AI use case

AI agent for device and connectivity troubleshooting on voice and chat

An AI agent that diagnoses and fixes a customer's broadband, mobile, TV or device problem by conversation on the phone or in chat, running line tests and remote resets through the operator's systems, guiding the customer step by step, and booking an engineer or handing over to a technician when the fault needs a person.

By Len Debets · Last verified 26 September 2026 · 4 public deployments

70%
Median containment rate
3 deployments.
About 45 million
Interactions handled
Vodafone (vendor claim).
USD 1.2 million to USD 8 million
Indicative value per year
An operator with 2 million broadband and mobile customers. Worked example, see how it is calculated.

What problem does it solve?

"My internet is not working" is one of the reasons customers contact a telecom operator. Many faults are resolved by the same few steps: check for an outage, run a line test, restart the router, check the cabling, change a setting on the phone. Yet each call ties up a trained agent, and a fault that could have been fixed remotely can still end in an engineer visit.

This use case adds an agent that can call the same diagnostics itself, read what the customer describes or photographs, and adapt the steps as it goes.

How does it work?

  1. Identify the customer and the service. The agent authenticates the caller and finds the affected line, device or package.
  2. Rule out the network first. It checks for known outages or maintenance in the area and, if there is one, explains it and gives an estimated fix time instead of troubleshooting.
  3. Run diagnostics. Through approved tools it runs a line test, reads the router or modem status and recent faults, and can trigger a remote reset or reprovisioning.
  4. Guide the customer. It walks the customer through the remaining steps in plain language, on voice or chat, and can interpret a photo of a device's lights or error screen.
  5. Escalate with the evidence. If the fault persists it books an engineer or passes the case to a human technician with the diagnostics already run, so nobody repeats the tests.
Audience
Customer facing
Autonomy
Supervised agent
Adoption
Early adopters
Channels
Phone and voice, Web chat, Mobile app, WhatsApp

What is it worth?

Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.

Value benchmarks for AI agent for device and connectivity troubleshooting on voice and chat
KPIMedianReported rangeData pointsClaimed by
Containment rate70%
44% to 73%
31 organization, 2 vendor
Interactions handledNot pooled
70,000 to 45 million
21 organization, 1 vendor

Value drivers: Lower cost to serve, Customer experience, Speed and cycle time, Employee productivity.

Indicative value

An operator with 2 million broadband and mobile customers

USD 1.2 million to USD 8 million

Human handled technical contact cost avoided per year

How this is calculated

Formula: customers * faultContactRate * resolvedShare * costPerContact. The low scenario uses every low input, the high scenario every high input.

InputLowHighBasis
Broadband and mobile customers customers, customers2,000,0002,000,000The reference operator.
Technical fault contacts per customer per year faultContactRate, contacts per customer per year0.40.8Editorial assumption, replace with the technical share of your contact reason report.
Share of fault contacts the agent resolves without a human resolvedShare, fraction of fault contacts0.250.5Editorial assumption: set below the Singtel benchmark on this page (73% of mobile and home troubleshooting cases resolved without a Customer Care officer in its initial results) because voice calls and complex home network faults are expected to resolve less often; replace with your own pilot data.
Cost of a human handled technical contact costPerContact, USD per contact610Editorial assumption for a technical support contact, which usually takes longer than a general one. Replace with your own fully loaded cost.

What it leaves out: Gross avoided contact cost only. It leaves out the larger saving from engineer visits that are avoided or better prepared, the cost of the AI and the diagnostic integrations, and any repeat contacts from faults the agent closed too early.

Who already uses it?

4 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.

Singtel

Singapore · Telecommunications · 2026

ProductionGrade B

In a partnership announced on 4 March 2026, Singtel upgraded the customer care assistant Shirley of Singtel Singapore with Sierra's agentic AI, starting with a pilot that went live in under ten weeks. Shirley verifies customer details, resolves mobile and home troubleshooting, completes roaming sign ups and understands local expressions including Singlish; customers purchased more than 200 roaming add ons independently. In its first six weeks it handled over 70,000 cases. Singtel Singapore also says it will deploy voice AI agents for outbound sales calls within defined compliance and governance standards.

  • Containment rate: 73%, mobile and home troubleshooting cases, initial results after launch
    "73% of mobile and home troubleshooting cases were resolved without requiring a Customer Care officer."
    Claimed by: organization
  • Automation rate: 76%, roaming sign up requests, initial results after launch
    "76% of roaming sign-up requests were completed successfully without requiring a Customer Care officer."
    Claimed by: organization
  • Interactions handled: at least 70,000, first six weeks after launch
    "Singtel went live in less than 10 weeks, and in the first six weeks since launch, ”Shirley” handled over 70,000 customer cases involving high volume requests in areas such as mobile issues and roaming services."
    Claimed by: organization

Virgin Media O2

United Kingdom · Telecommunications · 2026

PilotGrade B

In September 2026 Virgin Media O2 began a gradual rollout of an AI voice agent, built with Sierra, that handles selected routine broadband fault calls, which Virgin Media O2 says are a small proportion of total call volumes. Human advisors stay available to every caller, calls are monitored by Virgin Media O2 teams, and customer feedback is used to judge performance before the agent is extended. The launch sits alongside a specialist team of more than 500 agents for complex and sensitive cases and the Lumi AI advisor tool. No results have been published yet.

