Main site

Telecom AI agent · Retention offers

Retention-Offer Copilot AI Agent

Score churn risk from the signals you configure, select a save offer from your approved set, and hold the case for the named retention adviser, with the cancellation route open throughout.

Typical delivery
4–6 weeks
Deployment
Your stack
Adviser decides
Exit stays open
After launch
Agent Care

What this agent does

Prepares the save, not the decision to make it

In

  1. 01

    Ingest the account, its contract terms and the configured churn signals from supported billing or CRM sources.

  2. 02

    Normalise plan, tenure, arrears and contract-end fields, and carry each signal forward with the system it came from.

Reason

  1. 03

    Score retention risk against the thresholds configured for the segment, and show which signals moved it.

  2. 04

    Select a candidate offer from the approved offer set, never a price computed for the individual customer.

  3. 05

    Check the consent record for the channel, and stop where it is missing, revoked or built for a servicing call.

Decide

  1. 06

    Flag anything in the assembled case that would sit between the customer and cancelling.

  2. 07

    Route the case to the named retention adviser, who decides whether an offer is put to the customer at all.

Out

  1. 08

    Retain the signals, the candidate offer, the disclosures and the adviser's decision against the account.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent proposes an offer; the retention adviser decides whether it is made, and the carrier remains responsible for the charge.

Example workflow

One account, signal to decision

  1. 1Churn signal receivedBilling record, CRM note, contract-end date or an inbound cancellation request
  2. 2Case assembledTenure, plan, arrears, contract terms and the consent record, each with its named source
  3. 3Offer proposedOffer, term, price, disclosures, confidence
  4. 4Controls appliedApproved-offer-set checks, consent-tier checks, cancellation-path checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and no offer reaches a customer at any of them — the agent is assembling a case, and the adviser's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the retention adviser to approve.

Low confidence

Adds a compliance read first.

Adviser approval

The case is held with its signals, its candidate offer and the confidence.

Approve · Amend · Send to compliance review

Approved — released to the customer

  1. 6Billing systems updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedOffer acceptance, adviser amendments, cancellation completion and post-offer complaints

Amendments

Adviser amendments are counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Releasing a save offer or credit to a customer.
    • Placing anything at all in the cancellation path.
    • Contacting a customer whose consent record is missing.
    • Pricing an offer outside the approved offer set.

  • Automation boundary

    Agent acts unaided

    • ✓Score retention risk from the signals you have configured.
    • ✓Select a candidate offer from the offer set you approved.
    • ✓Assemble the case: signals, terms, consent and disclosures into the review queue.
    • ✓Hold the case for the named adviser, with the exit still open for the named owner.

    Any write happens inside the boundaries agreed at implementation, never ahead of the adviser's approval.

    • Re-contracting a customer or renewing a term.
    • Turning a servicing contact into an offer call.
    • Deciding what a cancellation is allowed to cost.
  • Changes to the offer set, thresholds or approval rules.

Example output

One save attempt, annotated

Everything the agent proposes is attached to the account it was drawn from.

Retention output · single account

Illustrative example
Account
Contract ending
Candidate offer
Bundle credit from the approved offer set, at the same monthly price
Stated term
12-month term
Source of record
Approved offer set
Confidence
91%
Consent tier
Prior express written

As received

Taken from the billing record and the consent register — nothing on this side is chosen by the agent.

Signals used

  • Contract-end date
  • Tenure band
  • Consent register entry

Why this offer

It comes from the approved set, not from a price computed for this customer.

What the score decides

Below the configured threshold the case picks up a compliance read before it.

Action

ApproveAmendSend to compliance review

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Accounts at churn risk

From the billing record

03Scoring

Score from configured signals

Draw on the contract terms, the tenure and the churn signals your retention team configured.

01Approved path

An offer is not an obstacle

Routine renewal cases arrive already assembled, with a candidate offer named.

02Human review

Send review to the risky cases

Consent gaps and low-confidence cases are marked, so the adviser's read starts where the exposure sits.

04

Build an evidence trail

The offer, the eligibility behind it and the adviser who released it stay on the account.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Billing and charging

    • Amdocs
    • Netcracker
    • CSG
    • Optiva
    • Ericsson BSS

  • CRM and care

    • Salesforce
    • Dynamics 365
    • ServiceNow
    • Zendesk

  • Churn and analytics

    • Snowflake
    • Databricks
    • Customer data platforms

Agent

Retention-offer copilot

  • Reads the account
  • Proposes an offer
  • Holds for approval

  • Contact and consent

    • Genesys
    • NICE
    • Twilio
    • Consent and preference stores

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the customer

The layers nest, and each says what it does not catch. The map below carries the rest.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modePull the agent back to scoring-only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, offer-set and threshold configuration changes.Track
L4TraceabilityRecord the signals, the candidate offer, the disclosures and the approval time.Record
L3Adviser approvalHold cases for the named adviser; it governs release, not whether the offer was the right one.Gate
L2Offer guardrailsTest cases against the approved offer set, the consent tier and the cancellation-path rule configured for that market; a failure returns the case.Restrict
L1Confidence thresholdsRoute low-confidence cases to a compliance read before the adviser sees them.Require review
Model coreCase assembled — offer, term, disclosures, consent tier and confidence
L1 – L2
Test whether a case may stand
L3
Puts the decision in an adviser's hands
L4 – L5
Keep the offer and the eligibility behind it
L6
Holds offers for the adviser when signals degrade

How Nestack evaluates it

Evaluate the whole save attempt — not only whether the customer stayed.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the offer the customer hears

  1. E1Final-output evaluationDid the candidate offer come from the approved set for that account?
  2. E2Step-level evaluationDid the agent use the right contract terms, thresholds and consent record?
  3. E3Tool evaluationDid it read and write the correct account and the correct field?
  4. E4Confidence calibrationDo low-confidence cases actually attract more adviser amendments?
  5. E5Slice evaluationHow does performance change across specific tenure bands?
  6. E6Business outcomeHow many cases needed an amendment, or a correction after the offer went out?

