Autonomous operations

What full autonomy looks like

TM Forum defines Level 4 as decision-making driven by intent and predictive analysis, with closed-loop management of service- and experience-driven networks through AI modelling and continuous learning. Stripped of the vocabulary, three things are true of a network operating that way.

• Zero wait — a service request, a change or a fix does not sit in a queue waiting for the one person who understands that domain. The system acts on intent as it arrives.
• Zero touch — the routine path from signal to resolution runs without a human in it. People set the policy, approve what the policy says needs approving, and handle the genuinely novel.
• Zero trouble — faults are predicted and prevented rather than detected and repaired. The customer never learns there was an incident, because there was not one.

“Zero-X”, and the Level 4 definition above, are TM Forum's.

Where the industry actually is

Most operators sit between Level 2 and Level 3, and some are moving quickly to Level 4 in particular domains. The distance from 3 to 4 is the one that matters and the one that is routinely understated: at Level 3 machines assist a human who still decides, and at Level 4 the machine decides inside a policy the human set. That is a change in who is accountable for a decision, not a change in tooling.

A level belongs to a scenario, not to a network. There is no such thing as an “L4 operator”, and any figure quoted for a whole estate is a figure for something that was never assessed.

What an operator has to gain

TM Forum names six areas where Level 4 changes the business. What each one means on an operations floor is the second column — that reading is ours.

What the research says it is worth

TM Forum research suggests autonomous-network maturity can deliver up to a 55% reduction in operations and maintenance cost, a 71% rise in customer satisfaction and a 21% increase in energy savings. Those are the industry body's figures, they are “up to”, and they are not attached to a named deployment — so we publish them tagged indicative and would not put them in your board paper as though they were ours.

Ask us instead what we have measured. The proof pages carry figures with a named scenario behind each one, and where a number came from a customer rather than our own instrumentation, it says so.

How an operator actually gets there

The sequence matters more than the destination, and it is unglamorous. An operator that automates a scenario it cannot yet measure has bought risk, not autonomy.

• Score one scenario honestly. Not the network — one operational flow, against the published methodology, with the evidence recorded. That gives a defensible starting number instead of an opinion.
• Fix the data before automating on it. Most operators discover at this step that the inventory and the network disagree, and every closed loop built on top of that disagreement inherits it.
• Automate with the autonomy as a setting. Run it in dry-run, then with a person approving every write, then unattended — promoted per automation as confidence is earned, not switched on across the estate.
• Prove the outcome on the real network. Independent confirmation that the change improved the customer's experience is the evidence a loop needs before anyone should let it run itself.

What we will not claim

No product confers an autonomy level, ours included. A level is scored for a scenario against a questionnaire and the evidence behind it, by an evaluator. We are certified to run that assessment and we build the platform that moves the score — those are two different statements, and a vendor that blurs them is selling you something.