Behind-the-scenes orchestration view

Executive Walkthrough

A customer experiences poor internet performance. Traditionally, resolution means five manual steps and days of waiting. Agentic AI completely changes the experience: it understands the issue, gathers network context, uses multimodal diagnostics and resolves problems autonomously.

Want the customer's point of view instead? Open the User Demo for the full guided router-reset conversation.

1. Call support
2. Wait in queue
3. Repeat issue
4. Run diagnostics
5. Book technician

Why it exists

Home internet faults generate the highest call volume and the most expensive truck rolls in telecom. Most of them are resolvable remotely — if someone can see the network, the devices and the gateway at the same time.

What it does

An orchestrator agent interprets the customer's own words, dispatches specialist agents in parallel, requests multimodal evidence such as a photo of the modem, then executes the fix on the network and confirms it with the customer.

How it changes support

Support shifts from a queue of tickets handled by humans to autonomous resolution, with people reserved for the small share of faults that genuinely need a technician on site.

The Agents Involved

Four specialist agents work in parallel under an orchestrator, with a fifth agent for assisted resolution when a site visit is genuinely required.

Customer Context Agent

Consults: Customer profile, Service tier, Equipment, Previous tickets, Home topology.

Network Operations Agent

Consults: Fibre status, Service health, Last 24 hours performance, Telemetry.

Device Analytics Agent

Consults: WiFi channels, Device congestion, Coverage maps.

Visual Diagnostics Agent

Consults: Camera capture, Computer vision analysis, LED state model.

Field Service Agent

Consults: Technician inventory, availability, routing, customer preferences.

Live line telemetry

MonitoringSampling every 0.70s
Inject fault

Scenario: Fibre degradationOptical signal drifting out of spec — intermittent dropouts and buffering.

Signal quality
88.0%
Throughput
720 Mbps
Packet loss
1.40%
Latency
22 ms

Business Outcomes

Modelled impact once agentic resolution handles first-line home internet faults.

80%
Issues resolved without an agent
Deflected from the contact centre
-62%
Truck rolls avoided
Faults fixed remotely instead of on site
3 min
Average time to resolution
Versus 2–4 days for a scheduled visit
+21 pts
NPS on resolved contacts
Proactive, explained, verified fixes

Microsoft Architecture

The solution runs entirely on Microsoft cloud and AI services, from the customer experience layer down to network operations and field dispatch.

Experience
Copilot StudioTeams / Mobile appAzure Communication Services
Reasoning
Azure AI FoundryGPT modelsAzure AI VisionAzure AI Speech
Context
Dynamics 365 Customer ServiceMicrosoft FabricCustomer 360
Operations
OSS/BSS connectorsNetwork telemetryAzure IoT device management
Assisted resolution
Dynamics 365 Field ServiceTechnician schedulingLive ETA notifications

Full layer-by-layer view on the Architecture page.