TEL：+86 158 1857 3751 
Japanese operators are accelerating their "AI-first" strategy, with the ambitious goal of reaching Level 5 of the TM Forum's "Autonomous Network (AN) Maturity" within 18–24 months.
Some scenarios have already self-assessed at Level 4, but the maturity curve is exponential—progress from 0 to 70 is quick, while advancing from 90 to 95 is extremely difficult. Therefore, the 18–24 month timeline is a realistic target. This schedule is generally in line with the industry, as 20% of operators have already reached Level 4 or 5 in certain areas, and 35% are expected to achieve this within two years.
Currently, no operator worldwide has officially reached Level 5. Network autonomy is divided into six levels: Level 0 (fully manual) to Level 5 (fully autonomous). As early adopters of AI, these operators internally adhere to a quantifiable "AI Triple 20" principle: any AI project must deliver a 20% efficiency improvement in marketing, operations, and customer experience.
Their AI strategy focuses on four key areas: customer experience, network, operations, and sustainability.
In terms of customer experience, AI first identifies the scope of fault impact, then automatically calculates compensation solutions, and adjusts the tilt angles of neighboring antennas to ensure coverage. "The entire process is self-developed, with AI integrated at every step."
On the network side, AI focuses on three key KPIs: quality, coverage, and congestion. Among 400,000 indoor and outdoor cells, AI proactively identifies high-traffic events and dynamically allocates resources.
For energy efficiency, instead of shutting down stations during idle periods—which could harm user experience—power is reduced in low-traffic areas. This model has already been implemented in over half of the network, achieving energy savings of 17% to 22%.
In operations, AI is used for anomaly detection, root cause analysis, automated handling, and closed-loop resolution. Several AI agents are currently being piloted, achieving an accuracy rate of 85%. At this stage, a "human-machine hybrid" model is adopted.
AI agents must justify their cost-effectiveness: although the unit price of tokens is decreasing, usage is increasing, which could result in saving $10 while spending an extra $40. To address this, operators have begun developing their own "Small Language Models (SLMs)." These models require significantly less computational power and infrastructure compared to large models, deploy faster, are easier to optimize, and allow data to remain on-premises. Inference can be performed without GPUs, offering better cost efficiency.