No outcome disclosed.

Vodafone

United Kingdom · Telecommunications · 2024

ScaledGrade C

TOBi is Vodafone's digital assistant on the website, the My Vodafone app, messaging and telephony, first launched in Italy and extended to 15 language versions. It handles billing questions, contract updates and simple technical troubleshooting, hands over to a live agent with a summary, and during the pandemic made the same sales offers as human agents, such as data boosts and upgrades. Microsoft reports that TOBi now fully resolves 70% of inquiries arriving through digital channels; an earlier Microsoft story quotes Vodafone on a 12% year on year fall in contacts to call centres after launch.

  • Containment rate: 70%, inquiries arriving through digital channels
    "Currently, TOBi handles nearly 45 million customer calls a month, fully resolving 70% of customer inquiries coming through the company’s digital channels."
    Claimed by: vendor
  • Interactions handled: about 45 million, per month
    "Currently, TOBi handles nearly 45 million customer calls a month, fully resolving 70% of customer inquiries coming through the company’s digital channels."
    Claimed by: vendor
  • Contact deflection: 12%, frequency of customer contacts to call centres, year over year after the TOBi launch
    "Since launching TOBi, we’ve reduced the frequency of customer contacts to call centers by 12 percent year-over-year"
    Claimed by: organization

Vodafone Germany

Germany · Telecommunications · 2020

ScaledGrade C

Vodafone Germany moved its messaging channels into one central team and put the TOBi chatbot in front of every WhatsApp, Apple Business Chat and SMS conversation, with a handover to a human agent on the same screen. TOBi understands more than 230 intents and can classify the photos and screenshots customers send, such as a bill or a router with a flashing red light. Genesys reports that the share of inquiries resolved by AI (which it calls first contact success) rose from 16% at launch to 44%.

  • Containment rate: 44%, share of messaging inquiries resolved by TOBi without a human
    "44% of inquiries now resolved by AI, up from 16%."
    Claimed by: vendor

How do you implement it?

A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.

Data you need

  • Troubleshooting guides per device and technology with an owner and review date
  • Outage and planned maintenance status by area, available by API
  • Fault contact reasons with their eventual root cause, for testing
  • A list of fault types that must always go to a technician

Systems to integrate

  • Network diagnostics and line test tools
  • Device management for routers and set top boxes (status, reboot, reprovision)
  • Outage and maintenance status service
  • Field service scheduling for engineer appointments
  • Contact centre platform for handover with diagnostic results

Complexity: Medium

Knowledge based troubleshooting is simple. The value comes from connecting diagnostics (line tests, device management, outage status) and engineer booking, which live in network and field systems that were not built for real time use by a conversation.

  1. 1

    Start with the faults that have a known fix

    Take the top fault reasons by volume and pick those with a reliable remote fix, such as router restarts, settings on the phone and account provisioning errors. Leave intermittent and physical faults to people at first.

  2. 2

    Connect diagnostics before you tune the conversation

    An agent that can run a line test and read router status can resolve faults that questions alone cannot. Build the diagnostic tools and their error handling first.

  3. 3

    Always check for an outage first

    Make the outage check the first tool call on every fault conversation, so the agent never walks a customer through resets during a network incident.

  4. 4

    Define the escalation evidence

    Agree with field operations what a technician or engineer needs from the agent (tests run, results, customer availability) and pass it in every escalation.

  5. 5

    Pilot on selected calls and listen

    Route a small share of routine fault calls to the agent, monitor calls, measure repeat faults within 14 days and expand only when they hold.

Guardrails

  • Outage check before any troubleshooting step
  • Only allow listed remote actions, each with a confirmation to the customer
  • Handover to a person on request at any time, on voice and chat
  • Detection of vulnerable customers and telecare or medical alarm users, who go straight to a person
  • Personal data masked before text reaches a model or the logs

KPIs to instrument

  • Resolution rate per fault type, counting a repeat fault contact within 14 days as unresolved
  • Engineer visits booked per 1,000 fault contacts, and visits that found no fault
  • Average handling time of escalated cases, compared with cases without the agent
  • Customer satisfaction on resolved and escalated conversations
  • Share of fault conversations that started during a known outage

Human in the loop

Technicians and engineers own every fault the agent cannot fix and every physical repair. Supervisors monitor live calls during the pilot, review a weekly sample of resolved cases against repeat faults, and approve every new remote action before the agent can use it.

