3GPP’s push toward an AI-native 6G standard signals that autonomous, intelligent networks are central to the next evolution, but the first finalized 6G release is still a few years away. In the meantime, the telecom industry is actively embracing AI agent technology as an alternative control plane to add autonomous capabilities to existing infrastructure.
That shift is not converging on a single stack. Agent technology is evolving rapidly, with vendor platforms, hyperscaler offerings, and open source build-your-own frameworks all in play.
Agent layers are moving into the middle of the stack
A multi-layered architecture is emerging that spans from the data substrate to business and user intent, with an agent orchestration layer in between. The goal is to enable autonomous agent processes and collaboration across the RAN, core, and cross-domain telco operations.
The analysis points to a familiar pattern: operators want the control benefits without waiting for standards to catch up. That is why the architecture being built now sits between raw telemetry and business intent, rather than replacing the network functions themselves.
Governance is the first constraint, not an afterthought
The industry should avoid introducing non-deterministic LLM agents into environments and network functions that require absolute predictability or do not need them. Multi-agent systems also create new security and interoperability challenges and expand the attack surface across vendor domains.
The TM Forum’s newly launched AI-Native Blueprint and other industry guidelines will be critical to ensuring multi-agent collaboration follows boundaries and governance.
Nokia, Ericsson and the hyperscalers are defining the near-term model
Nokia’s partnerships with Google Cloud, AWS, and Microsoft were demonstrated at DTW Ignite 2026. Google Cloud and AWS provide distributed compute for AI agent processing aimed at closed-loop automation at TM Forum Level 4 autonomy, while Microsoft Fabric integration unifies network telemetry to accelerate AI use case development.
Ericsson announced in June 2026 that it was expanding its Intelligent Automation Platform into the Intelligent Core. Together, these moves show where the market is placing its bets: on agent processing, telemetry integration, and cross-domain orchestration rather than one-off automation demos.
Commercialization is starting to show up in the tooling
Mavenir partnered with Red Hat to introduce a framework that meters and charges for AI token consumption through operator BSS. Amdocs demonstrated AI agents that turn plain-language customer requests into automated network actions for new premium on-demand services.
Telefónica Deutschland showed that AI agents in Blue Planet’s MDSO reduced 5G network slicing design time from weeks to minutes. Vodafone and Deutsche Telekom are among early adopters that achieve more than 90% automation in specific domains, while most operators remain at TM Forum Level 2 to 3 and are experimenting with multi-vendor solutions to avoid lock-in.
What network automation buyers should watch
The convergence of unified data layers, distributed compute inferencing, and cross-domain agent buses is reshaping the telecom industry’s legacy domains. Operators must upskill and align their cross-domain networks around the emerging agent architecture or risk maintaining unscalable, obsolete stacks.
For buyers, the practical question is not whether to buy into the 6G label. It is where agents can be governed, billed, and connected to existing operations without breaking deterministic workflows.