Industries

Telecommunications

Provisioning, billing and care that scale by subscriber, not by headcount.

Telcos run some of the oldest systems in software against some of the most impatient customers. The work that pays is rarely a rip-and-replace: it is putting a clean, evented layer over OSS/BSS so activation, billing and care stop contradicting each other — then automating the contact types that should never have reached a human.

Book A Call Run the numbers first

Google Cloud for telecommunications

The telecom stack maps closely onto the work we already do: OSS/BSS transformation, contact centre modernisation, subscriber insights and network root-cause analysis. We build against Dialogflow CX, BigQuery and Pub/Sub where they are the shortest credible path.

01 · Assure

Pub/Sub · BigQuery

Network and OSS events into one evented layer over the existing stack.
02 · Diagnose

Network root-cause analysis · Vertex AI

Faults grouped by cause rather than by alarm.
03 · Resolve

Contact Center AI · Dialogflow CX

Eligible contacts answered, with a clean handoff and a usable transcript.
04 · Monetise

Telecom Subscriber Insights · SMB MPaaS

Catalogue, entitlement and billing exposed as APIs.

68.4%: of telecom CEOs expect gen AI to significantly change how their company creates, delivers and captures value within three years; Source: Industry benchmark, paraphrased
13.8×: increase in users on an SMB platform in three months, from a client launch; Source: Client engagement, anonymised

SASB TC-TL — Telecommunications

SASB standards ↗

Telecom has the most mature disclosure set of any industry here: energy and the equipment-recovery line are both auditable today, which makes them a good place to start.

TC-TL-130a.1
Operational energy

total energy consumed (GJ), % grid, % renewable

Network and data-centre telemetry is already collected; we join it to site and tariff data so renewable share is reported, not estimated.
TC-TL-130a.1
Emissions intensity

scope 1 and 2 (tCO2e), per subscriber and per terabyte

Usage-based intensity is the number that survives a traffic-growth argument — and it falls out of the same pipeline.
TC-TL-440a.1
Product end-of-life

% of equipment recovered, by weight

Reverse-logistics events captured at the returns flow, reconciled against procurement volumes.
TC-TL-230a.1
Data security

breach count, % involving PII, subscribers affected

Subscriber data is the crown jewel here. Field-level PII tagging and access logging make the disclosure a report, not an investigation.

Sustainability disclosures, their SASB code, the metric and how we would instrument it
Standard	KPI	Metric	How we instrument it
`TC-TL-130a.1`	Operational energy	total energy consumed (GJ), % grid, % renewable	Network and data-centre telemetry is already collected; we join it to site and tariff data so renewable share is reported, not estimated.
`TC-TL-130a.1`	Emissions intensity	scope 1 and 2 (tCO2e), per subscriber and per terabyte	Usage-based intensity is the number that survives a traffic-growth argument — and it falls out of the same pipeline.
`TC-TL-440a.1`	Product end-of-life	% of equipment recovered, by weight	Reverse-logistics events captured at the returns flow, reconciled against procurement volumes.
`TC-TL-230a.1`	Data security	breach count, % involving PII, subscribers affected	Subscriber data is the crown jewel here. Field-level PII tagging and access logging make the disclosure a report, not an investigation.

Where we help

OSS/BSS without the big bang: An event layer over the existing stack, so order fallout and billing disputes are caught by a queue rather than by a customer. Strangler-pattern migration where the incumbent system has to keep running.
Care deflection that survives real traffic: Contact centre AI over your own knowledge and line diagnostics, with a hard handoff to a human and a transcript that an agent can actually use.
A marketplace the channel will sell: For CSPs moving into SMB services: catalogue, bundling, entitlement and billing exposed as APIs, with a partner onboarding flow measured in minutes rather than quarters.

Working on something in Telecommunications?

A 30-minute call is usually enough to work out whether this is a quick win or a rebuild — and we will tell you if it is the second one before you spend anything.

Book A Call

Calculator

What is a contact your AI could have answered worth?

Care cost scales with subscriber count — until deflection works. Model the contacts and the cost you would take out.

Indicative

Telecommunications

Annual run-rate saving

$36M

Care contacts per year: 24M
Contacts deflected / year: 6M
Equivalent agent FTE: 375

Subscribers 3 M

0.1 M 50 M

Care contacts per subscriber / year 8

1 24

Fully loaded cost per contact 6 $

Target deflection on eligible contacts 25%

5% 60%

Model assumptions

Average handle time assumed at 6 minutes; a 1,600-hour productive year.
Deflection applies only to contacts eligible for automation — billing, activation status, plan changes, diagnostics.
Excludes platform cost; the detailed model nets it out.

An order-of-magnitude model, not a quote. We will send the version with your real baselines, the assumptions replaced, and the implementation cost netted out.

Send me the full model

I agree to the Terms and Privacy policy.