The challenge
Lounge-access support hit predictable spikes around peak travel hours across web, WhatsApp, and IVR, and the offshore CSAT team was absorbing the surge in eight languages with uneven quality. Travellers expected a sub-minute response; the operations team was running 20-minute queues at peak.
How we approached it
Ajuni dropped Asha into the existing Freshdesk and Twilio stack, with WhatsApp Business handling the chat channel and an IVR bridge into the voice queue. Eight languages were tuned against six months of resolved tickets, with deflection routing only the cases that genuinely needed a human into the queue.
Outcomes in production
- 87% of inbound queries auto-resolved end-to-end
- 8-language coverage went live in a single rollout
- Peak-hour response time dropped under one minute
- Human-agent queue volume cut by two-thirds
- 6 weeks kickoff to production traffic
Stack & guardrails
Integration & deployment
- Azure India · production tier
- Claude Sonnet 4 · API
- Freshdesk integration
- Twilio voice · IVR bridge
- WhatsApp Business API
- IndicTrans2 · language pack
Compliance & audit
- DPDP Act 2023
- ISO 27001
- PCI-aware redaction
Timeline
-
Week 0–1
Channel + ticket audit
Six months of resolved tickets sampled across web, WhatsApp, and IVR.
-
Week 2–3
Language pack tuning
Eight languages tuned and validated against historical resolutions.
-
Week 4–5
Shadow on live traffic
Asha drafted responses for human review; deflection threshold tuned.
-
Week 6
Production cutover
Full inbound traffic routed; CSAT monitored daily for first 30 days.
