The Fair Practices Code does not govern the institution's intentions — it governs what the agency's agents actually say to borrowers on calls that the institution is responsible for. An institution that contracts a collections agency and does not listen to what that agency's agents say on its behalf is an institution that has created a compliance exposure it cannot see. The Collections Agency Monitor AI samples 5% of every agency's daily call volume, scores each sampled call against the FPC conduct rubric, and flags any violation — not to the agency first, but to the institution's own compliance officer, independent of the agency's self-reporting.
What the FPC requires of collections agents — and the 6 violations most commonly found
The RBI's Fair Practices Code for NBFCs, reinforced in the Guidelines on Outsourcing and the Code of Conduct for Recovery Agents, specifies that collection agents must: identify themselves and the institution at the start of every call; confirm the borrower's identity before discussing the account; restrict calls to permitted hours (8 AM to 7 PM on weekdays and Saturdays, no calls on Sundays and national holidays); refrain from using abusive, threatening, or humiliating language; not contact family members, employers, or neighbours without the borrower's specific written consent; and not make representations about legal action that are not accurate.
The six violations most frequently found in agency call sampling are: calling outside permitted hours (particularly 7:01–7:59 PM, when many agents complete their evening's calls), failing to identify themselves at the start of the call, contacting a third party without consent, using language that crosses into threatening territory ("we will send police to your house," "we will tell your employer"), making inaccurate legal representations ("your account has already been handed to the court"), and recording outcome codes that contradict what the call recording shows.
The call audit: 9-criterion scoring rubric
Agent must state their name and "[Institution Name] Recovery Services" within the first 20 seconds
The borrower's right to know who is calling and on whose behalf is fundamental. Identification at call start is the first criterion — and one of the most frequently failed. Many agents skip identification when calls are busy or when borrowers answer with immediate hostility.
Discussing an account with a family member or third party without confirming identity and consent is an FPC violation
If someone other than the borrower answers, the agent may only confirm whether the borrower is available — they may not disclose the account, the reason for the call, or any financial information to the third party. Checking the recording for this criterion reveals whether agents are discussing loan details with whoever picks up the phone.
Call timestamp verified against permitted hours — recording metadata cross-checked against logged call time
The call timestamp is extracted from the recording metadata and cross-referenced against the permitted hours window. A call timestamped 19:07 is a timing violation regardless of whether the conversation was civil. Timing violations are scored binary — the call was either within permitted hours or it was not.
AI-scored against a vocabulary and tone rubric · Escalated to human review for ambiguous cases
Call audio is transcribed and the transcription is scored against a list of prohibited terms (direct insults, family-related threats, employer contact threats, references to police or legal action in threatening rather than factual terms). The AI flags any call with a term or phrase in the alert vocabulary for human review — the human reviewer confirms or clears the flag.
If a third party is on the line or the call transfers to another person, the agent's handling is reviewed
Third-party contact is a categorical FPC violation — there is no graded version. If the recording shows the agent discussed the borrower's account with a family member, employer, or neighbour without the borrower's documented written consent, the call scores zero on this criterion and the finding is escalated as a conduct violation regardless of the overall call score.
Agents may reference SARFAESI or DRT proceedings where these have been initiated — not as threats where they have not
An agent who says "we will file a case in court tomorrow" when no legal proceedings have been initiated is making an inaccurate legal representation. The Collections Agency Monitor AI cross-references the call's legal claims against the legal track status for that account from the Legal Notice Agent AI's register. A claim of imminent legal action for an account with no active legal proceedings is flagged as inaccurate.
If agent logged "PTP obtained" but the recording shows the borrower refused — fabricated outcome code
This criterion cross-checks the outcome code the agent submitted (from SLA Metric 6) against what the call recording actually shows. A "PTP obtained" outcome code for a call where the recording clearly shows the borrower saying "I cannot pay right now" is a fabricated record — and a Category A data integrity violation that escalates immediately to agency management and the institution's compliance team.
Agent must direct payment to the institution's official accounts only — never to an agent's personal account or cash
An agent who accepts cash payment directly, or who provides their personal UPI ID for payment, is creating a serious financial controls breach — the institution has no record of the payment, the agent may keep it, and the borrower has paid without getting credit. Call monitoring for payment instruction accuracy checks that all payment directions are to the institution's official account numbers and UPI IDs.
A live call audit: Agent ID CA-4821, Credence Collections
2 · Language: borderline threatening tone
The institution is accountable for every call made in its name — whether it listened to it or not
When a borrower files a complaint with the RBI Ombudsman about a collections call, the complaint is against the institution — not against the agency, whose name the borrower may not even know. The institution's response to the Ombudsman requires it to produce evidence of what was said on that call and what action was taken. An institution that samples and scores calls has that evidence. An institution that does not sample calls has no defence beyond the agency's word against the borrower's. The Collections Agency Monitor AI's call auditing is not a performance management tool — it is the evidence the institution needs to defend itself when a call goes wrong, and the early warning system that prevents calls from going wrong in the first place.