Artificial intelligence is largely becoming a key component of network automation solutions. These solutions allow operators to maintain performance, reduce service disruptions, and operate more efficiently, by enabling proactive monitoring, predictive analytics, automated anomaly detection, intelligent traffic or service optimization, and accelerated root-cause analysis, At Intracom Telecom, a long-term focus has been on enhancing our wireless networks through AI-driven innovations that significantly improve network performance and reliability.
Fixed Wireless Access (FWA) broadband introduces a fundamentally different operational reality compared to fiber or cable because service quality depends on a shared and highly variable radio medium rather than a stable physical line.
Challenges in deploying mm-Wave FWA include LoS requirements, extensive site surveys and planning to ensure optimal placement of base stations and Terminal Stations (TSs), as well as a strong requirement for continuous monitoring and management to optimize signal quality, adjust bandwidth allocation and ensure uninterrupted high-speed. Additionally, FWA networks are prone to external situations where the service might remain technically connected but the user experience might be poor.
Such degradations are often driven by factors like partial obstructions in the LoS, antenna misalignment, rain-induced attenuation, and persistent signal deterioration caused by vegetation growth on the radio path.
AI/ML can help reduce operational complexity, enhance reliability, improve performance, and ensure cost-efficient scaling, addressing the most pressing pain points of existing and next-generation wireless deployments.
Towards relieving multi-domain ambiguity, AI/ML can significantly assist by introducing correlation-driven monitoring that combines TS telemetry, radio KPIs, Base Station performance indicators into anomaly detection and root cause classification models. Rather than relying on static thresholds, ML can learn baselines per cell and time-of-day, detect deviations that signal emerging congestion or interference, and classify incidents into buckets such as radio coverage impairment, congestion/capacity limitation. AI/ML can help by forecasting cell load and congestion trends, using historical traffic and radio utilization patterns, allowing operators to anticipate when and where performance will degrade and act proactively.
Generative AI can complement these AI/ML capabilities, by summarizing and explaining incidents and providing contextual information with supporting evidence thus reducing the need for human intervention and escalation. GenAI can also generate tailored next-step guidance for different audiences (NOC engineers, field technicians, or customer support).
Intracom Telecom is developing a uniMS™ AI conversational assistant, that receives user queries through natural-language interaction, pulls evidence across multiple network elements (Base Stations, TSs) and data layers (planning, performance, capacity), provides explainable insights on FWA network status, proposes the most likely root cause category with confidence and evidence, and recommends the next best actions.
The AI assistant is positioned as a natural-language interface to uniMS™ data and workflows, aimed at reducing time spent on repetitive diagnostics and turning raw telemetry/alarms into actionable operational insight.
Functionally, the AI assistant will:
The new AI assistant is designed to be tightly integrated with uniMS™ Platform and not as a parallel system. In addition, it could support troubleshooting by utilizing internal technical documentation to provide context aware responses via Retrieval-Augmented-Generation (RAG) functionality.
We report two representative user queries that illustrate the expected experience: an operator asks for sector ranking by disconnection rate over the last 24 hours and get an aggregated, prioritized list, or requests KPI trend analysis (RSSI/SNR) for a given TS across 24 hours and 7 days with deviations highlighted versus a baseline. In other words, the Copilot is designed to translate operator intent into the right data retrieval and analysis steps, and present results in an operationally useful form.
Autonomous networks shift operations and maintenance away from static rule-based automation – managed by humans, towards more dynamic intent-based automation that can enhance customer experience and guarantee SLA commitments at scale.
Rather than requiring engineers to define detailed workflows for every possible scenario, intent-driven networks enable operators to express what outcome they want (for example, quality thresholds or performance targets) and rely on AI, real-time data, and closed-loop feedback mechanisms to continually adjust the network to meet those goals.
Here, AI and machine learning provide the predictive and analytical layer that anticipates emerging issues, while Generative AI translates network behavior into human-readable explanations and recommendations. In more advanced settings, lightweight agentic capabilities can help the system decide when to act within predefined guardrails, rather than simply reporting problems to humans.
Intent-driven autonomy enhances traditional automation by aligning network behavior with operational goals rather than isolated tasks. For repetitive incidents and predictable performance degradations, an intent framework can enable the system to recognize patterns, evaluate utility against specified objectives (such as a minimum throughput or latency target), and enact corrective adjustments proactively.