Skip to content
AtheronLABS

You're visiting from the United States. Prices are shown in US dollars. Not right?

labs@atheron:~/industries/telecom/ai$ train agent --grounded

AI for telecom

Care agents and alarm correlation, each on its own route.

Carriers have two queues that AI can shorten: customers waiting to ask simple questions, and alarms waiting for someone to find the root cause. The first suits a frontier model working from your public help content. The second needs a model trained on your own network.

Start a projectSee the example projects

// what matters here

What is different in this industry.

01

The right route for each job

A care agent on a frontier model starts fast and handles high volume. An alarm model is trained on your own history, because no general model knows your network.

02

Account data through your tools

The care agent looks up an account only after the customer signs in, through interfaces you control, and never sees card numbers.

03

A person for the hard cases

The agent hands over with the conversation and what it tried, so the customer does not repeat themselves.

04

Scored on past incidents

The alarm model is measured on incidents your team has already solved, and it goes live only when it groups them the way they did.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAIFrontier model

Customer care agent

Hypothetical. Not a client, not a result.

// The problem

A carrier's care team answers the same questions about bills, outages and moving service all day, and customers wait while the hard cases sit in the same queue.

// What we would build

A care agent in the website chat and app that answers from the carrier's own help content, checks outages and explains a bill after the customer signs in, and hands to a person with the conversation attached. It uses a frontier model through its API: the fastest start for a high-volume product where the content is public and account lookups go through tools you control.

// What is in it

  • Answers grounded in your help centre and plan terms, with links
  • Outage status by address from your network systems
  • Bill explanations after the customer signs in, with no card data in the conversation
  • Handoff to an agent with the full conversation and what was tried
  • English and French, with an evaluation set run before every change

// Stack

  • Frontier model API
  • Next.js
  • PostgreSQL with pgvector
  • CRM and billing interfaces
  • AWS

// estimate

Build
≈ US$114,000 to US$175,000, delivered within 22 weeksCAD 163,000 to 249,300
Hosting
≈ US$3,660 a monthCAD 5,205 a month
Support
≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Frontier model
Model running cost
≈ US$3,480 a monthCAD 4,955 a month

Timeline by milestone

Discovery
2.9 to 3.4 weeks
Specification and evaluation plan
0.4 weeks
Design approved
2.4 to 3.9 weeks
Core features
1.4 to 1.9 weeks
Full build
0.9 to 1.4 weeks
Working pilot
3.9 to 6.4 weeks
Testing and fixes
0.9 to 1.9 weeks
Launch
0.9 to 1.4 weeks
Production
0.9 to 1.4 weeks

Outside our hands, and added to the calendar

Test access from your IT team
1 to 4 weeks
Penetration test by your chosen firm
2 to 4 weeks

How it is paid

Deposit 20%
CAD 32,600 to 49,860
Discovery 2%
CAD 3,260 to 4,986
Specification and evaluation plan 10.2%
CAD 16,626.00 to 25,428.60
Design approved 2%
CAD 3,260 to 4,986
Core features 6.2%
CAD 10,106.00 to 15,456.60
Full build 4.1%
CAD 6,683.00 to 10,221.30
Working pilot 30.7%
CAD 50,041.00 to 76,535.10
Testing and fixes 2%
CAD 3,260 to 4,986
Launch 2%
CAD 3,260 to 4,986
Production 10.8%
CAD 17,604.00 to 26,924.40
Holdback, 30 days after launch (10%)
CAD 16,300 to 24,930
Start from this example
Example projectWeb appAICustom LLM

Network alarm correlation model

Hypothetical. Not a client, not a result.

// The problem

A single fault produces hundreds of alarms across the network, and the operations centre spends the first hour working out the root cause by hand.

// What we would build

A custom model trained on the carrier's own alarm history and outage tickets that groups related alarms, proposes the likely root cause and the affected customers, and shows the operations team why, served in the carrier's own cloud account.

// What is in it

  • Alarms streamed from your network management systems
  • Related alarms grouped into one incident, with the likely root cause first
  • Affected services and customers listed for the care team
  • A model trained on your own alarm and ticket history, scored on past incidents
  • Operators confirm or split groups, and their corrections feed the next training round

// Stack

  • PyTorch
  • Kafka
  • Python
  • Next.js
  • AWS

// estimate

Build
≈ US$166,000 to US$253,000, delivered within 36 weeksCAD 235,800 to 360,000
Hosting
≈ US$7,020 a monthCAD 10,000 a month
Support
≈ US$5,010 a monthCAD 7,135 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Custom LLM
Model running cost
≈ US$6,530 a monthCAD 9,300 a month

Timeline by milestone

Discovery
2.9 to 3.4 weeks
Data and evaluation plan
0.4 weeks
Design approved
2.4 to 4.4 weeks
Core features
1.9 to 3.4 weeks
Full build
1.4 to 2.4 weeks
Training
8.9 to 15.4 weeks
Testing and fixes
0.4 weeks
Evaluation and red-teaming
3.9 to 4.4 weeks
Launch
0.9 to 1.4 weeks
Production
0.9 to 1.4 weeks

Outside our hands, and added to the calendar

Test access from your IT team
1 to 4 weeks
Penetration test by your chosen firm
2 to 4 weeks

How it is paid

Deposit 20%
CAD 46,920 to 71,760
Discovery 1.3%
CAD 3,049.80 to 4,664.40
Data and evaluation plan 8.2%
CAD 19,237.20 to 29,421.60
Design approved 1.3%
CAD 3,049.80 to 4,664.40
Core features 4%
CAD 9,384 to 14,352
Full build 2.6%
CAD 6,099.60 to 9,328.80
Training 24.8%
CAD 58,180.80 to 88,982.40
Testing and fixes 1.3%
CAD 3,049.80 to 4,664.40
Evaluation and red-teaming 16.5%
CAD 38,709 to 59,202
Launch 1.3%
CAD 3,049.80 to 4,664.40
Production 8.7%
CAD 20,410.20 to 31,215.60
Holdback, 30 days after launch (10%)
CAD 23,460 to 35,880
GPU time for training and testing, at cost
CAD 1,200. Billed up front, at cost, outside the milestones
Start from this example

// questions

Questions we get about this.

Does customer data go to the model provider?

The conversation is sent, so we keep personal details out of it where we can and look up accounts through your own tools. Where that is not enough, an open-source model in your cloud is the alternative, and we price it.

What does the alarm model learn from?

Your alarm history and the outage tickets that explain it. The more incidents with a known root cause, the better it groups new ones.

Can the care agent handle French as well as English?

Yes. It answers in the customer's language, and the evaluation set covers both.

Back to the industry: TelecomAbout the service: AI agents and models

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

Start a projectBook a call
AI for telecom: customer care agents and alarm models | Atheron Network Labs