TM Forum and Accenture have launched a multi-year programme designed to give telecom operators a common way to assess trust in AI systems as they move towards more autonomous operations.

The initiative will develop the Trustworthy AI Governance Certification, which TM Forum describes as an industry benchmark for assessing how communications service providers (CSPs) govern and manage AI. It will cover areas including governance maturity, workforce readiness, and operational accountability.

Announced recently at TM Forum Innovate Americas in Dallas, the programme focuses partly on autonomous and agentic AI systems that can make decisions or take actions rather than only provide recommendations.

TM Forum said the framework is intended to provide verifiable evidence around AI decisions and interactions between agents. The certification will be supported by an industry blueprint covering the design, deployment, and continuous monitoring of AI systems.

Telcos face a gap between AI trust and proof

The programme follows research from TM Forum and IBM’s Institute for Business Value that identified a gap between operators’ confidence in AI and their ability to demonstrate that their systems can be trusted.

The study surveyed 130 AI decision-makers at 130 telecom operators. It found that 72% of CSPs considered their AI capabilities trustworthy, but only 14% said they could provide evidence to support that assessment.

The research also examines how operators are moving from using AI to support human-led processes towards systems capable of making and executing more decisions autonomously. TM Forum said the consequences of malfunctions and hallucinations can become greater as that level of autonomy increases.

That creates a need to monitor AI inputs and outputs and retain evidence showing how systems reached their decisions.

Under the programme, operators will be able to benchmark their AI governance against common criteria and identify gaps in their processes. TM Forum said the framework will also cover training and certification aimed at preparing employees to work with more autonomous systems.

Guy Lupo, executive vice president of Trustworthy AI and Data at TM Forum, said trust needs to be supported by continuous, verifiable evidence rather than principles alone.

“Trust enables autonomy to scale, but trustworthiness only becomes operational when every principle is supported by continuous, verifiable evidence,” Lupo said.

Accenture will work with TM Forum on turning the proposed standards into operational practices. Arnab Chakraborty, chief responsible AI officer at Accenture, said agentic AI changes the role of the technology from supporting decisions to making them, requiring organisations to put governance and risk controls around those systems.

Autonomous AI moves closer to enterprise services

The governance work comes as operators introduce AI agents into customer-facing and operational systems.

At DTW Ignite in June, TM Forum hosted a session with Amdocs and Swiss operator Sunrise on the use of AI agents in customer relationship management and billing. The discussion covered consolidated data platforms, conversational customer interactions, and agent-to-agent interactions in areas including pricing and packaging.

TM Forum has also highlighted a constraint around billing, where processes require deterministic outcomes while AI systems have probabilistic capabilities. Its September report said operators are introducing AI into billing and revenue management as part of longer-term work towards AI-native business support systems.

AI agents are also being tested in network fault management and service assurance. The GSMA’s Agentic AI Testbed includes a Network Fault Troubleshooting Agent designed to ingest alarms, correlate events, identify the root cause of failures, and suggest remediation steps autonomously.

Some deployments are moving beyond recommendations. Australia’s NBN Co received TM Forum Level 4 Autonomy validation in June for a fixed-wireless service-assurance system developed with Ericsson.

The system monitors network performance, predicts degradation, identifies root causes, and can execute mitigation measures. It also supports automated fallback and rollback without human intervention during normal operations.

Autonomous operations are also being applied to enterprise services governed by service-level agreements. A current TM Forum Catalyst project involving operators including du, NTT Group, Ooredoo Qatar, Telkom Indonesia International, and VNPT is testing how agentic AI can connect commercial requirements with network operations.

The project is designed to translate business intent into service requirements, assess technical feasibility, and support dynamically priced services tied to service-level agreements. Its scope includes service creation, assurance, issue resolution, and SLA compliance.

An earlier phase also examined how AI-driven business support systems could support B2B sales for high-value 5G services.

Trust depends on auditability and control

As AI systems gain authority to take operational actions, operators need records showing what an agent did, which data it used, and whether the action produced the intended result. TM Forum and Accenture discussed this through the concept of an AI control plane at DTW Ignite earlier this year.

Chakraborty described the control plane as a layer that records AI decisions, the data sources accessed, and instances of model drift as systems operate. Guardrails can generate alerts, while the collected information provides a digital trail covering the AI lifecycle from design through deployment.

Traceability also becomes relevant when AI agents interact with other systems or agents. TM Forum’s new programme specifically targets verifiable evidence around autonomous decisions and agent-to-agent interactions, rather than relying only on an organisation’s assertion that its AI systems are trustworthy.

The Innovate Americas agenda also covers controls around autonomous decisions. An AT&T session addresses real-time state verification, escalation, and rollback, while a separate autonomous-networks panel examines how decision-making should be divided between humans and AI.

These governance questions now extend across service assurance, billing, service creation, and SLA management. TM Forum’s Trustworthy AI programme focuses on verifiable evidence around autonomous decisions, while its autonomous-network projects are testing AI across service creation, assurance, issue resolution, and SLA compliance.

The TM Forum and Accenture programme will use TM Forum’s AI-Native Open Digital Architecture as its standards foundation, with CSPs, vendors, hyperscalers, and consultancies contributing operational use cases through its AI-Native Collaboration Studios.

The certification will assess governance maturity, workforce readiness, and operational accountability. The broader blueprint will cover the design, deployment, and monitoring of AI systems.

(Photo by Roman Budnikov)

See also: TM Forum calls for controls on telecoms AI agents

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the IoT Tech Expo and Cyber Security & Cloud Expo. Click here for more information.

Telecoms Tech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.