A digital consumer lender answers routine support tickets with AI and escalates the rest
An AI agent that classifies, answers and routes customer tickets in English and Hindi for a digital lending app, with RBI-aligned escalation for grievances.
- of tickets resolved without an agent
- 62%
- of tickets resolved without an agent
- median first response
- 4 h → 6 min
- median first response
- languages: English & Hindi
- 2
- languages: English & Hindi
- grievances routed to a person
- 100%
- grievances routed to a person
The client
Kissht offers consumer loans through its app across India. Most support tickets ask about EMI dates, statements, NOCs and payment status, while a smaller share are formal grievances with regulatory response times.
The challenge
- ✕Ticket volume spiked around EMI dates, stretching response times to hours.
- ✕Agents answered the same questions about EMI dates and NOCs hundreds of times a day.
- ✕Grievances were sometimes mixed in with routine tickets and missed their deadlines.
- ✕Customers wrote in English, Hindi and Hinglish.
The solution
Ticket classification
Each ticket tagged by intent, language and urgency within seconds of arrival.
Grounded answers
Routine questions answered from the customer's loan record and approved help content only.
Grievance routing
Complaints detected and routed to the grievance team with deadlines tracked.
Agent assist
Draft replies and loan context shown to human agents for the tickets they handle.
How we delivered
- 01Two weeks labelling historical tickets to measure accuracy before launch.
- 02Shadow mode for two weeks with agents reviewing every AI answer.
- 03Gradual rollout by intent, starting with EMI dates and statements.
- 04Weekly accuracy reviews with the compliance team.
The results
- ✓62% of tickets are resolved without an agent.
- ✓Median first response fell from four hours to six minutes.
- ✓Every grievance reaches a person within its deadline.
- ✓Agents spend their time on the complex cases.
“EMI week used to bury the support team. Now the routine questions answer themselves and our people handle the cases that need them.”