Autonomous Networks

Summary: “Autonomous Networks” (multi-speaker session on AI-native / autonomous telco)

1) Scaling autonomous network level L4 (TM Forum view: business + ops outcomes)

• Goal: Evolve from autonomy deployed mainly for operational efficiency to autonomy that also drives business value and revenue growth.
• Why it matters: Traditional bandwidth-based products face revenue pressure as growth slows; autonomy can enable new monetizable offerings.
• TM Forum benchmark insight: About 75% of surveyed operators (~80 CSPs) plan a moderate/significant increase in EN (autonomous network) L4 investment over the next year.

Observed operator results (examples)

• China Mobile: claims 6,000 person-years of annual labor savings.
• Telia / Telra (as transcribed): reduced 5G network slicing time-to-market by 30%.

ZeroX / SelfX vision

• “Zero X” (consumer/enterprise): zero-weight, zero-touch, zero-trouble experiences.
• “selfX” (internal): infrastructure that is self-serving, self-fulfilling, self-assuring for planning/marketing/ops/maintenance teams.

TM Forum framework emphasis

• A transformation framework with core elements plus operational/industry evaluation and certification.
• An AN L4 industry blueprint supports a phased two-stage implementation approach.

Agentic architecture (high-level)

• Agents act as a brain/command center:
  • acquire/analyze data,
  • decide,
  • invoke tools for closed-loop execution:
    • awareness → analysis → decision → execution.
• Human co-pilot support remains for tasks not fully automatable; experts intervene for complex issues and to preserve security.

High-value scenario guidance

• Prioritize high-value use cases to clarify scope, improve time-to-value, and strengthen ROI.
• Focus on differentiated experiences tied to deterministic SLA (e.g., bandwidth, latency), often combined with security.

Phase 1 vs Phase 2 (TM Forum updated evolution model)

• Phase 1: end-to-end autonomous closed-loop at the scenario level, using user intent (users + network ops personnel).
• Phase 2: expand to multi-scenario collaboration across different service types, aiming for network-wide optimality, including business innovation and revenue growth scenarios.

---

2) “AI at the core” / Telco → TechCo via agentic, contextual, anticipatory operations (Circles / Sanjay K)

Industry shift

• Connectivity becomes table stakes; differentiation comes from best digital experiences (competition with internet/hyperscalers).
• Survey-style claim (GSMA): Many telcos plan TechCo transformation, but few show tangible progress.

Proposed AI-native telco operating model

• Predictive & autonomous: agents act before offers are conceived.
• Contextual & anticipatory: detect likely overpaying/churning (e.g., consumption vs plan) and proactively recommend a better plan.
• Continuous partner/service orchestration: embed connectivity into partner touchpoints and enable new business models.

Example monetization scenarios

• POVO (KDDI mentioned): partnerships with Japan Rail and Lawson; data-driven free data / embedded connectivity.
• TikTok embedded connectivity: sense low data usage and offer upgrades within the TikTok experience.

Architecture principles

• AI must sit at the core of enterprise architecture and enable replacement/modernization of legacy systems (“zero glitch transformation,” per transcript).
• Requires a unified data lake (the “garbage in, garbage out” emphasis), then agentic AI on clean data.
• Enables a multi-market learning flywheel through scalable agent actions.

Expected outcomes

• Improved NPS, ARPU, and broader revenue impact—framing progress toward IMO (“AI-native mobile operator”) business value.

---

3) 5G + cloud-native + AI-native architectures for autonomous smart infrastructure (Transforma Insights / Suruchi Dingra)

Evolution of infrastructure control

• Sensor data → centralized correction (scalar systems)
• IoT analytics → still human-centric decisions
• Today at massive scale: autonomy requires closed-loop execution

Core closed control loop architecture

• 5G enables observation
• Cloud-native orchestrates computation
• AI predicts + automates
• Digital twins provide continuous learning

Where impact is highest

• Smart grids and smart cities prioritized as the most impactful autonomous domains.

Smart grid use cases

• Self-healing grids: detect faults, isolate, restore automatically.
• Real-time balancing: match renewable generation vs demand.

Communications requirements emphasized

• Ultra-low latency
• Massive IoT connectivity
• High reliability
• Network slicing to separate mission-critical traffic classes
• Edge computing integration to reduce latency and enable local decisions

Smart city examples

• Adaptive traffic management, emergency response coordination
• AI-enabled public safety video analytics

Also mentioned

• Manufacturing, transportation, ports (autonomous inspection/cranes), and healthcare (noted regulatory constraints).

