A lending institution with 50 collection agencies cannot monitor them all. A human collections head can stay close to 5, keep reasonable visibility over 10, and is essentially blind to the remaining 40 — until a complaint arrives, a fraud surfaces, or an RBI inspection reveals conduct that has been occurring undetected for months. The Collections Head AI monitors all 50 simultaneously, continuously, and with the same rigour it applies to the first.
The Governance Gap That Exists in Every Collections Function
Collection agency governance is structurally difficult because the principal-agent problem is embedded in the model. An agency is contracted to recover money. Its agents are incentivised to recover money. The practices they use to recover money — the calls they make, the language they employ, the pressure they apply — are invisible to the lender unless the lender has the infrastructure to make them visible. Most lenders do not.
What exists in most institutions is a quarterly review meeting with the agency, a complaint-driven monitoring process that only surfaces the most egregious conduct, and an annual performance review that looks at recovery rates without looking at how those rates were achieved. This governance model is inadequate — not because the agencies are necessarily acting improperly, but because it creates no systemic incentive for proper conduct and no early warning when conduct deteriorates.
The Collections Head AI creates that systemic incentive by making agency performance and conduct visible — in real time, across every dimension that matters — and by publishing a ranked agency leaderboard that every agency can see. Agencies that know they are continuously monitored on both recovery rate and conduct metrics behave differently from agencies that know they are reviewed quarterly.
The 8-Dimension Agency Performance Score
The Collections Head AI scores every agency across 8 dimensions — 4 operational and 4 conduct-based — and combines them into a composite agency health score updated weekly. The score is transparent: every agency receives their own scorecard with commentary on what is driving movements in their score.
Resolution Rate by Bucket
Percentage of assigned accounts resolved (payment received or restructuring agreed) per bucket per month. Benchmarked against peer agencies assigned similar borrower risk profiles to control for assignment bias.
Contact Rate & Right-Party Contact
Percentage of assigned accounts reached and percentage of contacts that are right-party (the borrower, not a family member or neighbour). Low right-party contact rate may indicate list quality issues or agent shortcuts.
Promise-to-Pay Conversion
Percentage of borrower commitments that result in actual payment. Low conversion indicates agents are eliciting commitments without genuine borrower intention — a leading indicator of future roll-back and false reporting.
Field Visit Outcome Rate
For agencies handling Bucket 2 and 3 field work: percentage of field visits resulting in documented outcome (payment, commitment, or address verification). Unproductive visits with no documentation are flagged.
FPC Compliance Rate
Percentage of monitored calls that are fully compliant with RBI Fair Practices Code — no prohibited language, correct identification, proper time-of-call compliance, no third-party disclosure. Drawn from AI call monitoring (Article 3).
Borrower Complaint Rate
Formal complaints received per 1,000 accounts assigned. Tracked against agency-level baseline and peer benchmark. Rising complaint rate from a specific agency is the strongest early conduct warning signal.
Data Security & Documentation
Agent-level compliance with data handling protocols: borrower data not shared with unauthorized parties, field visit notes submitted within SLA, call recordings retained per policy. Assessed through audit sample and system logs.
Agent Attrition & Training Currency
High agent attrition within an agency is a conduct risk indicator — it prevents consistent FPC training and creates onboarding gaps. AI tracks agent roster changes and flags agencies where trained-agent percentage falls below threshold.
The Real-Time Agency Leaderboard
| Agency | Rank | Resolution Rate | FPC Compliance | Complaint Rate | Composite Score | Trend | Status |
|---|---|---|---|---|---|---|---|
| Resolve Associates | #1 | 84.2% | 96.1% | 0.4 / 1,000 | 91/100 | ↑ +3 | Preferred |
| NorthStar Collections | #2 | 81.7% | 93.4% | 0.6 / 1,000 | 87/100 | ↔ 0 | Preferred |
| Apex Recovery Services | #11 | 78.3% | 81.2% | 1.4 / 1,000 | 72/100 | ↓ −4 | Watch |
| FastTrack Recoveries | #38 | 82.1% | 64.3% | 4.1 / 1,000 | 51/100 | ↓ −11 | Flagged |
| Swift Collections Ltd | #44 | 71.8% | 58.7% | 6.2 / 1,000 | 44/100 | ↓ −16 | Escalated |
What Happens When an Agency Is Flagged
A flagged agency — one whose composite score has fallen below the threshold or whose FPC compliance rate has triggered the conduct alert — does not simply receive a lower score. The Collections Head AI initiates a structured governance response: assignment suspension for the affected account categories, a cause analysis brief delivered to the Collections Head within 24 hours, an agency remediation notice specifying what must change and by when, and a monitoring intensification protocol that increases the AI's call sampling rate for that agency from 10% to 100% until compliance is demonstrated.
For FastTrack Recoveries in the leaderboard above — whose FPC compliance rate has fallen to 64.3% with a complaint rate more than 10x the top-ranked agency — the Collections Head AI has already suspended their Bucket 2 assignment, flagged 14 specific call recordings for immediate human review, initiated a show-cause notice draft for the Collections Head to issue, and calculated the portfolio reallocation required to maintain overall recovery targets while the agency is under review. The human Collections Head reviews and approves; the AI has done everything except sign the notice.
Visibility Is the Only Effective Agency Governance Tool
You cannot govern what you cannot see. A quarterly review and a complaint-driven monitoring process creates a governance gap that agencies — consciously or not — tend to fill with conduct the institution would not endorse. The Collections Head AI closes that gap by making every agency's performance and conduct visible in real time. When agencies know they are always being measured, they behave as if they are always being watched. That is the governance effect.