Back to Case Studies
TelecommunicationsUnited KingdomAI Agents

Major UK Telco: How a Wholesale Operations Team Agentified the NTE Shift Provisioning Journey

An autonomous AI agent reads unstructured reseller messages, validates charges in context, books Openreach appointments, and closes the loop the next day — escalating only the cases that genuinely need a human.

~30
Portal steps automated per order
4
Core systems orchestrated
3
Autonomous decision paths
10+
Fields extracted from free-text

About the Client

The client is one of the UK's largest wholesale broadband and network operators, providing fixed-line provisioning services to a broad base of resellers and communications providers. Their wholesale operations team supports reseller-raised service requests across a continuous, high-volume pipeline.

The Challenge

A Manual Provisioning Chain Across Multiple Systems

One of the most repetitive processes handled by the operations team is the NTE Shift — a request to move a customer's Network Terminating Equipment to a new location within the premises.

The original process was entirely manual. Team leads filtered the Grab Queue in their Fault Management System (FMS), copied fault IDs into Excel, and distributed them by hand. Associates then logged into their Network Asset Management System (NAMS) to perform structured lookups and verify reseller aliases. After verification, they switched into the external Openreach Sabor portal, working through roughly thirty manual steps to book each appointment.

The critical point of failure was the free-text Reseller Interactions message. Buried in unstructured prose were contact details, available dates, site safety notes, and explicit confirmation of charge acceptance. Traditional enterprise RPA tools could click through the external provisioning portal, but none could reliably read the free-text. Rule-based extraction broke down the moment a reseller altered their phrasing.

The process demanded an agentic model: deterministic bots for structured system work, and an AI agent for true language comprehension and decision-making.

The Solution

An Agent-Orchestrated Provisioning Pipeline

The deployment utilized an architecture built on a single, powerful principle: the AI agent is the orchestrator. Deterministic checks remain in code, while reasoning over language is managed by the agent, ensuring each task is routed to the exact right tool.

STEP 01

Automated Ticket Intake from FMS

A poller bot continuously scans the Grab Queue in the internal Fault Management System for NTE Shift tickets. For every matching ticket, it captures the CLI (Calling Line Identification), the reseller name, and lifts the entire raw Reseller Interactions message as a single text blob — bypassing the manual Excel allocation entirely.

STEP 02

Deterministic Pre-Checks in Code

Before AI is invoked, the NAMS Bot performs structured, binary checks within the asset management system. It verifies the routing path, extracts the local loop unbundling (LLU) reference, and validates the reseller alias. If these deterministic checks fail, the bot rejects the ticket directly — preserving AI computing power for complex reasoning tasks.

STEP 03

Agentic Extraction, Reasoning, and Tool Selection

The AI Agent receives the validated payload and the raw reseller message. It extracts all required booking fields from natural language and validates charge acceptance in context. The agent then autonomously calls one of three tools: book the appointment, reject the ticket, or trigger a human handoff with a structured summary explaining what was extracted and what action the advisor needs to take.

STEP 04

External Booking, Snooze, and Same-Loop Closure

When a booking is approved, the Booking Bot executes the full ~30-step appointment flow in the Openreach Sabor portal, matching preferred slots and submitting the order. The Action Bot snoozes the ticket automatically. On the resume date, the Check Bot verifies the appointment status in OR-HUD (the Openreach tracking dashboard) and completes the ticket lifecycle.

Four Core Systems Orchestrated

Fault Management System (FMS)
Ticket intake, queue scanning, and status updates
Network Asset Management (NAMS)
Routing path verification and reseller alias validation
Openreach Sabor Portal
External ~30-step appointment booking workflow
OR-HUD Tracking Dashboard
Day+1 automated follow-up and lifecycle closure

The Results

Provisioning Without the Manual Chain

Ticket AllocationEliminated entirely via automated polling
Reseller Message ReviewAI agent extracts fields directly from free-text; no brittle regex required
Charge AcceptanceValidated contextually by AI, rather than by basic string-matching
Appointment BookingBooking Bot executes ~30 portal steps automatically using the structured AI payload
Next-Day Follow-UpCheck Bot runs automatically on the resume date to close or amend
Human ExceptionsAdvisors receive AI-generated summaries; manual takeover only for true exceptions

Why It Matters

This NTE Shift transformation highlights what separates modern agentic automation from traditional RPA. Classic bots can click through a portal, but they stall the moment a process depends on understanding unstructured human language.

By making the agent an autonomous actor rather than a passive advisory step, this solution manages real operational variation at scale. It delivers a highly practical operating model: deterministic automation where the rules are absolute, AI reasoning where language is variable, and human intervention only where uncertainty remains.

"What sold us was that the platform understands the request, triggers the right bot, and escalates only when human judgement is genuinely needed."
WO
Wholesale Operations Leader
Major UK Telco

Ready to automate your telecom provisioning?

Talk to an automation specialist who understands telecom operations, provisioning workflows, and agentic AI.