Ericsson launches AI-driven rApp as a Service on AWS
Ericsson has introduced an agentic rApp delivered as a cloud service on Amazon Web Services (AWS), aiming to speed operators’ shift from manual automation toward truly autonomous networks.
Overview of the rApp-as-a-Service offer
rApps are applications that run on the non–real-time RAN Intelligent Controller (non‑RT RIC) within the Service Management and Orchestration (SMO) layer defined by the O-RAN Alliance. By offering an “Agentic rApp as a Service” on AWS, Ericsson is packaging policy-driven and AI-assisted RAN optimization as a managed, cloud-delivered capability. The goal: faster onboarding, lower integration friction, and a more repeatable path to closed-loop assurance across multi-vendor 4G/5G networks.
Why agentic AI matters
Agentic capabilities bring reasoning, planning, and action-taking to operations. In practice, this means the rApp can analyze intents and policies, correlate telemetry, propose or execute changes, and learn from outcomes—within operator-defined guardrails. Coupled with SMO and near‑RT RIC/xApps, an agentic rApp can elevate automation from static playbooks to adaptive, goals-driven control loops across the RAN lifecycle.
Standards alignment and reference architectures
The rApp construct comes from the O-RAN Alliance’s non‑RT RIC framework, which interfaces with near‑RT RIC/xApps for time-sensitive control and with OSS/BSS for service intent. Autonomous operations also align with TM Forum’s Autonomous Networks levels and ETSI ZSM principles for zero‑touch management. Expect Ericsson’s offer to emphasize standards-based interfaces to ease adoption alongside existing SMO, SON, and analytics functions including 3GPP-aligned NWDAF.
Why it matters for autonomous RAN now
Operators are facing twin pressures—a tougher RAN cost curve and rising service expectations—while preparing for Open RAN, 5G-Advanced, and enterprise SLAs that require assured performance.
From automation to autonomous operations
Most CSPs have pockets of scripting and SON, but closed-loop autonomy remains fragmented. GenAI and agentic patterns are accelerating the move from rule-based workflows to intent-driven operations that can adapt in real time. An rApp delivered “as a service” creates a pragmatic on-ramp to those capabilities without long, bespoke integration cycles.
RAN cost pressures and ROI levers
Energy costs, dense site portfolios, spectrum refarming, and new mid-band deployments are squeezing margins. rApps can target high-ROI use cases—energy savings, interference mitigation, coverage-capacity balancing, anomaly detection—to improve utilization and reduce truck rolls. As traffic patterns shift with FWA, gaming, and enterprise private networks, autonomous optimization becomes a lever for both OPEX and experience.
Cloud delivery benefits on AWS
Running rApps on AWS offers elasticity, global reach, and faster release cadence. For many CSPs, this means quicker pilots, consistent security posture, and simplified lifecycle management versus running every component on-prem. When coupled with clear data boundaries and integration patterns, cloud delivery shortens time to value.
How AWS-hosted rApps transform RAN operations
Agentic rApps expand the scope and speed of closed-loop control while clarifying ownership between SMO, RICs, OSS/BSS, and data platforms.
High-ROI early rApp use cases
Expect initial focus on energy optimization (sleep modes and cell activation policies), parameter tuning (PCI/ANR, load balancing), QoE-driven capacity steering, and proactive fault isolation. As confidence grows, operators can add intent-based slice assurance, intelligent RAN feature activation (e.g., MIMO layers), and guided operational runbooks for NOC/ROC teams.
Key integrations and data flows
Value comes from clean interfaces: intent and policies from OSS/BSS and service orchestration; telemetry via SMO, near‑RT RIC, and data lakes; and actuation through standards-based southbound APIs. Align the rApp with O-RAN policies (A1), enrichment information, and policy distribution so that decisions in non‑RT time inform near‑RT behaviors and vice versa.
Data governance, trust, and guardrails
Agentic systems must operate under strict controls. Define where data resides, how models are governed, and which actions require human-in-the-loop. Establish rollbacks, canary changes, and observable SLOs. Align with security, privacy, and sovereignty requirements upfront to prevent slowdowns later.
Challenges and risks to manage
Cloud-delivered autonomy touches architecture, process, and regulation—each with risks that need active management.
Interoperability and avoiding lock-in
rApps promise portability via O-RAN constructs, but practical differences in SMO/RIC implementations persist. Insist on open APIs, documented data schemas, and exit strategies so rApps can work across multi-vendor domains and future-proof Open RAN transitions.
Observability and ROI measurement
Autonomous changes must be explainable. Instrument the rApp’s decisions with human-readable rationales, pre/post KPIs, and cost/benefit tagging. This is critical for finance (business cases), engineering (trust), and regulators (auditability).
Security, data sovereignty, and compliance
Clarify where inference and data processing occur, how identities are managed, and how policies are enforced across borders. Map controls to frameworks such as GSMA NESAS/SCAS, ISO/IEC 27001, and local telecom requirements, especially for RAN data that may be considered sensitive.
Next steps for CSPs and enterprises
Treat this as a catalyst to systematize autonomous network adoption with measurable guardrails.
Start with measurable closed loops and KPIs
Prioritize one or two rApp-driven loops with clear KPIs: energy savings per site, call-drop reduction, or throughput uplift. Run controlled A/B trials, codify rollback, and publish results internally to build momentum.
Build an rApp/xApp roadmap
Create a 12–18 month plan that sequences foundational rApps (policy, enrichment, anomaly detection) before advanced intents (slice assurance, enterprise SLA optimization). Ensure alignment between non‑RT rApps and near‑RT xApps to avoid policy conflicts.
Align org, skills, and guardrails for agentic ops
Stand up a cross-functional pod—RAN engineering, operations, cloud, security, and data science—with a product owner for autonomy. Invest in prompt engineering, policy-as-code, and observability skills so teams can tune agent behavior safely.
Bottom line on Ericsson’s Agentic rApp
Ericsson’s Agentic rApp as a Service on AWS is a pragmatic step toward autonomous networks, offering standards-aligned, cloud-delivered closed loops that can cut costs and improve experience while laying groundwork for Open RAN and 5G‑Advanced.
Strategic takeaway: operationalize autonomous networks
Use this moment to formalize your autonomous networks strategy: anchor on open interfaces, start with high-ROI rApps, harden governance, and scale iteratively—so GenAI-driven agents become a trusted, auditable part of your RAN operating model.