Floor — the outcome the carrier answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, placed at the stage each one originates.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • JD-03

      Stale contract state

      Term or consent status read from a superseded record.

    Stage gathers

    Account, contract terms, consent record and offer set

  2. 02 · Reasoning

    2 modes

    • JD-04

      Offer before the exit

      The case is built so the offer precedes the cancellation route.

    • JD-06

      Consent-tier mismatch

      A servicing consent record is used for an offer contact.

    Stage proposes

    Offer, term, disclosures and confidence

  3. 03 · Tool / write

    2 modes

    • JD-02

      Off-set price

      An offer is assembled outside the approved offer set.

    • JD-05

      Repeat save attempt

      The same account is worked twice inside the quiet window.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • JD-01

      Missing price disclosure

      A required pricing or renewal disclosure is not surfaced.

    Stage returns

    The offer the adviser releases to the customer

  5. 05 · Change / Version

    1 mode

    • JD-07

      Silent offer-set drift

      A model or rule change widens what the agent will offer.

    Stage tracks

    Model, prompt, offer set and threshold config

  • Sev-1 · an offer sent outside the set
  • Sev-2 · a wrong offer reaches the customer
  • Sev-3 · signals degrade, case routes to review
See the complete Telecom failure-mode catalogue

Affected slices

One cohort can carry most of the rework

In aggregate the amendment rate looks tolerable; in one slice it is not. A few cohorts absorb most of the adviser amendments, so Nestack reports by slice, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Inbound cancellation requests3.8%3.6×Review
Win-back after disconnection2.3%2.2×Review
Mid-contract renegotiation1.7%1.6×Watch
Standard renewal at term end0.8%0.8×Normal

Bar: adviser-amendment-rate lift vs. the portfolio average · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on a regression case

Nothing closes on a review meeting. It closes on a case the next release has to pass, and that suite is what the next save attempt is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Adviser-amendment rate rises in a tenure slice.

  2. 02

    Diagnose

    If the rise sits in one cohort, a retention lead reads the cases behind it until the cause narrows to one.

  3. 03

    Improve

    Every change ships against a version with the offers attached.

  4. 04

    Verify

    The release is blocked while an affected case fails.

  5. 05

    Learn

    The case becomes permanent, and the offer rules are re-tested.

  6. Learn → Detect

    The return edge. The next detection is measured against a suite longer than this one.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, offer workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Retention workflow discovery and boundary definition.Week 1
  2. 02Billing, CRM and consent source assessment.Week 1, Week 2
  3. 03Approved offer set, thresholds and disclosure rules.Week 1, Week 2
  4. 04Account ingestion and signal normalisation.Week 2
  5. 05Risk scoring and offer selection.Week 2, Week 3
  6. 06Confidence scoring and case routing.Week 3
  7. 07Adviser approval workflow.Week 3, Week 4
  8. 08Billing and CRM integration.Week 5
  9. 09Cancellation-path cases.Week 4, Week 5
  10. 10Guardrails and offer controls.Week 4
  11. 11Offer-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne market, one deskProductionProduction billing systemsAdvancedMultiple markets / brands
Introduced at Pilot
Scoring and offer selection✓✓✓
Adviser approval✓✓✓
Offer-quality baseline✓✓✓
Introduced at Production
Reporting by tenure band—✓✓
Approval workflow in your systems—✓✓
Approved write-back—✓✓
Billing-system integration—✓✓
Introduced at Advanced
Multi-jurisdiction offer rules——✓
Multi-stage retention approvals——✓
High save-attempt volume——✓
Multi-market offer controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your account fields and contract data

    Account ingestion and signal mapping

    Week 1

  2. 02Representative save attempts and their outcomes

    Scoring baseline, signal extraction and offer-set binding

    Week 2

  3. 03Your approved offer set and disclosure rules

    Offer-set, threshold and disclosure-rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Billing, CRM and consent assessment, then integration setup

    Week 2

  5. 05Offers you would not want made

    Cancellation cases and failure-mode testing

    Week 4

  6. 06Where a save attempt has to stop

    Confidence scoring, case routing, guardrails and approval controls

    Week 3

  7. 07Named retention advisers to review cases

    Adviser approval workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

The bands sit on the weeks the work occupies, so the fifth carries evaluation and launch together.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Retention workflow discovery, offer-set mapping and the automation boundary

W2Billing and CRM integration and the scoring baseline

W3Offer workflow, confidence logic and approval controls

W4Evaluation suite, cancellation-path checks and failure-mode testing

W5Consent-register integration, pilot cases and targeted corrections

W6One renewal cycle handled under the retention desk, then handover

Reading the band
A bar sits on the weeks its work is named in, and nothing else. The fifth week genuinely carries two kinds of work.
At the end of W6
The cycle closes validation and Agent Care picks up monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Telecom AI agent

Build a retention-offer copilot around your approved offer set.

Show us your offer set, your consent records and who signs an offer off. Which of those does a customer meet before the cancellation route? We'll map the workflow around that answer.

Typical delivery
4–6 weeks
Deployment
Your billing and CRM
Adviser approval
Exit stays open
After launch
Agent Care available

Nestack Agents · Retention-offer copilot

AGT-TL-11 · Agent Care available after launch

Build a retention-offer copilot

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your accounts

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.