What Counts as Mis-Selling on a Call?
Mis-selling in financial services is broadly defined as selling a product to a customer by providing incorrect, incomplete, or misleading information. On a call, this typically happens in five ways:
- False claims: Stating that a product has features, returns, or protections it does not actually have
- Omission of material facts: Deliberately not mentioning key limitations, exclusions, charges, or risks
- Unsuitable product recommendation: Recommending a product that is not appropriate for the customer's stated situation or needs
- Pressure selling: Using coercive language, artificial urgency, or threats to force a decision the customer would not otherwise make
- Impersonation or misleading association: Implying government backing, guaranteed safety, or comparison with products the customer trusts
🚨 RBI, IRDA, and SEBI have all increased penalties for mis-selling in the past three years. Individual agents and their managers can now face personal liability in addition to organisational penalties.
Real Mis-Selling Phrases That Appear on Indian Sales Calls
These are real categories of language that appear on Indian financial services calls and create regulatory risk when transcribed and reviewed:
"Isme guaranteed return milega — 12% har saal"
No market-linked product can guarantee returns. This statement on a call is an IRDA violation for insurance products and an SEBI violation for mutual fund-linked products.
"Yeh government ki scheme hai" (for a private product)
Implying government backing for a private financial product is fraud. AI detects government association language when applied to products that lack it.
"Aaj hi lena hoga, kal offer band ho jayega"
Artificial urgency to prevent informed decision-making is a known mis-selling tactic. Detected as high-pressure language in sentiment analysis.
"Koi charges nahi hain, bilkul free hai" (when charges exist)
Omitting processing fees, GST, or renewal charges constitutes mis-selling. AI cross-checks the transcript for disclosure of all known product charges.
"Yeh same hai jaise FD" (for a ULIP or market-linked product)
Comparing a market-linked product to an FD in terms of safety or returns misleads customers about risk. This category of phrase is detectable through product comparison language patterns.
💡 A single IRDA complaint, if upheld, results in a mandatory refund of premiums, penalties on the insurer, and marks on the agent's IRDAI licence. The cost far exceeds the value of the commission earned.
How AI Detects Mis-Selling on Calls
AI mis-selling detection works in two layers:
Layer 1: Transcript-Based Phrase Detection
After transcribing the call, the AI scans the full text for known mis-selling language patterns — guarantee claims, government association language, pressure phrases, fee omission markers, and unsuitable product comparisons. These are flagged as compliance risks in the audit output.
Layer 2: Sentiment and Consent Analysis
Beyond specific phrases, the AI analyses whether the customer showed genuine understanding and willingness, or whether their responses suggest confusion or coercion. A customer saying "okay okay" with flat sentiment while an agent rushes through disclosures is a different risk profile from a customer who asks clarifying questions and provides positive confirmations.
The Cost Equation: Auditing vs Complaints
Consider a 50-agent insurance sales team making 200 calls per day. At 2% manual QA coverage, 4 calls are reviewed daily. If 5% of calls contain mis-selling language, approximately 10 calls per day have a compliance risk — and 9.8 of them are never heard by anyone.
If just 1% of those unreviewed calls results in a complaint — that is roughly 2 complaints per month, at a conservative cost of Rs 1 lakh each in refunds and penalties. Rs 24 lakh per year in complaint costs, all preventable with better call coverage.
AI auditing of all 200 daily calls costs a fraction of one complaint.