India's digital lending market has a paradox: lenders have built sophisticated credit models, fast disbursement rails, and digital-first productsbut they're losing more than half their approved applicants at the video KYC step.
Why VKYC Drop-off Is So High
Video KYC requires the applicant to:
- Have a stable internet connection
- Position their phone camera correctly
- Hold up their PAN card in the right orientation and lighting
- Understand and respond to a set of compliance questions
- Complete a liveness check (specific gestures or movements)
For applicants in Tier-2 cities who are not tech-savvy and not comfortable with English instructions, this process is genuinely confusing. A human VKYC agent helpsbut human agents are only available during business hours, have variable quality, and can't handle the volume during application peaks.
The Language Barrier in VKYC
A VKYC agent who speaks only English or metro Hindi is structurally less effective with a Kannada-speaking applicant from Hubli or a Tamil-speaking applicant from Madurai. The applicant may not fully understand the instructions, may make mistakes in the process, and the agent may have to repeat instructions multiple timesleading to timeouts, failed attempts, and eventual drop-off.
Key finding: In one Bengaluru lender's data, Kannada-speaking applicants had a 74% drop-off rate at VKYC vs a 41% drop-off rate for Hindi-speaking applicants with the same VKYC agent. After Agni deployed in Kannada, the Kannada cohort drop-off fell to 18%below the Hindi baseline.
How AI-Guided VKYC Works
Agni guides the applicant through the VKYC process via voicein the applicant's preferred languagewhile the visual elements (camera, document) are handled through the existing VKYC interface.
The AI:
- Detects the applicant's language in the first 10 seconds
- Gives step-by-step instructions in that language ("Please hold your PAN card so the entire number is visible")
- Handles common failure states ("The image is a bit darkcan you move to a brighter area?")
- Handles compliance questions in the language ("Do you confirm that you are [Name], applying for a loan of ₹[Amount]?")
- Records the consent acknowledgment in the applicant's language
The Compliance Dimension
RBI's VKYC guidelines require specific disclosures and consent captures. Agni's VKYC guidance script is pre-approved against these requirementsevery session is compliant by construction, not by agent adherence.
Results at Scale
Across Agni's fintech VKYC deployments:
- Average VKYC drop-off reduced from 58% to 23%
- Average session completion time reduced from 38 minutes to 11 minutes
- First-attempt success rate improved from 61% to 94%
- Available 24/7no queue, no shift dependency
The incremental approved disbursals from recovered VKYC drop-off represent, for a lender doing 5,000 approved applications per month, approximately 1,750 additional loans disbursed per monthat zero additional acquisition cost.
