For years, the pitch for automating network operations was about cost: fewer manual interventions, fewer truck rolls, lower operating expense. At TM Forum's DTW Ignite event in mid-2026, that framing changed. The industry body's latest guidance on scaling toward what it calls Level 4 autonomy adds revenue growth as a core objective alongside efficiency -- the argument being that a network smart enough to run itself is also a network you can package and sell differently.
What "Level 4" actually means
The autonomous-network scale runs from fully manual operations up to a network that effectively drives itself. Level 4 is the point where, in the framework's own phrasing, intent replaces instruction: operators set business goals and guardrails, and the system senses, decides and acts on its own within them, rather than waiting for a human to write and run a script for every situation. Everything below that is essentially reactive automation -- it only handles the scenarios someone anticipated in advance.
Almost nobody is fully there yet. TM Forum's 2026 research, based on a survey of 80 operators worldwide, found only around 4% report operating at Level 4 today, while 81% are targeting Level 4 or above by 2030. The trend line is still steep, though: a mid-2026 update counted 40 Level 4 validations awarded to 17 operators, up from a near standing start.
The proof points are getting concrete
The most cited example is Rakuten Mobile, which earned TM Forum validation for Level 4 autonomy on a live Open RAN network in Japan -- a production deployment rather than a lab trial. The reported result was roughly a 20% cut in radio-access-network energy use through AI-driven closed-loop control, with no measurable hit to customer-experience metrics. That matters because the radio access network accounts for the majority of a mobile network's energy draw, and energy is a large slice of overall operating cost. Writing for Rakuten Symphony, the company's Udai Kanukolanu summed up the change in roles: "The human moves up the loop from firefighter to architect."
Why the revenue framing matters more to a small ISP than the technology
A regional WISP or fibre ISP is not about to deploy a self-optimising Open RAN. The Level 4 machinery -- digital twins, semantic data layers, agent frameworks -- is built for carriers with thousands of engineers. But the underlying logic is exactly what a smaller operator should be copying, and can, today, on the commercial side of the business.
The pattern is simple: sense a condition, decide within preset rules, act without a human in the loop -- and use that same intelligence to protect revenue, not just contain cost. That is close to what a good billing and RADIUS platform already does for an ISP. A payment lands via M-Pesa or Kopo Kopo, and the subscriber's session is reconnected automatically. A subscription lapses, and access is suspended without an operator logging into a router at midnight. Usage and plan data across the customer base points to who is a realistic candidate for a higher tier. None of that needs a data-science team -- it needs the billing system and the network talking to each other in real time.
Africa is already running the customer-facing version of this
The service-and-revenue side of autonomy is further along in the region than the network side. Safaricom's WhatsApp assistant Zuri reportedly handles more than 70% of customer queries in Kenya, including bill payments and bundle purchases -- automation that cuts support load while keeping a sales channel open around the clock. Airtel has run an AI-driven spam and fraud SMS alert service in Uganda, and in 2026 MTN backed a funding round for a company building AI-native radio equipment aimed at Africa's power-constrained networks. Operators here are not waiting for Level 4 to be finalised before acting on the idea behind it.
The takeaway
The headline from 2026's autonomous-networks conversation isn't that self-driving networks have arrived -- they mostly haven't. It's that the industry has stopped treating automation purely as a cost-reduction project and started treating it as a way to deliver, and charge for, a better-run service. For a smaller ISP, the entry point to that is unglamorous and available now: a billing and RADIUS system that enforces policy automatically and turns subscriber data into decisions, so that growth in customer numbers doesn't simply become growth in manual work. That is the logic behind how XpressRADIUS ties real-time RADIUS control to payment data -- automated disconnect and reconnect, multi-currency mobile money, and a live view of who is due for an upgrade.
Want to see how automated disconnect, reconnect and real-time RADIUS control work for a growing ISP?
Start a free trial