Nestack Agent Care
Industries / Telecom / Settlement agent

Telecom AI agent · Wholesale settlement

Wholesale Settlement & Dispute AI Agent

Reconcile a partner's records against the agreement you signed, assemble the evidence behind each discrepancy, and surface every configured dispute window with the time left — a named settlement manager decides what is filed.

4–6 weeksTypical delivery
Your stackDeployment
Window-trackedManager approval
Agent CareAfter launch

What this agent does

Reconciles the claim, does not file it

In
01

Ingesting partner statements, usage records and rate schedules from supported settlement, mediation or file sources.

02

Normalising record identity across the two sides, and carrying each line forward with the file it came from.

Reason
03

Matching the partner's records against your own for the same period, route and trunk group.

04

Applying the rates, effective dates and settlement terms configured from that one agreement.

05

Calculating what the agreement says is owed, and showing the arithmetic behind each line.

Decide
06

Assembling the evidence pack for a discrepancy, with the records and versions it rests on.

07

Surfacing each dispute, notice and audit window with the time left before it closes.

Out
08

Retaining the reconciliation, the evidence, the manager's decision and the outcome against the settlement.

09

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

Product statement

The agent reconciles and evidences; a named settlement manager decides what is filed, withdrawn or settled, and the carrier stays the party to the agreement.

Example workflow

One settlement period, end to end

AgentHuman
1Partner statement receivedSettlement statement, usage records, or a wholesale-billing extract
2Records matchedUsage detail, route and trunk group, rate schedule and agreement version, each with its source
3Discrepancy proposedAmount, evidence, window and confidence
4Controls appliedAgreement-version checks, bilateral data scope, window arithmetic and confidence threshold
No human action required

Stages 1 to 4 run unaided, and nothing is filed at any of them — the agent is reconciling, and the manager's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the settlement manager to decide.

Low confidence

Adds a wholesale analyst read first.

Manager approval

The discrepancy is held with its evidence, its window and the confidence.

Approve · Amend · Send to analyst review
Approved — released to file
6Settlement systems updatedOnly where write access and approval policy allow it
7Outcome evaluatedDispute outcome, evidence sufficiency, window margin and amounts recovered
Amendments

Every amendment a manager makes is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items
Filing, withdrawing or settling a dispute claim.
Agreeing a rate or a commercial term with a partner.
Characterising traffic, jurisdiction or usage factors.
Withholding or suspending payment of a billed amount.
Automation boundaryAgent acts unaided
Reconcile the partner's records against the agreement on file.
Calculate what that agreement says is owed, line by line.
Assemble the evidence pack for a discrepancy, with its sources.
Surface every configured window with the time left on it.
Any write happens inside the boundaries agreed at implementation, never ahead of approval.
Writing off, waiving or crediting an open claim.
Declaring a partner an access-stimulating carrier.
Using data from outside your own agreements and records.
Changes to agreement, rate or window configuration.

Example output

One disputed line, annotated

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

Reconciliation output · single lineIllustrative example
Partner
Disputed line
Amount claimed
Agreement version
Confidence
Window status
Regional interconnect
Tandem switching billed on transit minutes
$18,400.00
ICA amendment 7
91%
Open — flagged before expiry
As receivedTaken from both carriers' records and the agreement on file — nothing on this side is inferred by the agent.
Records used Partner statement line Your own usage records Rate schedule in force
Why this is disputedThe agreement's rate for that trunk group is not the rate billed.
ActionApproveAmendSend to analyst review
What the score decidesBelow the configured threshold the discrepancy picks up an analyst read before it.

Value

Where AI adds value

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

Where the value landsValue 01 – 04
Every billed lineFrom both carriers' records
03Reconciliation

Match against the agreement

Draw on your own records and the rates, terms and windows configured from that one agreement — never another partner's.

01Approved path

Reconcile, and watch the window

Routine matching across statements and usage records arrives already done.

02Human review

A missed window cannot be reopened

Once a dispute window closes the claim is extinguished, not reduced — so the time left is put in front of the manager.

04Build an evidence trail

The dispute, the records it rests on and the manager who filed it stay on the settlement.

Integrations

Typical integrations

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

Settlement and wholesale billingAmdocs · CSG
Netcracker · wholesale billing APIs
Mediation and recordsUsage records · file drops
Mediation platforms · exchange feeds
AgreementsInterconnection · roaming
Rate schedules · amendments

Agent

Wholesale settlement and dispute

Reads both sides
Reconciles the period
Holds for approval

WorkflowEmail · ticketing
Dispute and approval systems
Observability & evaluationOpenTelemetry · 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 dispute

Each control encloses the last. Whatever slips past all of them is named below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model
L6Rollback / safe modePull the agent back to reconciliation reporting when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, rate-table and agreement-configuration changes.Track
L4TraceabilityRecord the records used, the calculation, the evidence and the filing decision.Record
L3Manager approvalHold discrepancies for the named manager; it governs filing, not whether the claim is right.Gate
L2Policy guardrailsTest each output against the bilateral data scope and the agreement's own terms; a failure returns the line.Restrict
L1Confidence thresholdsRoute low-confidence discrepancies to an analyst read before the manager sees them.Require review
Model coreDiscrepancy proposed — amount, evidence, window and confidence
L1 – L2Test whether a discrepancy may stand
L3Puts the filing in a manager's hands
L4 – L5Keep the dispute and the records behind it
L6Cuts back to reconciliation reporting when signals degrade

How Nestack evaluates it

Evaluate the whole settlement run — not only the closing figure.

