Fixed, mobile, and converged operators

AI worker teams for service assurance and retention

Telco churn is largely manufactured in the support queue. The information needed to fix most faults exists before the customer calls — the failure is that nobody checks the network before asking for a reboot.

FIXED, MOBILE, AND CONVERGED OPERATORSCONSTRAINT ENVELOPE4 sector constraints designed to firstHUMAN AUTHORITY — NEVER AUTOMATED· Complex and infrastructure fa…· Save offers and commercial co…· Disputes and formal complaintsACTS WITH APPROVALThresholds and gates on anything with cost or risk attachedACTS WITHIN POLICY3 blueprints start hereAutonomy widens outward only on evidence from the recordearned autonomy
The authority boundary we design to in Telecommunications — autonomy widens outward from the centre, and only on evidence.

Our view

Where we think the real opportunity is

The defining telco support failure is not rudeness or wait time, it is being asked to troubleshoot a fault the network already knows about. That happens because the agent cannot see network state, provisioning, device profile, and billing in one place inside the seconds a conversation allows — so they fall back to a script. A worker that queries all four before speaking removes the single most corrosive experience in the sector. The second insight is about churn: cancellation intent usually surfaces inside a technical call, and the standard response is a retention script delivered by whoever happens to be on the line. We think that is backwards. Save conversations are commercial judgements, and the automation's job is to detect the signal and route it fast with full fault history, not to survive the call.

Operating reality

How demand actually behaves here

Volume has a shape, and the shape is what breaks teams. This is what we assume about your operation before we design anything.

  • Diagnosis spans four systems minimum

    Network monitoring, provisioning, device profile, and billing each hold part of the answer, and no agent can hold all four contexts at conversational speed.

  • Repeat faults are logged as new tickets

    Because contacts attach to tickets rather than to the line, the pattern that actually predicts churn is invisible until the cancellation call arrives.

  • Churn intent arrives inside technical conversations

    The cancellation is rarely the reason for the call. It is the third unresolved fault, which means the churn signal is a support-queue artefact.

  • Contact rate per subscriber is high and elastic

    Small service degradations generate large contact volumes, so quality of resolution has a direct and rapid effect on cost to serve.

Where the hours go

The cost we would try to move first

Almost every operation spends its most expensive time on its least valuable work. Naming that precisely is what makes a pilot measurable rather than impressive.

  • Handling time inflated by system-hopping

    A meaningful share of every technical call is the agent navigating interfaces while the customer waits, and it is time that produces no diagnostic value.

  • Truck rolls that were not necessary

    Field visits dispatched because remote diagnosis was inconclusive are the most expensive outcome of a weak first conversation.

  • Repeat contacts on unresolved faults

    The same fault handled three times costs three times, and the third contact is the one that also costs the subscriber.

  • Churn attributable to service experience

    Acquisition spend replacing subscribers lost to fault-handling failures is the largest and least attributed cost in this list.

The trust envelope

What constrains the design, and what never gets automated

We design to the boundary first. A control added after launch is a control nobody believes — and in this sector the boundary is not negotiable.

Sector constraints

  • Verified access before service or billing detail

    Account access levels determine what may be discussed, and the level has to be established before the conversation, not assumed from the calling number.

  • Explicit confirmation before any billable change

    Plan, tariff, and add-on changes need recorded customer confirmation, because a disputed charge originating in an automated conversation is a regulatory problem.

  • Approved diagnostics only

    Workers may run a pre-approved test set. Ad-hoc network commands are outside scope, both for safety and because their results need expert interpretation.

  • Regulated complaint and switching processes

    Complaints and porting carry defined processes and clocks that automation must hand over to rather than participate in.

Stays human, permanently

These are structural boundaries, not trust levels waiting to be relaxed.

  • Complex and infrastructure faults

    Multi-line, multi-site, and network-side faults need an engineer, and the worker's contribution is a complete brief rather than an attempt.

  • Save offers and commercial concessions

    Discounts, credits, and contract variations are commercial decisions with margin consequences, made by a person.

  • Disputes and formal complaints

    Once a customer disputes a charge or escalates formally, the regulated process is human-owned from that point.

Where to start

Sequencing matters more than scope

The first blueprint should be the one whose success criteria your team already agrees on. Each step below links to the blueprint that implements it.

  1. Start

    Account verification and known-incident checks, read-only

    Verify the account, check known incidents, and triage. Nothing is changed and no diagnostics run yet — and this alone eliminates the reboot-during-a-known-outage conversation.

    Read the blueprint
  2. Then

    Approved diagnostics and the safest remediation class

    Enable the approved test set and the narrowest class of configuration fixes, with approval gates on anything touching provisioning or charges.

    Read the blueprint
  3. Next

    Churn routing, then arrears

    Turn on repeat-fault and cancellation-intent routing with full history, then reuse the consent and identity governance for arrears recovery.

    Read the blueprint

Systems we would expect to read and write

  • Network monitoring and incident feed
  • OSS/BSS provisioning
  • Billing and charging
  • CRM and ticketing
  • Device and CPE management

Blueprints that apply

3 operating blueprints for Telecommunications

Each one carries the full team, process, and control detail — the mechanics this page deliberately does not repeat.

Push back on this

Objections we actually hear in Telecommunications

Answered as we would answer them in the room, including the cases where the right answer is that this is not for you.

Our fault space is far too complex for automation.
The whole fault space, yes. That is why scope is defined by fault class rather than by queue: known-incident checks and the two or three highest-volume fault classes, with everything else routed to a specialist with a complete brief. Measuring first-contact resolution per fault class also tells you honestly where the blueprint helps and where it does not, which is a better outcome than a blanket claim either way.
Automating retention calls would damage our churn numbers.
We agree, which is why the blueprint does not automate them. Cancellation intent routes to a human retention specialist with the full fault and contact history attached. What is automated is detecting the signal early — including the repeat-fault pattern that predicts it — so the save conversation happens with context and before the customer has decided.
Our OSS/BSS estate is old and poorly documented.
That is the real project, and it is worth being honest that integration effort in this sector is usually larger than the conversational build. The mitigation is to start read-only against the systems that already expose an interface — typically network monitoring and CRM — and to prove value before touching provisioning. If nothing is integrable, this will not work, and that is worth establishing in week one rather than month three.

Other sectors

How the argument changes elsewhere

Next step

Argue with this in a working session.

Bring your own numbers. We will map your version of the operating reality, agree what must stay human, and scope the first blueprint against a measurable win — or tell you it is not worth doing yet.