---

4) Quantum computing / “quantum memory mirrors” for AI (Cubit Nexus / Dr. Katherine K — concept only)

Focus

• A quantum approach described as “quantum memory mirrors” using entangled light to “forget bad habits in AI” (as transcribed).
• Mention of a hybrid quantum programming language (Quill, claimed IP).

Quantum computer types discussed

• Cryogenic (near absolute zero)
• Room-temperature photonic
• Thermal computing

Key claim

• Quantum superposition + quantum algorithms enable simulations (including life-sciences dose optimization) with reduced time vs classical computing.

Practical emphasis

• Training/reskilling and building hybrid quantum neuromorphic skills.
• Quantum ecosystem availability via cloud (AWS credit mention).

---

5) AI-native networks: from network intelligence to business value at scale (Johan Lansstead)

Main argument

• AI is the beginning, not the destination—the destination is profitable growth.

Opportunity

• Differentiated connectivity enabled by guaranteed service performance through SLAs and network slicing.

Why AI is needed

• Complexity across:
  • multi-generational stacks
  • cloud-native breakdown
  • open-RAN horizontal vendor diversity
  • and AI needs for control/monitoring/optimization

Where business impact rises

• Use cases move “up the stack” toward business-relevant control.
• Fault management predicted as a major near-term business-impact area (3–5 years), alongside other higher-layer controls.

Conclusion theme

• Manage complexity, enable differentiated connectivity, and scale operational feasibility.

---

6) Generative intelligence + security: Telco Aegis autonomous security ecosystem (Mock Shihoudinisma SLA Santo)

Problem

• Attackers operate at machine speed; defenders often at human speed.
• Security operations face alert overload and fragmented tools.

Vision

• Move from reactive security to autonomous protection.

Proposed system: Telco Aegis

• A unified intelligent layer combining threat intelligence from multiple sources and ecosystem partners.
• An AI-powered autonomous security engine with a continuous lifecycle:
  1. Detect/Aggregate & correlate threats + telemetry
  2. Validate via agentic AI (risk evaluation, verify threat, decide response)
  3. Respond via orchestration/automation across integrated security + network platforms
  4. Learn from outcomes to improve the knowledge base

Collaboration emphasis

• Collective defense + federated/industry ecosystem blueprint
• Shared anonymized threat intelligence target

Demo use cases (as transcribed)

• Detection simulation and recombination for APN/UNE (transcript shows “AP UNCC 3886” as reference)
• Subscriber location tracking attack detection with auto-mitigation

Benefits claimed

• Reduce detection/response time (MTD/MTDR from days to minutes—transcribed)
• Higher precision/accuracy
• Safer automation with validation before live deployment

---

7) Observability in the age of AI: agentic AI ops + autonomous remediation (Richard Simon / T-Systems)

Shift in observability

• From reactive/passive monitoring → proactive, AI-driven approaches.

Drivers listed

• agentic AI ops
• autonomous infrastructure
• hybrid/multicloud complexity
• telco blind spots
• data silos
• need for real-time visibility/pattern recognition
• cost and governance pressures
• expansion to data observability and AI stack observability

Agentic AI ops

• Combine ML + generative + agentic intelligence for proactive, context-aware analysis.
• Support natural-language querying for patterns, causes, and remedies.

Autonomous remediation

• Agents monitor business services and automate corrective actions.
• Workflows evolve to remove human bottlenecks.

Business/operational impact

• Reduced downtime
• Operational teams (SRE/platform engineers/cloud ops) must update practices
• AI governance essential (policy, security, controlled operation of AI agents)
• Consolidate/normalize data models; ensure clean telemetry reaches observability tools.

---

8) Open cloud + AI-native telecom ecosystems via open operator platform (Vasil Zaro / Greece/Intercom Telecom + SIG Delhi + “Etsy Open” / Open OP)

Key claim

• Transformation must be simultaneous AI-native and data-native.

Main bottleneck

• Not only better AI models, but trustworthy, high-quality, continuously evolving data with governance.

DataOps framing

• Data quality expands beyond missing values/duplicates to:
  • provenance
  • governance and trust
  • explainability
  • regulatory compliance
  • semantic interoperability

Federated data ecosystems

• Intelligence requires combining telemetry across domains (radio, user mobility, energy, digital twins, KPIs).
• No single operator/country has all diversity → federated collaboration + sovereignty.

Project approach

• EU flagship initiative (SIG Delhi) building end-to-end AI/data framework.
• Alignment with IDSA / EU data spaces principles (as transcribed).
• Etsy Open / Open OP turns research into reusable software and bridges research–industry–open source–standards.

