Collections Training Agent AI

Invoked when: daily drill cycle runs each morning — every active collections agent receives their daily FPC scenario before their shift begins

• –Delivers a daily Fair Practice Code compliance simulation to every collections agent — a scenario-based drill in which the agent responds to a borrower situation that tests their understanding of the RBI's collections conduct requirements. Each drill scenario covers one of the seven FPC conduct areas: permitted calling hours (no calls before 8 AM or after 7 PM), prohibited language (no abusive, threatening, or humiliating language), contact frequency limits (maximum contacts per day and per week as specified in the RBI guidelines), family and employer contact restrictions (the conditions under which a borrower's family or employer may be contacted), consent to location visit (a field visit requires prior notice and the borrower's consent), settlement communication (verbal settlement commitments must be followed by written confirmation), and dispute acknowledgment (a borrower who raises a dispute must be told the resolution timeline immediately). Each daily drill focuses on one area in a rotating schedule, so each area is covered at least once per week.
• –Scores each agent's drill response against the FPC rubric — marking the response as compliant, partially compliant (the correct action was taken but the language or framing was suboptimal), or non-compliant (the agent's response describes an action that would violate the FPC). Drill scores are recorded per agent per day and aggregated into the weekly FPC adherence metric — the percentage of daily drills in which the agent demonstrated compliant conduct. A weekly FPC adherence rate below 85% is a training risk signal; below 70% is a compliance risk signal that is escalated to the collections head.
• –Updates drill scenarios within 5 business days when a new RBI circular modifies the Fair Practice Code or introduces new collections conduct requirements. The updated scenarios are reviewed by the CCO before they are activated — ensuring that the drills accurately reflect the new regulatory position and do not inadvertently train agents to a superseded standard. Agents receive a brief explanation of the regulatory change alongside the updated drill scenario so they understand why the correct response has changed.

Invoked when: the QA scoring process flags a call for script coaching — coaching feedback delivered to the agent within 48 hours

• –Reviews every flagged call transcript for specific script adherence and language quality issues — going beyond the pass/fail FPC scoring to identify exactly where in the call the agent's language deviated from the approved script or from the communication quality standard. The coaching feedback identifies the specific exchange (the borrower's statement and the agent's response), the specific issue with the agent's response (language that was too aggressive, a commitment made without authority, a failure to acknowledge the borrower's dispute, or a missed de-escalation opportunity), and the specific alternative phrasing that would have been compliant and effective.
• –Frames coaching feedback constructively — presenting the alternative phrasing as the better approach rather than dwelling on the failure. A collections agent who receives feedback framed as "here is what works better" retains and applies it more effectively than one who receives feedback framed as "here is what you did wrong." The coaching feedback includes a model script for the specific scenario type — a borrower who is abusive, a borrower who claims to have already paid, a borrower who invokes their right to a payment plan — so that the agent has a ready reference for the next similar call.
• –Tracks the agent's coaching feedback history — noting whether the issues identified in prior feedback sessions are recurring or improving. An agent who receives the same coaching feedback on the same script issue across three consecutive weeks is not applying the coaching; this pattern is escalated to the collections head with the agent's feedback history attached. An agent whose flagged issues are different each week may be improving on prior issues while encountering new challenges — a different profile that requires a different coaching response.

Invoked when: call transcripts are available from the call recording system — all calls QA-scored within 24 hours of call completion

• –Scores every collections call transcript against the QA rubric — a 10-point assessment across five dimensions: FPC compliance (calls within permitted hours, no prohibited language, correct contact frequency — 3 points), script adherence (opening and closing script followed, correct product identification, required disclosures made — 2 points), de-escalation quality (borrower distress recognised and addressed, tone modulated appropriately, call not prematurely terminated — 2 points), commitment accuracy (payment commitments made within the agent's authority, settlement terms not improvised, correct documentation promised — 2 points), and dispute handling (dispute acknowledged, resolution timeline communicated, dispute registered correctly — 1 point). A call that scores below 6 out of 10 is flagged for coaching; a call where the FPC compliance dimension scores below 2 out of 3 is flagged as a potential FPC violation for supervisor review.
• –Produces a weekly QA scorecard for the collections head — ranking all agents by their weekly average QA score, identifying the agents whose scores have improved, declined, or remained static relative to the prior week, and highlighting the specific QA dimension where team-level performance is weakest. The weekly scorecard is the collections head's primary tool for performance review; the QA dimension breakdown tells them where to focus team coaching, not just which agents to watch.
• –Cross-references QA scores with the Dispute Manager Agent AI's dispute log — checking whether agents with confirmed conduct disputes also have a pattern of low QA scores in the FPC compliance dimension. A dispute that is supported by a pattern of low QA scores on the same conduct issue is a more serious compliance concern than an isolated incident; the cross-reference gives the collections head and the CCO the full picture when assessing the severity of a conduct complaint.