The best support call is the one that never happens.The best support call is the one that never happens.
DvSum is building the operating model that gets network operators there — starting with 41.7% call containment at Liberty Latin America, in production today.
NTF dispatches only · misdiagnosis stopped at source
The Problem
Contact center AI was built for CRM data. Tech support needs your network.
Wrong tool for the job
Contact centers run on scripts and CRM data — it works for billing questions, account changes, service upgrades. But 40% of your calls are tech support. Those calls can’t be resolved with CRM data. They need live CMTS telemetry, provisioning state, CPE health, and in-home Wi-Fi data — simultaneously. No script covers that.
The compounding cost
Most contact center AI can’t tell the difference between a plant issue and an in-home problem. So your agent navigates 4–6 systems manually while the subscriber waits. They guess. A plant issue gets dispatched as a CPE swap. A billing block gets diagnosed as a modem fault. Every wrong diagnosis compounds — in your operations budget and in your churn rate.
AHT
8 minutes to diagnosis
Agent spends the first 90 seconds of every tech support call just figuring out which domain to look at first — before any real troubleshooting begins.
First Call Resolution
Wrong domain = callback next week
Misdiagnosis doesn’t just add handle time — it generates the wrong action. The subscriber calls back. FCR drops. Your agents handle the same problem twice.
NFF Truck Rolls
Tech finds nothing. Every time.
Plant issue treated as CPE swap. Provisioning error dispatched as field visit. The truck rolls, the tech finds nothing, and the cost hits your operations P&L.
Why tech support is different
Any one of these five domains could be the cause. Your agent has to check all of them — manually.
01
Billing
Account suspended, restricted, or misconfigured — no amount of modem rebooting fixes this.
02
Provisioning
Service incorrectly configured in OSS — SVLAN, disabled package, CPE never fully activated.
03
Wi-Fi
In-home interference, weak RSSI, no band steering — nothing to do with the plant.
04
CPE / Modem
FEC uncorrectables, T3/T4 timeouts, upstream power out of range — modem swap or reboot.
05
Plant / Network
Node-level or segment-level issue — upstream SNR degraded across multiple modems.
Most contact center AI doesn’t know what FEC uncorrectables mean — and cannot tell the difference between a plant issue and an in-home problem. DvSum does.
The Operating Model
From reactive to zero-touch. Three stages. One platform.
REACTIVE
When they call, your agent already has the answer.
Five-domain diagnosis before the agent says hello. Right verdict, right action, first time. No dead air. No NFF truck rolls. No guessing across four screens.
Live in production
PROACTIVE
Before they call, DvSum already saw it coming.
QoE degradation detected before the subscriber notices. Automated outreach goes out. The inbound call never enters the queue. At-risk subscribers identified before churn, not after.
Coming next
PREVENTIVE
The subscriber never knew there was a problem.
Always-on baseline per subscriber. Deviation triggers remediation before escalation. SLA breach surfaced before it happens, not after. The network acts. The subscriber stays.
The destination
Most contact center transformation stops at Stage 1. DvSum is the platform that takes you through all three.
Subscriber Experience
The experience we help you deliver.
"
My internet just works.
DvSum detected a degrading upstream node, applied a remote fix at 3 AM. Subscriber never lost connection. Never called.
DvSum detected CPE anomaly, pushed remote remediation. Subscriber received: "We detected and resolved a Wi-Fi issue in your home."
"
When I did call, it was actually quick.
Agent had full diagnostic history on screen before pickup. No "have you tried restarting your router?" Resolved on the first call.
How DvSum Gets You There
DvSum doesn't wait for the call. It already knows — and when the call arrives, the answer is ready.
Assemble
One view. Before the agent speaks.
CMTS telemetry + provisioning state + CPE health + managed Wi-Fi data + CRM + call history — unified into a single subscriber health snapshot, refreshed continuously. No dead air. No reset script as a first move.
Isolate
Plant → Modem → Wi-Fi → verdict.
Wall-to-wall triage chain: Plant → Modem → Gateway → Wi-Fi → Device. Each step either exits with a verdict or passes to the next — in seconds. The output is plain language: "Outside plant signal issue — escalate to NOC, do not dispatch." No RF engineering required.
Sense
Plant-side degradation flagged before they call.
Continuous monitoring surfaces FEC drift, upstream SNR decline, and intermittent modem flap — and links it to specific subscribers before they call. When that subscriber does call, the plant-side finding is already flagged. Not because an alarm fired. Because DvSum was watching.
Pre-Compute
Warm result. Under 3 seconds.
Every subscriber's signal chain is analyzed continuously on a rolling basis and cached. When the call arrives, the agent retrieves a warm result — not a cold query. Not faster compute on demand — pre-computed. The same analysis that could have triggered a proactive outreach is instantly available as the triage answer.
In Production
Live at Liberty Latin America. Real contact center. Real numbers.
41.7%
Call containment rate
Network-related tech support calls only · LLA · Genesys Engage
Call open to agent verdict · network-related calls · LLA
50%
NTF truck roll reduction
NTF dispatches only — misdiagnosis stopped at the contact center
All metrics from Liberty Latin America production deployment. Genesys Engage integration. 41.7% containment applies to network-related tech support calls only. AHT reduction is blended across call types — not a flat rate on every call. NTF truck roll figure applies to NTF-specific dispatches only.
Same desktop. Same IVR. The intelligence underneath just got a lot smarter.
DvSum surfaces the diagnosis on the agent’s existing desktop — no new tab, no new system. When the call arrives, a co-pilot card auto-populates: domain verdict, confidence score, recommended action, and subscriber context. The agent’s first sentence is a statement: “I can see what’s causing this — let me try a remote fix.”
IVR / Virtual Agent
1Subscriber connects → DvSum invoked in background
2Warm analysis retrieved — 5-domain verdict ready
3Virtual agent delivers verdict + remote action offer in natural language
4Subscriber confirms → resolved, no human agent needed
CSR / Genesys Co-Pilot
1Call arrives at agent → co-pilot card auto-surfaces
2Domain verdict + confidence + recommended action — no toggling
3Agent one-clicks remote fix or dispatches with pre-loaded RCA
4First call resolution — right diagnosis, right action, first time
DvSum Intelligence Layer
Called as an API · surfaced in the existing desktop · invisible when not needed
DvSum is not a replacement for your contact center platform. It is the intelligence layer underneath — called as an API, surfaced in the existing desktop, invisible when not needed.
Integrations
Works with your existing contact center stack and OSS/BSS — without ripping anything out.
DvSum integrates with Genesys Engage today — in production at Liberty Latin America. Genesys Cloud integration is in beta, with GA planned for Q4 2026. For other contact center platforms, DvSum surfaces via API — if your platform supports co-pilot cards or iframe embeds, we can connect.
On the act layer: every recommended action — remote reset, Wi-Fi channel push, dispatch work order — requires agent or subscriber confirmation before execution. Every action is logged with its full evidence chain. Human agents stay in the loop. You define the boundary.
See what your contact center is missing before it costs you a subscriber.
We'll show you what DvSum surfaces on a real call from your network. No sandbox. No demo environment. Your subscribers, your nodes, your agent desktop.
