NRBY - NOAX
Vol. I
A ThesisWhy AI Alone Will Not Fix Network Operations

AI cannot outperform the process
that labels its data.

In today’s network operations, a ticket can close while the problem continues. If that closure becomes the training label, the model learns a false success.

ClosedSolved
ScrollThe thesis
00Core thesis

The telecom industry has automated work creation.
It has not automated work governance.

If AI learns from ticket closure instead of network convergence, it will optimize the wrong outcome.

02The current telco process

Two processes. Two different futures.

Today, automation turns signals into work. Tomorrow, NRBY - NOAX decides whether the work should exist.

TodayUNGOVERNED

Automation creates work.

Signal
Automation
Ticket
Dispatch
Resolution
Repeat
TomorrowGOVERNED

NRBY - NOAX decides if work should exist.

Signal
Automation
NRBY - NOAX
Approve Observe Suppress Combine Reroute Escalate Capital Review
Decision
Action
Convergence
Memory
03The audit snapshot

The evidence.

Measured from a 12-month Tier-1 network audit

0
OSP dispatches reviewed
0%
Machine-created dispatch demand
0
Proven waste dispatches
0
Technician hours consumed
0/day
HFC repeat dispatches
0
Revolving-door nodes
$0.0M
Proven OSP waste
04Why the old AI fails

A model that learns from closure learns the wrong thing.

The model sees
Signal
Dispatch
Ticket Closed
Labels it: Success
What actually happened
Node still degraded
Customer calls again
New ticket created
Same dispatch repeated
Wrong label
Ticket Closed
Correct label
Network Converged

“The model learned the closure. It did not learn the outcome.”

05Bad training labels

Garbage labels create expensive intelligence.

Example 1
Ingress Repaired
Ticket Closed
Signal Returns
AI learns
Ingress Repaired = Success
Reality
No convergence
Example 2
No Trouble Found
Ticket Closed
Issue self-healed before crew arrived
AI learns
Dispatch completed
Reality
Dispatch was unnecessary
Example 3
Repeated Dispatch
Same Resolution
Same Node Still Degraded
AI learns
More work required
Reality
Escalation required
06Time contains information

Time contains information.

Two identical alerts at 7:00 PM can require two different decisions.

Example A7:00 PM
Healthy Node
Improving health trend
No prior failures
ActionObserve
Example B7:00 PM
3 prior dispatches
Declining health
Recurring failure
ActionEscalate
0.0%

of self-healed intermittent dispatches were created between 1pm and midnight.

A 4-hour observation window would have prevented 2,447 dispatches annually.

“The signal was not the problem. The missing context was.”

07The repeat loop

The loop closes tickets. It does not close problems.

Repeat loop
Signal → Dispatch → Closed → Signal → Dispatch → Closed → Signal → Dispatch → Closed →
0
Revolving-door nodes
0
Truck rolls
$0.00M
Spent on nodes still degraded

“Closure records activity. DTF records whether the activity worked.”

08Machine-created demand

The problem is not automation.
The problem is ungoverned automation.

Machine-created work is not inherently bad. Some automation performs as well as humans. But at machine scale, small decision errors become millions in waste.

fsm2remedy78.6%
ospparser9.3%
custdownbot2.7%
nrby_user0.3%
humans9.2%

The control point is not after dispatch.
It is before labor is committed.

09Capacity is not always the answer

Not every symptom is a capacity problem.

A customer symptom is not a capital justification.

Symptom
Capacity Constrained
Action
Capital Review
Symptom
Capacity Candidate
Action
Join Utilization Evidence
Symptom
Chronic Plant
Action
Engineering Review
Symptom
Ingress Driven
Action
Plant Cleanup
Symptom
Self-Healing
Action
Observe / Suppress
Symptom
Governance Failure
Action
Combine / Reroute / Stop Repeat

“The wrong capital project does not eliminate operational demand. It only moves spend from OPEX to CAPEX.”

NRBY - NOAX separates capacity problems from plant problems, governance problems, and self-healing events before labor or capital is committed.

10FTTH first impression

Fiber has a different failure mode.

0
Fiber customers monitored
0
With service issues in last 30 days
0
With optical power issues
0
New customers landed on troubled ports in 3 months

In fiber, the highest-value intervention may happen before the customer is activated.

Decision
Port Health
Install Readiness
Remediate Port
Activate Customer
Action: Do not light up customers on degraded ports.
11What NRBY - NOAX solves

NRBY - NOAX changes the decision layer.

Governance
NRBY - NOAX

Decides whether work should be approved, observed, suppressed, combined, rerouted, escalated, or moved to capital review.

ApproveObserveSuppressCombineRerouteEscalateCapital Review
Memory
DTF

Remembers what was tried, whether it worked, whether the signal returned, whether the node converged, whether another dispatch should be blocked.

TriedWorkedReturnedConvergedBlocked
Execution
NRBY

Routes the right work to the right team with the right evidence at the right time.

Right WorkRight TeamRight EvidenceRight Time
11·SIMTry it · Decision simulator

Choose a scenario. Make a decision. Watch the label change.

Every decision produces a different outcome — and a different training label. The same signal can teach the model truth or teach it noise, depending on what the system decides.

01 Choose a scenario
Context
Time: 7:00 PMNode health: improvingNo prior failures in 30 daysSignal cleared 11 min after creation
02 Make a NRBY - NOAX decision
03 Outcome & training label
Pick a decision to see how the labels and outcome change.
04 Training dataset · what the model learns from you
Dataset integrity
0%
● Optimal 0● Acceptable 0● Wasteful 0● Harmful 0
0 samples in dataset

Train at least one decision to see how the model's worldview takes shape.

Every label you train with becomes a row in the model's worldview. NRBY - NOAX is the layer that decides which labels are true enough to learn from.

12The new training labels

Better AI starts with better truth.

Old labels
Ticket Closed
Dispatch Completed
Resolution Entered
New labels
Converged
Flat
Degraded
Repeated
Self-Healed
Escalation Required
Capital Required

When the labels change, the models change. AI can finally learn which actions improve the network, not which actions close tickets.

13Final

The future is not more automation.

It is governed automation.

Automation
creates work
NRBY - NOAX
governs the decision
DTF
remembers the outcome
NRBY
executes the action

AI does not fix a broken process.
It scales it.
NRBY - NOAX changes the process first.