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

Surface — the claim the partner sees
Depth of coverage ▼
E1Final-output evaluationDid the calculated amount follow the agreement on file?
E2Step-level evaluationDid the agent use the right agreement version, rate schedule and period?
E3Tool evaluationDid it read and write the correct settlement account and record?
E4Confidence calibrationDo low-confidence discrepancies actually attract more manager amendments?
E5Slice evaluationHow does performance change across settlement regimes and partners?
E6Business outcomeHow many claims were amended after filing, or withdrawn by a manager?
Floor — what survives the dispute window

Failure modes

Where each failure originates in the agent

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

Agent lifecycleDirection of processing →
01 · Retrieval1 mode
UX-03

Superseded rate schedule

A rate read from an amendment no longer in force.

Stage gathersStatements, usage records, rate schedules and windows
02 · Reasoning2 modes
UX-04

Traffic mischaracterised

A record is treated as access when the agreement says otherwise.

UX-06

Cross-partner leakage

One partner's rate data reaches another partner's reconciliation.

Stage proposesAmount, evidence, window and confidence
03 · Tool / write2 modes
UX-02

Window mis-dated

A window is counted from the wrong trigger date.

UX-05

Duplicate record matching

Late-arriving records are counted twice in one period.

Stage writesOnly where write access and approval policy allow it
04 · Output1 mode
UX-01

Discrepancy overstated

A claimed amount exceeds what the agreement supports.

Stage returnsThe discrepancy the manager files and the partner receives
05 · Change / Version1 mode
UX-07

Silent rate-table regression

A model or table change shifts what the agent will claim.

Stage tracksModel, prompt, rate tables and agreement config
Sev-1 · acts outside the boundary Sev-2 · a wrong claim reaches the partner Sev-3 · records degrade, line routes to review

Affected slices

A healthy total can hide one settlement type

An aggregate amendment rate is one number covering every agreement at once; the slice view is where a single settlement type shows up carrying most of the rework. Nestack reports the manager-amendment rate by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example
SliceFailure rateLift Lift vs. thresholdStatus
International settlement7.2%3.7× Review
Roaming usage records4.5%2.3× Review
Transit and tandem billing2.7%1.4× Watch
Domestic reciprocal compensation1.8%0.9× Normal
Bar: amendment-rate lift vs. the domestic reciprocal-compensation baseline · 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 test, not a memo

The cycle does not end in a post-mortem. It ends in a case the next release has to clear, and that suite is what the next settlement run is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect

Manager-amendment rate rises in a settlement slice.

02Diagnose

Read the two record sets and the agreement behind them until the cause narrows to one.

03Improve

Each change carries a version and the disputes that forced it.

04Verify

Release waits on the affected cases going green again.

05Learn

The case is kept for good, and the dispute playbook is rewritten.

Learn → DetectThe return edge. The next run is detected against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

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

Workstream Week 1Week 2Week 3Week 4Week 5Week 6
01Settlement workflow discovery and boundary definition.
02Record and rate-source assessment.
03Agreement, rate-schedule and dispute-window mapping.
04Record ingestion and normalisation.
05Reconciliation and evidence logic.
06Confidence scoring and window routing.
07Manager approval workflow.
08Settlement-system integration.
09Dispute-window regression cases.
10Guardrails and filing controls.
11Settlement-trail instrumentation.
12Deployment, documentation and Agent Care handover.
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 tier PilotOne partner, one agreement ProductionProduction settlement systems AdvancedMultiple partners / regimes
Introduced at Pilot
Reconciliation to your agreement
Manager approval
Reconciliation-quality baseline
Introduced at Production
Reporting by partner and route
Dispute workflow in your systems
Approved write-back
Settlement-system integration
Introduced at Advanced
Multi-regime settlement rules
Multi-stage wholesale approvals
High settlement volume
Multi-agreement settlement controls
Build price From $5,000 From $8,000 Custom quote
Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from buildBuild 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.

You bringWe build with it
01Your interconnection and roaming agreements Agreement, rate-schedule and window mappingWeek 1
02Representative settlement periods Record ingestion, normalisation and the reconciliation baselineWeek 2
03Your dispute process and escalation ladder Dispute-process mapping and automation-boundary definitionWeek 1
04Access to relevant APIs, feeds or exports Record and rate-source assessment, then integration setupWeek 2
05Disputes you would not want filed Window cases and the evaluation suiteWeek 4
06What must reach a manager before a claim is filed Confidence scoring, window routing, guardrails and approval controlsWeek 3
07Named settlement managers to review claims Manager approval workflow, then pilot and production validationWeeks 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, which is why the fifth carries two kinds at once.

Phase W1W2W3W4W5W6
Discovery W1
Build W2 – W3
Evaluate W4 – W5
Pilot & Launch W5 – W6
Week focus W1Settlement workflow discovery, agreement mapping and the automation boundary W2Record and rate-source integration and the reconciliation baseline W3Reconciliation workflow, confidence logic and approval controls W4Evaluation suite, window cases and failure-mode testing W5Settlement-system integration, pilot partners and targeted corrections W6One settlement cycle reconciled under the wholesale desk, then handover
Reading the bandA bar covers the weeks its work is named in, and nothing else. The week 5 overlap is real, not padding.
At the end of W6The cycle closes the validation and Agent Care takes over monitoring.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Telecom AI agent

Build a settlement agent around your wholesale agreements.

A settlement agent is only as accurate as the agreements behind it. You bring the agreements, the rate schedules and the dispute windows; we map the reconciliation workflow and name what the manager keeps.

Nestack Agents · Wholesale settlement and disputeAGT-TL-16 · Agent Care available after launch