Next step mentioned

• A “data lab module” to make the platform AI + data-native, including native governance/ops and federated data space capabilities.

---

9) AI-native infrastructure → business value and autonomous operation (Stella Co / Taipei)

Core evolution

• Reactive ops (manual troubleshooting) → AI embedded across infrastructure for self-optimizing and autonomous operations.

Component highlighted: ARN (AI RAN)

• Continuously collects/analyzes radio & traffic data.
• Dynamically optimizes:
  • traffic analysis
  • intelligent resource allocation
  • energy saving
  • QoE optimization

Business opportunity from AI infrastructure

• Share AI server/GPU investment to offer enterprise services:
  • AI inference
  • private AI models
  • AI-as-a-service (monetize infrastructure utilization)

AI troubleshooting agent

• Monitors anomalies, analyzes likely root causes, recommends actions, and sometimes automates resolution.
• Example: UE can’t connect; agent finds duplicate IPs across interfaces and suggests a fix in seconds.
• Also supports customer self-troubleshooting during emergencies.

---

10) AI-native monetization in 5G: closing the “commercial loop” (Ahmed Al-Salami)

Central thesis

• AI-native infrastructure/automation creates intelligence, but revenue needs a commercial loop.

Core problem

• AI optimization doesn’t automatically monetize; the commercial layer must translate AI decisions into chargeable outcomes in real time.

Proposed architecture shift

• Policy control allocates resources using AI insights.
• Online/converged charging turns decisions into revenue.
• Charging becomes part of the real-time decision loop (not a back-office step after usage).

Real-time closed commercial loop example

• AI detects autonomous-factory traffic needs ultra-low latency.
• Policy control adjusts resources.
• Charging immediately applies SLA-based pricing (no batch/override delay).

Industry building blocks (vendor examples cited in transcript)

• Ericsson: intent-based charging (tight coupling between slice provisioning and charging policies)
• Netcracker: partner ecosystem monetization + revenue sharing automation
• (Third approach mentioned) cloud-native converged charging with microservices/APIs

New revenue opportunities

• Network Slice as a Service (pay for business outcomes/guarantees, not only volume)
• “open EBI monetization” (real-time capabilities exposed as a commercial service—wording transcribed)
• AI personalization (context-aware upsell, e.g., premium speed upgrade at the right moment)

Bottleneck discussed (mass scale beyond pilots)

• Biggest bottleneck across speakers:
  • data fragmentation + legacy OSS/BSS integration (“data readiness gap”)
• Additional recurring blockers:
  • trust/governance
  • organization change
  • AI governance/security
  • workflow integration
  • measuring value

---

11) Q&A consensus themes (explicit bottleneck discussion)

Common bottlenecks and concerns raised across multiple speakers:

• Data readiness gap / fragmentation (plus telemetry + real-time digestibility)
• Integration with legacy OSS/BSS and operational workflows (AI must fit existing telecom systems)
• Trust and governance (audit trails, explainability, regulatory compliance, safe deployment)
• Organizational readiness (skills retraining, employee trust, human-in-the-loop where needed)
• Security & governance of AI agents in production
• Measuring/value realization (CFO reluctance; need demonstrable ROI and business metrics)

---

Main speakers / sources (as identified in subtitles)

• Keshan Gundes — founder; presenter/host for TM Forum L4 scaling topic
• Sanjay K (Singapore) — Circles; “AI at the core telco imagined”
• Suruchi Dingra (Gurugon, India) — Transforma Insights; “5G and cloud AI native architectures…”
• Dr. Katherine K (Genea/Switzerland, per transcript) — Cubit Nexus; “quantum memory mirrors…”
• Johan Lansstead (Stockholm, Sweden) — AI-native networks overview
• Mock Shihoudinisma SLA Santo (Jakarta, Indonesia, per transcript) — Telco Aegis; gen-intelligent security ecosystem
• Richard Simon (London, UK) — T-Systems; “observability in the age of AI”
• Dr. Vasil Zaro (Greece) — Intercom Telecom / SIG Delhi / “Etsy Open” / Open OP
• Stella Co (Taipei, Taiwan) — AI-native infrastructure for business value & autonomous operation
• Amit/Ahmed Al-Salami (Yemen, per transcript) — AI-native monetization in 5G (closed commercial loop)

External reference/source mentioned

• TM Forum (benchmarks, L4 framework/industry blueprint)
• GSMA survey (TechCo agenda discussion)
• Additional vendor names mentioned in passing in transcript context (e.g., observability/security tooling such as Datadog and New Relic/Dynatrace).