.
- Interactions handled: 2.5 million, FY25, proactive actions
"It automatically fixes many issues before customers notice a problem – in FY25 it performed 2.5 million proactive actions, preventing nearly 1 million support calls by resolving issues in advance."
Claimed by: organization
du
United Arab Emirates · Telecommunications · 2025
du and Nokia implemented a 5G Advanced autonomous network slicing solution that continuously measures slice performance and adjusts radio policies by itself, so du can guarantee capacity or low latency for enterprise, event, gaming and broadcasting customers. Machine learning keeps an enterprise customer's capacity intent in all network conditions and enforces low latency slice policies for premium gamers in crowded areas. The companies present it as an industry first implementation; no operational results are published.
No outcome disclosed.
stc Group
Saudi Arabia · Telecommunications · 2024
Nokia deployed its MantaRay Cognitive SON, an AI powered feature of its self organizing network platform, in stc's commercial network in Saudi Arabia for the first time. The system optimises radio parameters autonomously. Nokia reports that during a period of high traffic it processed more than 10,000 actions, raised the utilisation rate of loaded cells by about 30 percent and average user throughput by 10 percent while traffic rose 40 percent, and that it reduced manual work. The results are stated by the vendor.
No outcome disclosed.
KDDI
Japan · Telecommunications · 2022
KDDI deployed Nokia's AVA Performance Degradation Detection and Resolution (PDDR) solution nationwide to monitor its 4G and 5G radio network around the clock. The model detects performance degradations that raise no alarm, so called silent cells, classifies the likely root cause, and hands recoverable cases to KDDI's own recovery system, which tries to fix them automatically. Recovered cells feed back into the training data. KDDI started on 4G in 2019 and extended the system to its 5G NSA network in 2021.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Real time performance, alarm and configuration data across radio, transport and core
- Accurate inventory and service topology
- History of past incidents, actions and outcomes
- External context such as event calendars, weather and planned works
Systems to integrate
- Network controllers, SON platforms and orchestration systems per domain
- Network data lake or data fabric
- Digital twin or simulation environment
- Trouble ticketing and change management
- Service and slice management systems
Complexity: High
Autonomy needs a unified data foundation across vendors and domains, programmable network interfaces, a reliable inventory, a way to test actions safely and a governance model that engineers trust. Most operators get there one closed loop at a time.
- 1
Pick closed loops with bounded risk
Start with loops where the action is reversible and the benefit is clear, such as capacity adjustments for planned events or moving workloads away from failed hardware.
- 2
Build the shared data foundation
Agents need one view across domains and vendors. Invest in the data layer and inventory before adding more agents.
- 3
Define intents and guardrails together
Write down the target outcome, the allowed actions, the limits and the approval rules for each loop, with the engineers who will be accountable.
- 4
Prove it in shadow and in a twin
Run agents in recommendation mode and test their actions in a digital twin before letting them act on the live network.
- 5
Raise autonomy step by step
Move each loop from recommend to act with approval to act within limits as measured accuracy allows, and keep an easy way to switch back.
- 6
Coordinate the agents
As loops multiply, add coordination so agents in different domains do not work against each other, and keep one audit trail.
Guardrails
- Explicit allow list of actions per agent, with limits and rollback
- Human in the loop approval for major network changes and for any action outside the limits
- Automatic stop and rollback when service indicators degrade after an action
- Every action logged with its reason, data and outcome, and explainable to engineers
- Change freezes and maintenance windows enforced for autonomous actions
KPIs to instrument
- Time to detect and resolve deviations from intent, per loop
- Share of events handled end to end without human action
- Actions rolled back and incidents caused by automation
- Service level attainment for assured services
- Engineer time spent on routine tasks
Human in the loop
Engineers define intents, allowed actions and limits, approve major changes and review the audit trail. Autonomy levels are set per loop and raised only on evidence. Operations leaders can pause any agent instantly.
Common failure modes
- Agents that fight each other
- A radio agent and an energy agent undo each other's changes. Coordinate loops and give them shared intents.
- Autonomy without a trusted data foundation
- Agents act on stale inventory or partial data. Fix data and topology first.
- Unexplained actions
- Engineers cannot see why an agent acted and stop trusting it. Require explanations and full logs.
- A small error at network scale
- One wrong automated change is repeated across thousands of elements. Use canaries, rate limits and automatic rollback.
What are the risks and rules?
EU AI Act
Depends on design
Annex III point 2 lists AI systems intended as safety components in the management and operation of critical digital infrastructure as high risk; Recital 55 ties this to the digital infrastructure in the Annex to Directive (EU) 2022/2557, which includes providers of public electronic communications networks. Recital 55 defines such safety components as systems that directly protect the physical integrity of the infrastructure or the health and safety of persons and property, and excludes components used solely for cybersecurity. Loops that only optimise performance or capacity are usually not safety components, but a loop that protects physical integrity or life safety services can be, so operators should assess each closed loop and document the outcome.
Rules that apply
Guidance
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 2 covers AI systems intended as safety components in the management and operation of critical digital infrastructure.
- Recital 55, safety components of critical infrastructure (European Union, Europe). Explains which critical digital infrastructure is meant and what counts as a safety component, and excludes components used solely for cybersecurity.
- NIS2 Directive, securing network and information systems (European Commission, Europe). Risk management and incident duties for telecom operators that also apply to automated operations.
Controls to put in place
- Inventory of autonomous loops with owners, intents, allowed actions and autonomy levels
- Pre deployment testing in a digital twin or staging network
- Complete, tamper evident audit trail of agent actions
- Kill switch and rollback procedures tested regularly
- Periodic review of autonomy levels against measured accuracy
Frequently asked questions
- Are autonomous networks real or still a vision?
- Parts are live. Deutsche Telekom's RAN Guardian agent went live in Germany in November 2025 and reduced the time to manage major events from hours to around a minute. Telstra showed a self healing proof of concept that moved network applications away from failed hardware in minutes. Full end to end autonomy across all domains is still ahead.
- What does intent based mean?
- The operator states the outcome, such as a latency target for a 5G service, and the system works out and executes the actions to achieve and keep it. du and Nokia describe autonomous network slicing that adjusts radio policies to keep premium service levels.
- Who is accountable when an agent changes the network?
- The operator. Vodafone's framework with Google Cloud and TM Forum stresses policies, explicit guardrails and human in the loop approval for major network changes. Every agent needs an owner, an allow list and an audit trail.
How to cite this page
Blits.ai AI Use Case Library, "Agentic AI for autonomous, intent based network operations", last verified 26 September 2026, https://www.blits.ai/ai-use-cases/autonomous-network-operations. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published