AI Voice Agents for Fault Management & Billing at a Fast-Growing UK Consumer ISP

Client Profile

A fast-growing challenger broadband provider in the United Kingdom, known for its customer-first approach, transparent pricing, and commitment to ethical business practices. The company operates on a lean model — small team, strong brand, and a growing subscriber base that expects a premium support experience to match the premium positioning.

The ISP purchases wholesale connectivity from upstream infrastructure providers and delivers it to residential customers under its own brand, managing the full customer lifecycle from sign-up through support and retention.

Industry: Telecommunications (Consumer Broadband) · Region: United Kingdom · Products Used: VoiceFlow AgentIQ · WhatsApp Integration · CRM Connector

The Challenge

Growth was creating a support crisis. As the subscriber base expanded, two specific pain points were consuming a disproportionate amount of agent time and operational budget:

Pain Point 1: Fault Calls Were Mostly Not Faults

The ISP was receiving over 1,000 fault-related calls per month. Analysis showed that approximately 70% of these were “no-fault-found” scenarios — cases where the customer’s internet was actually working correctly, or where the issue was with the customer’s own equipment (router placement, device configuration, WiFi interference) rather than the network.

Each of these calls consumed 10–15 minutes of agent time: verifying the account, running through diagnostics, checking the upstream network, and ultimately concluding that there was no fault to fix. At 700+ calls per month, this represented hundreds of wasted agent hours — time that should have been spent on genuine faults, complex technical issues, or proactive customer engagement.

The problem was compounded by the diagnostic workflow itself. Agents followed a manual checklist that couldn’t adapt to the specific situation. A customer calling about intermittent evening slowdowns received the same diagnostic script as someone reporting a complete outage — even though the causes, solutions, and urgency were completely different.

Pain Point 2: Billing Collections Were Manual and Expensive

The ISP operated an 8-stage dunning process for billing collections — from initial payment reminder through to final disconnection notice. Every stage required manual action: generating the communication, sending it via the appropriate channel, waiting for a response period, and then triggering the next stage if no payment was received.

This process consumed significant agent time, was prone to inconsistency (different agents executing stages at different times), and created customer experience issues when communications were delayed or duplicated. The emotional labour of collections calls — asking customers to pay overdue bills — also contributed to agent burnout.

Integration Complexity Added Friction

The ISP used a proprietary CRM system (not an off-the-shelf solution) alongside an external knowledge management platform and upstream wholesale systems for network diagnostics. Any AI solution needed to integrate with this specific toolset — not with generic, pre-built connectors, but with custom API integrations tailored to their architecture.

Our Approach

We deployed two specialised AI agents, each purpose-built for one of the two core challenges. Rather than a single generalist agent trying to handle everything, the specialised approach allowed each agent to be deeply optimised for its specific domain — with tailored conversation flows, integration points, and success metrics.

What We Built

Agent 1: Fault Management Agent

An AI voice and chat agent designed to be the first point of contact for all fault-related enquiries. The agent combines natural conversation with real-time network diagnostics to resolve the majority of queries without human intervention.

Core capabilities:

• Intelligent triage — within the first 30 seconds of the call, the agent classifies the likely fault category (complete outage, intermittent connectivity, slow speeds, specific device issue, upstream network fault) and routes the conversation down the appropriate diagnostic path. No more one-size-fits-all scripts.
• Real-time network diagnostics — the agent queries the upstream wholesale network in real-time, checking for known outages, planned maintenance, and line-level performance metrics. If there’s a confirmed network issue, the customer is informed immediately with an estimated resolution time — call handled in under two minutes.
• Guided customer-side troubleshooting — for no-fault-found scenarios, the agent walks the customer through targeted diagnostics: router restart, WiFi channel check, device-specific configuration, interference identification. The guidance is conversational and adaptive — if the customer says they’ve already rebooted the router, the agent skips that step.
• CRM integration — every interaction is logged in the proprietary CRM with structured data: fault category, diagnostic steps performed, outcome, and any follow-up required. If the issue requires escalation, the human agent receives a complete diagnostic report rather than starting from scratch.
• Pattern recognition — the agent tracks recurring issues per customer and per area. If a customer calls about slow speeds for the third time in a month, the conversation adapts: “I can see we’ve looked at this before — let me check if something has changed on the network side rather than going through the usual steps again.”

Agent 2: Billing & Collections Agent

An AI agent managing the full dunning lifecycle — from gentle payment reminders through to formal collections processes.

Core capabilities:

• Automated dunning workflow — the agent executes each stage of the 8-stage collections process on schedule, generating and sending communications via the appropriate channel (email, SMS, voice call) at the correct intervals. No manual intervention required to advance through stages.
• Payment processing — when a customer responds to a collections communication (or calls proactively about their bill), the agent can process payments immediately, update payment methods, and set up payment plans within defined parameters.
• Empathetic collections conversations — billing difficulties are often linked to personal financial stress. The agent is designed to handle these conversations with sensitivity: acknowledging the situation, explaining options clearly, and avoiding aggressive or pressuring language. The goal is payment recovery, not customer alienation.
• Account management — beyond collections, the agent handles general billing queries: explaining charges, confirming payment dates, processing refunds, and updating account details.
• WhatsApp integration — billing reminders and conversations available via WhatsApp Business API, meeting customers on the channel they’re most responsive to. Open rates on WhatsApp billing reminders significantly exceed email.

Projected Impact

Metric	Target
Fault call volume reduction	70% through AI-first resolution
No-fault-found call handling	Automated resolution without human agent involvement
Fault classification accuracy	90%+ on first contact
Availability	24/7 (from business hours)
Dunning process consistency	100% automated, on-schedule stage progression
Payment recovery rate	Improvement through timely, multi-channel communication
Agent time freed	Hundreds of hours per month redirected to complex cases

Why This Matters

For a challenger ISP, support economics directly determine scalability. Every subscriber added increases support volume — if that volume requires proportional headcount growth, the unit economics of the business deteriorate as it scales. AI-first support breaks this relationship: adding subscribers adds volume, but not cost.

The fault management agent is particularly significant because it addresses a problem that’s universal across consumer broadband. Every ISP deals with no-fault-found calls. Every ISP has agents spending time on issues that turn out not to be issues. The difference is that this ISP now resolves those calls in two minutes with an AI agent instead of fifteen minutes with a human — and the customer often gets a better experience because the diagnostic is faster, more thorough, and available at any hour.

The collections automation addresses the other end of the customer lifecycle with equal impact. Consistent, timely, empathetic billing communications recover more revenue than manual processes that depend on agent availability and prioritisation. And by removing the emotional burden of collections calls from human agents, it reduces a significant driver of support team burnout.