TM Forum Autonomous Networks
Autonomy measured against the industry standard, not our own marking
Most autonomy claims are self-assessed, which is why few people believe them. TM Forum publishes a six-level scale, an evaluation methodology and a questionnaire library precisely so the answer can be evidenced. We are a certified evaluator, and we apply it to your network.
The six levels
TM Forum's published scale, restated. Almost every operator that describes itself as “doing autonomous networks” is somewhere between L1 and L3.
- L0
Manual Operation & Maintenance
Offline manual implementation; assisted monitoring only.
- L1
Assisted Operation & Maintenance
Manual implementation with online recording of events.
- L2
Partial Autonomous Networks
Automation driven by statically configured rules.
- L3
Conditional Autonomous Networks
Automation driven by dynamically programmable policies; senses real-time change.
- L4
High Autonomous Networks
AI-assisted decisioning across domains; predictive closed loops; continuous learning.
- L5
Full Autonomous Networks
Self-evolving closed-loop automation across services, domains and lifecycle.
A level belongs to a scenario, not to a network. There is no such thing as an “L4 operator”, and a vendor claiming to make you one has already told you they have not read the framework.
Scored per scenario, against TM Forum's published packages
There is no single autonomy score for a network. The assessment takes one operational flow and scores it on the Intent, Awareness, Analysis, Decision and Execution activity chain, following the IG1252 methodology. TM Forum's Autonomous Networks Level 4 solution packages define those flows, and the twelve published so far are below.
Service-oriented
| Mobile | Home broadband | Private line | Edge & computing | |
|---|---|---|---|---|
| Service delivery | IG1502A | IG1502G | ||
| Quality assurance | IG1503B | IG1503D | ||
| Complaint handling | IG1504A | |||
| Problem management | IG1505A | IG1505G |
Network-oriented
| Main template | RAN | Core | FAN | IP | |
|---|---|---|---|---|---|
| Network planning | IG1521C | ||||
| Fault management | IG1523 | ||||
| Change management | IG1524B | ||||
| Quality optimisation | IG1525E | ||||
| Energy efficiency optimisation | IG1526A |
Published, service-orientedPublished, network-orientedNot published yet
Published packages as at August 2026. TM Forum's matrix covers further scenarios still in development, which are not listed here. IG1502G includes ODA components.
How we take a customer up the curve
The sequence matters more than the destination. An operator that automates a scenario it cannot yet measure has bought risk, not autonomy.
Assess the scenario, not the network
Pick one operational flow and score it against the framework with the evidence recorded. This gives you a defensible starting number instead of an opinion.
Fix the data before automating on it
Assessments almost always surface inventory and configuration gaps first. Automating on top of a model that disagrees with the live network is how you industrialise a mistake, so this comes before anything else.
Automate with the human in the loop
The same automation runs in testing, then human-approve, then autonomous. Autonomy is a dial per scenario, not a switch for the estate.
Prove the outcome independently
Nothing is promoted on the strength of its own logs. A change is verified against the model and, where customer experience is what matters, against real devices.
Re-assess and show the movement
Score the same scenario again so progression is evidenced rather than asserted. This is what a board, and increasingly a regulator, actually asks for.
The assessment itself
A structured, auditable assessment for a chosen scenario, delivered as an Evaluator against the published methodology, with a scored result, the recorded evidence and a roadmap to the next level.
What this looks like in practice
The framework is a measuring instrument, not an implementation. These are the parts of the platform that move a scenario up a level, and the transition each one bears on.
Correlated inventory and topology
Gives automation something true to act on. Most operators discover at this step that the model and the network disagree.
Configuration audit and write-back
Detects drift against policy and corrects it, moving from static rules toward policy-driven action.
Closed-loop automation with an autonomy dial
Alarm to diagnosis to action, promoted from testing to human-approve to autonomous as confidence is earned.
Agentic AI control plane
Cross-domain decisioning from stated intent, with provenance on every answer and an approval gate wherever the operator wants one.
Real-device verification
Independent confirmation that a change improved the customer's experience — the evidence a closed loop needs before it can be trusted to run itself.
These mappings are ours, offered as a way to reason about sequencing. TM Forum scores a scenario against its own questionnaire and evidence; no product, ours included, confers a level.
The verification half, in detail
The last row is the one operators most often skip. Independent confirmation that a change improved the customer's experience is what turns a closed loop from a demonstration into something you would let run unattended.
What the certification does and does not mean
Worth stating plainly, because this distinction gets blurred in our market. The framework defines two roles: the Evaluator, who runs the assessment, and the Evaluated, whose network is scored. Digitata Networks acts as Evaluator. TM Forum itself does not carry out assessments — it operates a separate validation service that verifies a submitted self-assessment used the tooling correctly and that the evidence matches the scores. That validation confirms process and evidence. It does not confirm that the network physically performs at the assessed level.
So, precisely: we are certified to run the assessment, and we can facilitate submission for TM Forum validation. Neither of those is TM Forum certifying your network, and anyone implying otherwise is selling you something.
Find out where you actually are
One scenario, scored against the framework, with the evidence recorded and a roadmap to the next level. It is the least speculative way to start.