← Agent catalogue

AI Agent Profile · LendingIQ · Bengaluru

Collections Agency Monitor AI

Function: Agency Performance ManagerInvoked via: agency operations pipeline — weekly SLA batch + event triggerRuntime: AWS Bedrock · ap-south-1Model: Claude Sonnet 4Context window: 200K tokens

DivisionCollections

Resume

What this agent does

The Collections Agency Monitor AI tracks the performance and compliance of LendingIQ's empanelled external recovery agencies — monitoring SLA adherence, surfacing conduct signals from complaint data, reconciling agency-collected payments against the NPA register, and flagging agencies whose performance or conduct warrants a formal review. It is the continuous oversight layer that sits between agency engagement and the human collections manager's accountability.

Primary functions

SLA Tracking

Weekly — all active agency allocations

INVOKED WHEN: weekly SLA review cycle is due or an agency reports a contact or resolution event

  • Reads each agency's allocated account list and reported activity — contact attempts, contacts achieved, PTPs captured, payments collected, accounts resolved — against contractual SLAs: contact rate within defined days, minimum PTP capture rate, and resolution rate by DPD band. Produces a scorecard per agency showing actual vs contracted performance on each metric.
  • Benchmarks each agency's performance against the portfolio average across all empanelled agencies — so the collections manager sees not just whether an agency meets its SLA, but whether it is above or below the peer group.
  • Flags SLA breaches immediately when detected — an agency that has not contacted 60% of its allocated accounts within the required window has breached its SLA and the flag goes to the human collections manager before the breach compounds.
Output: Weekly SLA scorecard per agency — each metric with actual vs contracted vs peer benchmark. SLA breach flags with breach magnitude and days since allocation.

Code-of-Conduct Audit

Continuous — complaint monitoring + monthly sampling

INVOKED WHEN: a conduct-related complaint is attributed to an agency, or the monthly review is due

  • Reads conduct-related complaints from the Grievance Redressal Agent AI attributed to external recovery agencies — harassment, contact outside permitted hours, misrepresentation, or abusive language — and maps them to the specific agency. Tracks complaint volume per agency over rolling 30-day and 90-day windows.
  • Computes the complaint rate (complaints per 100 accounts allocated) to normalise across different agency sizes — a complaint rate, not just a complaint count, is the meaningful metric for comparing agencies of different sizes.
  • Does not independently assess whether a complaint constitutes a conduct breach — that determination requires the human collections manager's investigation. The agent identifies the pattern and routes; the human investigates.
Output: Conduct monitoring report — complaint volume and rate per agency, 30-day and 90-day trend, peer comparison. Immediate flag when complaint rate exceeds the configured threshold.

Payment Reconciliation

On payment event + Monthly full reconciliation

INVOKED WHEN: an agency reports a collection, or the monthly reconciliation cycle is due

  • Reconciles agency-reported collections against CBS payment postings — confirming that every payment an agency claims to have collected has been posted to the correct account in CBS within the required timeline. Unreconciled collections are flagged immediately as a potential misappropriation risk.
  • Reconciles commission payable to each agency against its contractual commission structure — calculating commission on verified collections only. Produces the commission payable statement for finance team authorisation.
Output: Reconciliation report — agency collections vs CBS postings, unreconciled items flagged, commission payable statement for finance team. Immediate flag on any unreconciled collection above the amount threshold.

Hard guardrails

Will notInstruct recovery agencies on how to contact borrowers. Agency instruction is the human collections manager's responsibility.
Will notApprove or clear a conduct complaint without human investigation.
Will notRelease commission payments. Commission payable statements are produced for finance team authorisation.

Known limitations

Agency-reported activity data depends on agency self-reporting accuracy.An agency that under-reports contacts to avoid SLA scrutiny will produce misleading data. Supplement with quarterly borrower contact confirmations — a sample outbound call to borrowers allocated to each agency to confirm whether they received contact and when.
Conduct complaints are lagging indicators — serious incidents may not surface as complaints.Mystery call sampling — periodic test calls to the agency's contact scripts — is the proactive conduct control that detects problems before complaints surface.
Agent Profile · Collections Agency Monitor AI · LendingIQ · BengaluruLast updated April 2026 · For internal use

Important Reads

Learn more about how to deploy Collections Agency Monitor AI to your lending workflow.