Negotiating network slices with AI

Telecom and enterprise AI systems are beginning to negotiate and provision network slices

The telecommunications industry is rapidly moving toward a future where network connectivity is traded as dynamically as cloud computing. Instead of static contracts and manual provisioning, the next evolution of enterprise networking relies on the convergence of network slicing, intent-based networking, and AI. The goal is a fully automated marketplace where enterprise AI systems and telecom networks negotiate, provision, and monitor guaranteed network resources in real time, completely machine-to-machine.

This shift is already moving into actual product roadmaps. With companies like Singtel and Verizon running test deployments, and Nokia showcasing intent-based slicing at Mobile World Congress 2026, the pieces are falling into place. But while the underlying technology is advancing quickly, there are still significant technical, legal, and economic hurdles standing between today’s early deployments and a truly autonomous, AI-brokered network economy.

The core technology

At its heart, network slicing is a foundational 5G capability — and one that carves a single physical network into multiple independent virtual networks. It’s kind of like a highway that’s split into multiple lanes. One lane is reserved for autonomous vehicles that need ultra-low latency. Another handles IoT sensors that need broad coverage but can tolerate some delay. A third serves consumer video streaming with its appetite for high bandwidth. Each slice behaves like its own standalone network, tuned for specific performance characteristics, like latency, bandwidth, reliability, security.

Where it gets more interesting for automated negotiation is in AI-powered dynamic SLA enforcement. Traditionally, network resource allocation has been static. A telco provisions a set amount of capacity, and that’s what you get, regardless of what’s actually happening on the network. AI could help change that a little. It watches network conditions in real time and shifts resources across slices to make sure Service Level Agreements are continuously met. So, if congestion spikes in one part of the network, the AI reallocates within a slice to keep guaranteed performance intact. The technology could catch anomalies in milliseconds, and reroute traffic to maintain uninterrupted service. 

The emerging B2B marketplace

The concept that bridges network slicing and automated commerce is intent-based networking. Rather than an enterprise having to specify low-level technical configurations, intent-based systems let them simply declare what they want in near-plain language. The network’s AI takes it from there, automatically configuring slices to deliver those outcomes.

Nokia announced its first intent-based 5G slicing solution at MWC 2026, pairing its slicing tech with AWS AI platform capabilities. Under the hood, APIs coordinate data analytics, AI inferences, and slicing policies. Singtel, meanwhile, has commercially deployed network slicing at major events and is pushing it into healthcare, ports, and security. These are production deployments, not hypotheticals — even if they still involve a fair amount of human-mediated setup. Clearly, increasingly automated interactions where enterprise AI systems and telco AI systems transact directly through standardized APIs.

How the negotiation works

Let’s take an automated factory as an example. The factory’s AI specifies its intent — a guaranteed 1ms latency slice for a fleet of autonomous guided vehicles, two hours, 50 Mbps bandwidth. This request is structured and machine-readable.

From there, the system queries available telco APIs for pricing and availability. Think about how AWS on-demand cloud instances work today. You don’t negotiate with an Amazon salesperson every time you spin up an EC2 instance, you check the API, see the price, and provision. Same model here, except the “instance” is a guaranteed slice of a 5G network instead of a virtual machine in a data center.

Once terms are locked in, the telco’s AI provisions the slice and starts monitoring it continuously. Real-time KPI tracking confirms that latency, bandwidth, and reliability guarantees are being met moment to moment. If network congestion hits, the telco’s AI shifts resources within the slice to keep SLA guarantees intact. Predictive analytics can even see demand spikes coming before they arrive, pre-positioning resources to head off any degradation.

Settlement runs through automated billing tied to verified SLA compliance. Did the slice actually deliver 1ms latency for the full two hours? The monitoring data provides a record, and payment processes accordingly. 

Challenges ahead

Commercial B2B automation for network slicing is still largely aspirational. The demos and early deployments are real, but they represent emerging capabilities rather than a mature marketplace. There’s a meaningful distance between a successful MWC demo and a production system an enterprise is willing to trust with its mission-critical operations.

Legal frameworks might be an issue too. Who carries liability if a network slice fails mid-operation? If the enterprise’s AI negotiated the slice autonomously, the telco’s AI provisioned it autonomously, and something goes wrong — the chain of responsibility gets genuinely murky. Regulatory approval for automated, mission-critical network negotiations may end up requiring human oversight at some stage, which would naturally cap how fully autonomous the process can become. These are the kind of questions that regulators, insurers, and courts will need to sort out before high-stakes applications can run on AI-brokered network guarantees.

Multi-operator complexity layers on additional difficulty. Most current examples involve a single telco. But what happens when an automated logistics route spans two countries with different network operators? Automated negotiation across multiple carriers, each with different APIs, pricing models, and SLA structures, is dramatically more complex than working with one provider. Standards for cross-operator slice negotiation remain early-stage.

Then there’s the economics, which are still pretty opaque. The industry talks about “premium slices” and “monetization opportunities,” but actual pricing models for dynamically provisioned guaranteed slices haven’t been publicly detailed. Whether enterprises can justify paying for on-demand network guarantees remains to be seen. On top of all that, businesses face a substantial integration burden. Connecting enterprise IT systems to telco APIs demands standardization work that’s still very much underway, and anyone who’s worked with complex corporate IT environments knows they’re notoriously resistant to rapid change.

Future trajectory

Even with all these challenges, the timeline is compressing. Nokia’s MWC 2026 demonstration of intent-based slicing suggests commercial-grade solutions could show up within months, at least for controlled use cases. 

Longer-term roadmaps go well beyond basic slice negotiation too. Industry players are looking at quantum-safe networks with AI-enforced cryptographic guarantees, ensuring that even as quantum computing threatens current encryption, network slices stay secure. Multi-access slicing, spanning 5G, fixed wireless, and fiber, would deliver a unified experience regardless of the underlying transport, all negotiated and managed through a single AI interface. 

The vision is compelling: network resources traded as fluidly as cloud compute, machines negotiating with machines in real time to deliver guaranteed performance for the applications that need it most. Turns out, the technology is arriving faster than the legal, economic, and organizational frameworks required to support it. That gap is going to define how quickly the automated B2B slicing marketplace becomes real — and how much human oversight stays in the loop when the stakes are measured in lives rather than latency numbers.