Common failure modes

Troubleshooting during an outage
Customers are asked to reset routers while the network is down. Make the outage check mandatory and visible.
False resolution
The customer says it works now, the fault returns the next day. Measure repeat faults, not the end of the conversation.
Endless loops
The agent repeats the same steps with a frustrated caller. Limit attempts and escalate with the evidence.
Missing the vulnerable customer
A customer relying on a telecare alarm is left in automation. Detect these signals and hand over at once.

What are the risks and rules?

EU AI Act

Limited risk (transparency)

A customer facing troubleshooting agent must tell people they are interacting with AI unless that is obvious (Article 50). It is not high risk as long as it is not used as a safety component in the management and operation of critical digital infrastructure (Annex III, point 2); running line tests and resets for one customer's service does not make it one.

Guidance

Controls to put in place

  • AI disclosure at the start of every call and chat
  • Documented list of remote actions with owners, tests and rollback
  • Call recording and transcript retention in line with local rules
  • Regression tests on real fault scenarios for every change to prompts, tools or model
  • Monitoring of repeat faults and complaints that mention the agent

Frequently asked questions

What share of technical support contacts can an AI agent resolve?
Few operators publish results, and published voice results are scarce. Singtel, which launched its agentic assistant across chat and voice, reports that 73% of mobile and home troubleshooting cases were resolved without a Customer Care officer, without saying which channel produced that figure. Genesys reports that first contact success on Vodafone Germany's messaging channels rose from 16% at launch to 44%, for a TOBi chatbot built on IBM Watson before generative AI. Virgin Media O2 began with selected routine broadband fault calls in September 2026 and has not published results yet.
Should an AI agent troubleshoot on the phone or only in chat?
Both work. Chat lets customers read instructions and send photos, such as the flashing router light that Vodafone Germany's TOBi recognises, while voice reaches customers whose home internet is down. Singtel launched its agent across chat and voice, and Vodafone runs TOBi on digital channels and telephony. Vodafone Germany's published figure covers messaging only, and Virgin Media O2 started with a narrow set of voice calls that its teams monitor.
How does this relate to outage communication?
They share the outage check. The troubleshooting agent must know about an outage before it starts, and the outage communication agent tells affected customers before they call.

How to cite this page

Blits.ai AI Use Case Library, "AI agent for device and connectivity troubleshooting on voice and chat", last verified 26 September 2026, https://www.blits.ai/ai-use-cases/device-and-connectivity-troubleshooting-agent. Licensed under CC BY 4.0. Method: how we verify use cases.

Changelog
  • 27 September 2026: First published

Related use cases

Telecommunications

AI agent for network outage detection and customer communication

An AI agent that turns network alarms into a clear picture of which customers are affected by an outage and why, tells them proactively by message, app or phone with a cause and an estimated fix time, answers their questions during the incident, and updates them until service is restored.

Deployments
1 public, best grade B
Autonomy
Supervised agent
Telecommunications

AI copilot for field technicians and dispatch optimization

AI that decides whether a fault needs a site visit at all, predicts what work and parts a job will need, helps plan and update appointments, and gives technicians on site guided diagnosis and instant answers from manuals and past jobs, so more jobs are fixed on the first visit.

Deployments
2 public, best grade B
Autonomy
Copilot
Telecommunications

AI agent for telecom bill explanation and billing disputes

An AI agent that explains a customer's telecom bill line by line, in plain language and on any channel, answers why a charge changed or appeared, corrects clear errors within set limits and opens a billing dispute with the evidence attached when a human has to decide.

Deployments
4 public, best grade B
Reported first contact resolution
60%
Vodafone, organization claim
Telecommunications

AI assistant for telecom order to activation and eSIM onboarding

An AI assistant that takes a new or existing customer from order to a working service: it collects and checks the order details, guides number porting, eSIM download or SIM activation and installation appointments, tracks the order and fixes or escalates the step that is stuck, on messaging, app, web or phone.

Deployments
3 public, best grade B
Reported automation rate
76%
Singtel, organization claim
Cross industryBanking

AI agent for first line contact centre service

An AI agent that answers the first line of inbound customer contact on phone, chat and messaging, resolves general and routine questions end to end in the customer's own language, and routes everything complex, sensitive or regulated to the right human team with the context attached.

Deployments
18 public, best grade B
Median containment rate
47%
7 deployments
Telecommunications

AI for predictive network maintenance in telecom

Machine learning that spots the early signs of network failure, such as degrading cells, faulty customer equipment, ageing hardware or planned digging near fibre, and triggers a preventive fix, a remote reset or a targeted intervention before customers lose service.

Deployments
6 public, best grade B
Autonomy
Supervised agent