Mediation, rebuilt for the job it now does

Networks that run themselves

The industry is working towards networks that spot a problem, decide what to do and fix it, with people setting goals instead of working every ticket. TM Forum calls this Level 4 autonomy. In its survey of 80 operators this year, 81% are targeting it by 2030 and 75% plan to spend more on getting there.

The reason is money. PwC expects telecom service revenue to grow by around 2.8% a year to 2029, while Ericsson puts mobile traffic up 23% in a year to more than 220 EB a month. With complexity rising too, the cost of running the network has to fall. McKinsey reports that operators using AI across fault handling at scale have seen 55 to 80% lower network operations centre costs and 30 to 40% faster repairs.

Agents need data they can trust

AI agents are what take operations from automated to autonomous. Deutsche Telekom’s RAN Guardian agent has cut the time to manage major network events from hours to around a minute. A TM Forum Catalyst backed by T-Mobile USA, Telefónica and others showed a NOC where specialised agents link faults to business impact and fix them without a ticket or a bridge call.

Every agent decides based on what it is given. If the data is late, incomplete or inconsistent between domains, it acts on the wrong picture. Analysys Mason tracks 46 agentic AI activities among operators, finds most still at pilot stage, and says the foundations have to be right before they can scale.

Why Mediation is back on the telco strategy menu

Mediation started as billing plumbing: collect usage records, check them, remove duplicates, normalise them and pass one reliable record to charging. Because it already reached into the network, operators gradually fed the same data to fraud, assurance, planning and regulators. A billing component became shared data infrastructure, while the platforms underneath stayed on-premises, proprietary and hardware-bound.

That history puts mediation in the right place now. It already sits where network data becomes business data, and it already does the collecting, cleaning, enriching and auditing that agents depend on. Appledore Research calls it a fast track to getting operator data ready for AI. In the TM Forum Catalyst, the agents work from a shared data foundation that mediation feeds.

Treating mediation as a legacy cost line now carries a risk. If it is slow, rigid or inconsistent, every agent inherits the problem, and an agentic programme is only as reliable as its oldest component.

What an agentic mediation platform has to be

Real time, with quality you can see. Streaming alongside batch, rules changed through configuration rather than code, and quality checks and pipeline health visible as data flows, so problems surface before customers notice.

Cloud-engineered, not cloud-hosted. A monolith moved into virtual machines still scales, upgrades and depends on scarce skills as before. A platform designed for containers scales each part independently, keeps configuration in version control and releases like any other modern software.

Built for volume, and for machine callers. Agents ask many more small questions than analysts and expect a current answer each time. Operators should test two things using their own traffic volumes: how much work each unit of hardware can handle, which sets the running cost, and how well the platform responds when many requests arrive at the same time.

Headless, with control built in. The platform should be operable by agents as well as through a user interface. Model Context Protocol, an open standard now supported across the major AI providers, is the most visible route. Control matters more than access. Telenor Sweden wants “clear separations between microservices that can change things and those that can only see things”. That means scoped permissions, a log of every action and human approval before anything reaches production.

What operators get

An operator that adopts agentic mediation reaches autonomy sooner, because it builds on a layer it already owns. Its agents work from complete, traceable data that is easier to trust and to explain to a regulator. One converged platform costs less to run than separate silos, and captures the real-time usage that new services such as guaranteed experiences and network APIs need. Billing, assurance and AI run from one governed pipeline, with one set of records for every team and agent.

Four questions to ask

Can it stream current, quality-checked data and show its own health? Was it built for containers, or only hosted on them? Does it hold up when machines ask at once? Can an agent’s access be scoped, logged and approved?

SOURCES

TM Forum, “Autonomous Networks Leadership Forum spotlights Level 4 shift from cost-cutting to revenue growth”, DTW Ignite 2026. https://tmforum.org/news-insight/newsroom/autonomous-networks-leadership-forum-spotlights-level-4-shift

PwC, “How telecoms can win in the age of AI: perspectives from the Global Telecom Outlook 2025 to 2029”. https://www.pwc.com/gx/en/1/industries/technology-media-telecom/telecom-outlook-perspectives.html

Ericsson, Mobility Report, mobile traffic update (Q2 2026, June 2026 edition). https://www.ericsson.com/en/mobility-report/dataforecasts/mobile-traffic-update

McKinsey & Company, “Issue Brief: AI-driven telecom networks”, 27 February 2026. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/issue-brief-ai-driven-telecom-networks

Deutsche Telekom, “Deutsche Telekom and Google Cloud Collaborate for Superior Network Experience with Agentic AI”, 25 February 2026. https://www.telekom.com/en/newsroom/latest-updates/media-information/2026/2/mindr-ai-agents-in-telekom-network

TM Forum Catalyst C26.0.924, “Agentic NOC: AI-native operations for the autonomous telco”, 2026. https://www.tmforum.org/catalysts/projects/C26.0.924/agentic-noc-ainative-operations-for-the-autonomous-telco

Analysys Mason, “Agentic AI in telecoms: trends and analysis 1H 2026”, Adaora Okeleke, 12 August 2026. https://www.analysysmason.com/research/content/reports/agentic-ai-telecoms/

Appledore Research, “Telco AI Enabler: Mediation’s Defining Role”, John Abraham, 23 February 2026 (published in partnership with CSG). https://assets.mobileworldlive.com/wp-content/uploads/2026/06/26094648/Mediation-as-AI-Enabler-Appledore-Whitepaper.pdf

TM Forum Inform, “How telcos are building safe environments for AI agents”, Dawn Bushaus and Joanne Taaffe, 24 February 2026. https://inform.tmforum.org/features-and-opinion/how-telcos-are-building-safe-environments-for-ai-agents