Telecom operators are moving toward AI-driven, autonomous networks while still depending on complex, siloed legacy systems. That tension is now limiting progress; AI can reason, and automation can act, but neither can compensate for fragmented data, incomplete visibility, or an assurance layer designed around human intervention.
Cost makes the challenge more urgent. Most operators already spend nearly 80% of revenue on operating expenses. Against that backdrop, every new investment faces scrutiny, even when AI is positioned as the route to greater efficiency.
The early results explain the scepticism. Omdia’s April 2026 survey of global communications service providers found that most respondents put the operating expense impact of AI at between 1% and 5%, well below the 20% to 40% reductions the market had been promised. The constraint is not simply the AI. It is the environment in which the AI is expected to operate.
When assurance systems are fragmented, expensive, and built for manual intervention, they restrict what AI can achieve. Modernising assurance reduces operating costs whilst creating the trusted intelligence needed for autonomy.
Legacy assurance keeps operating costs high
Legacy assurance was designed to tell an engineer that something was wrong, then leave diagnosis and resolution to people. That model now carries a permanent operating cost.
Traditional monitoring concentrates on whether the network is functioning rather than subscriber quality of experience. It checks control-plane signalling, confirms that a connection has been established, and verifies that a device has attached. It often stops short of showing what the customer is actually experiencing. Visibility may also end before the Radio Access Network (RAN), even though this is where much of the subscriber experience is determined. Where RAN visibility does exist, it commonly focuses on cell and node health rather than the customer experience; the network can appear healthy while the subscriber session is failing.
Infrastructure cost creates a second constraint. Legacy probing bottlenecks user-plane traffic, requiring a separate packet-broker tier. As traffic grows, the monitoring estate significantly expands, making comprehensive user-plane visibility increasingly expensive.
The third constraint is fragmented data, spread across separate systems and difficult to integrate with operational support systems, business support systems, and service management. When an issue occurs, engineers reconstruct the evidence manually and close the ticket once the network symptom appears resolved. The customer experience, however, may never have recovered, so the subscriber churns rather than calling back. That is the silent gap between technical recovery and genuine service recovery, and it carries a direct revenue cost.
These same blind spots hold back AI. Agents may be capable of reasoning and acting autonomously, but their decisions are only as reliable as the network intelligence they receive. Data quality was the most frequently cited barrier to wider AI adoption in the Omdia survey, identified by 51% of communication service providers. If the evidence is incomplete, fragmented, or unreliable, the agent is being asked to act on a partial view of reality.
RADCOM changes the economics before automation begins
The first opportunity to reduce operating expenses lies within the assurance infrastructure itself. RADCOM processes more traffic with fewer resources, reducing the hardware, power, and supporting infrastructure required to maintain network visibility.
Compression ratios of 3:1 for the control plane, 4:1 for voice, and 10:1 for the user plane enable substantially more traffic to be processed with less hardware. A single 2RU RADCOM server delivers 800 Gbps of user-plane capacity, consuming 63.9% less power than a legacy estate and providing at least four times the throughput. Packet-broker functionality is built into the probe, eliminating an entire infrastructure tier, along with its cost and potential failure points.
The platform is delivered as a cloud-native network function. It can run on the operator’s chosen Kubernetes distribution, on any hyperscaler, on standard commercial off-the-shelf servers or on compatible probe hardware that has already been purchased. This gives operators a practical route to modernisation without tying the business case to a wholesale infrastructure replacement.
The economic impact is material. Independent business-case modelling by ACG Research across small, medium and large network profiles places RADCOM’s total cost of ownership 37% to 43% below an incumbent baseline, with operating expense 44% to 48% lower over the modelled period. In other words, operators can reduce assurance costs before automation begins, without sacrificing end-to-end network visibility.
Automation then compounds the savings
Once the assurance foundation is modernised, automation can deliver more value because it is working with correlated, customer-level intelligence rather than disconnected technical signals.
RADCOM ACE provides IMSI-level visibility from the radio access network to the core, correlating each subscriber journey from end to end. Customer-impacting issues can be identified in seconds rather than reconstructed manually. Open application programming interfaces push this intelligence into operational support systems, business support systems and service management platforms, placing it directly into the workflows that can act on it.
RADCOM Neura, the company’s AI agent suite, turns that intelligence into action. It reduces the time and effort required to resolve issues by automating root cause and impact analysis, validating customer complaints against live network conditions, grouping related incidents and guiding resolution across domains. The result is shorter ticket cycles and fewer issues reaching the customer.
Modernisation without a forklift upgrade
Operators should not have to replace the existing estate before the savings begin. RADCOM is hardware-agnostic and can run on compatible probe hardware that operators have already purchased, which allows the migration to be funded progressively by the efficiencies it creates.
The transition is organised around six workstreams:
- Operational continuity aligns key performance indicators and carries historical data forward
- Infrastructure transition reuses existing probe hardware wherever practical
- Change adoption supports the current user base
- Systems integration preserves existing connections while adding new ones
- Regulatory compliance remains embedded throughout the transition
- Testing and validation confirm the quality of data, alarms, analytics and recommendations before production useÂ
Vendor lock-in made the old model expensive. Open architecture, hardware choice and a structured migration path make the new model affordable.
The most economical route is also the fastest
The journey towards autonomous networks does not need to begin with another large investment competing for limited budget. It can start by making the assurance layer less expensive to run and complete enough to support confident action.
That is the shift RADCOM enables. Assurance costs materially less to operate than the estate it replaces while producing the real-time, customer-aware intelligence on which autonomous operations depend. Infrastructure savings are realized first, automation savings follow, and neither depends on a forklift upgrade.
For operators deciding whether they can afford to modernise, the question should be reversed. Assurance modernisation is not a cost that must be justified before the autonomy programme begins. It is the investment that helps pay for it. It lowers operating expense today and creates the trusted intelligence required for greater autonomy tomorrow. That provides a credible path towards TM Forum Level 3 autonomy and beyond, where the network can optimise itself and resolve issues within defined domains without human intervention.
