5G network slicing has been technically possible since 2019, but telecom operators treated it like a science experiment—too expensive to configure, too rigid to monetize. Nokia and AWS just deployed the automation layer that makes on-demand slicing economically viable, and two carriers are testing it in live networks right now.
du in the UAE and Orange across Europe and Africa are running commercial pilots of agentic AI that monitors network KPIs and adjusts policies continuously, responding to live events without manual intervention. The system is being demonstrated at Mobile World Congress 2026 this week. Why it matters: operators have promised network slicing for seven years without delivering scalable business results. This is the first deployment that could actually work at scale—if regulators allow it.
The system makes real-time decisions no human operator would approve
Nokia’s solution ingests seven external data sources—locations, events, traffic conditions, incidents, maps, weather, and timetables—then autonomously reallocates bandwidth based on what it predicts will happen next.
Picture this: AI detects a stadium filling up for a concert. It reallocates bandwidth from residential users to premium gaming customers near the venue. Then an ambulance enters the coverage area. Who decided the priority rules? What happens when the AI gets it wrong?
That’s the unresolved question. According to AI News, “how regulators will view AI control of critical communication infrastructure” remains an open problem. Telecom networks carry emergency traffic. Autonomous AI agents making operational decisions raise accountability questions no one’s answered yet.
And regulators haven’t approved autonomous control of emergency communication infrastructure—a pattern we’ve seen with autonomous agents making governments nervous across multiple sectors.
Manual slicing killed the business case—automation might save it
Here’s why telecom operators spent seven years promising network slicing and never delivered: manual configuration made it economically worthless.
The old model required RF engineers to define static slices, locked in for months, with no dynamic pricing. Orange previously said “enterprise customers expect connectivity to behave more like cloud computing, where resources can scale on demand.” They just couldn’t build it until now.
The new model runs on AWS Bedrock integrated with Nokia’s AirScale base stations and MantaRay SMO orchestration platform. The system triggers event-based slicing, pay-per-use pricing, and scales like cloud compute. Network engineers who spent weeks manually configuring slices are facing the same automation pressure as high-skill jobs facing automation in other technical fields.
But the deployment complexity is real. Nokia’s stack requires operators to integrate three separate systems (base stations, orchestration, AI modules) with AWS cloud services. That’s a heavy lift for carriers running legacy infrastructure.
The ‘autonomous’ system still requires human babysitters
Here’s the honest limitation: operators will run supervised pilots for months, possibly years, before removing human oversight.
According to AI News, “operators typically introduce automation gradually, keeping human oversight in place while validating system behaviour under real conditions.” Engineers must validate every AI decision before full autonomy. The efficiency gains will arrive slower than the press releases suggest.
Questions remain about deployment at scale. The technology works in controlled pilots. But no telecom regulator has approved autonomous AI control of emergency communication traffic. Every operator will need to prove the system won’t prioritize a stadium concert over an ambulance dispatch.
That’s the trade-off. The promise is plug-and-play network optimization. The reality is cautious, supervised rollout under regulatory scrutiny.
Orange wants enterprise customers to get cloud-like flexibility. Regulators want to know who’s liable when autonomous systems fail. Both things are true. Neither is solved yet.
