Can the assistant run on in-country infrastructure?

Yes. We size the model to the task, and a right-sized model that runs in-country usually beats a frontier model behind a residency problem.

How do you measure the assistant?

Containment, escalation quality and repeat contact, reported in the same view as the human queue.

Do you need access to raw CDRs?

Only where the use case requires it, and then aggregated and minimised at ingest with the basis recorded.

What does the first release cover?

One journey end to end, usually billing questions, because it is high volume and easy to measure.

Riyadh, telecom

Telecom AI in Riyadh lives or dies on subscriber data quality

Churn models, network analytics and Arabic service bots all read the same subscriber record. When that record disagrees with billing, every model built on it is wrong in the same direction.

Book the Data Foundation Assessment Everything we do in Riyadh

Fix the subscriber record, then the models get interesting.

CSTPDPLNCA ECCNDMO

Key takeaways

What you need to know, in a minute

One subscriber view

Network, billing and care records reconciled into a single view with the matching rules documented.

Arabic that handles dialect input

Customers speak in dialect and expect a formal reply. The assistant is built for that gap, not against it.

Consent respected in targeting

Marketing and retention audiences are built from consented attributes, and the basis is recorded per campaign.

Commercial

Retention scoring with reasons

A churn score paired with the reason and the offer that fits it, so the care agent has something to say.

Care

Arabic first-line support

Billing questions, plan changes and fault reports handled in Arabic, with a clean handover to a human when the case needs one.

Network

Network incident triage

Alarms grouped into incidents with a likely cause attached, so the field team is dispatched once.

Use cases

Where this applies in practice

What usually stalls these programmes

Billing and CRM disagree

Reconciliation is the first deliverable. Nothing useful sits on top of two conflicting truths.

Bots trained on English scripts

Translated scripts fail on the second turn of a real Arabic conversation. We start from Arabic transcripts.

No feedback loop

Outcomes are written back, so the next version is trained on what actually happened.

Telecom in Riyadh, questions we get

Other sectors in Riyadh

Government and public sector

Bad data, bad AI. We fix the data layer under the service first.

Open page

Banking and finance

A model without lineage is an opinion. We build the lineage first.

Open page

Healthcare

One patient, one record, one version of the truth.

Open page

Retail and commerce

Catalogue first, customer second, forecast third.

Open page

Energy and utilities

An asset register that matches reality is the whole project.

Open page

Jeddah

Clean the intake and the rest of the automation stops fighting you.

Let's put your AI ambition on the table

Sixty minutes with a senior Bilytica partner. We stress-test your plans against what actually works in your sector, agree which moves are worth funding first, and you leave with a written 90-day plan and honest numbers, whether you work with us or not.

Start the conversation