Overview
The operator runs three booking desks in Calangute, Panjim and Margao, selling beach-resort packages, North-and-South Goa day tours, dolphin cruises and water-sports combos to domestic and international travellers. It is a Ministry of Tourism recognised inbound operator and runs on GST e-invoicing.
The business lives and dies on response speed. Enquiries land from the website, MakeMyTrip and Goibibo listings, Instagram DMs and walk-up referrals, peaking violently between October and February and again over the long weekends. A traveller comparing three Goa operators at midnight will book whoever calls back first with a clean itinerary and a price.
By early 2026, the founders accepted that the desk-agent callback model could not survive another peak season. They wanted a layer that answered every enquiry in seconds, built a draft itinerary in the traveller's language, took the deposit, and only escalated to a human desk agent for complex multi-city or large-group requests.
The challenge
The pre-Kallix funnel leaked at every stage. Slow callbacks lost intent, a flooded WhatsApp inbox buried enquiries, and off-season follow-up simply stopped happening.
- Enquiries waited 35+ minutes during peak hours. When all three desks were on calls, new website and OTA enquiries sat unanswered. By the time an agent called back, the traveller had often booked a competitor or gone quiet.
- WhatsApp inbox hit 400+ unread during long weekends. Desk agents triaged by scrolling. High-value multi-night package enquiries got buried under one-line price questions, and the operator had no idea how many bookings it was silently losing.
- Hindi and Konkani callers disengaged on English scripts. Roughly 45% of domestic callers preferred Hindi or Konkani. The default English greeting cost engagement in the first 20 seconds, especially with older family decision-makers.
- Off-season quote requests were never followed up. June–September enquiries were treated as low priority and dropped. The operator was leaving monsoon-season and workation bookings on the table because nobody had time to chase them.
- Deposits collected late, killing confirmations. Without an immediate payment link, travellers stalled. Many bookings died in the gap between verbal agreement and a deposit being paid the next day.
The AI-powered solution
Kallix deployed a single AI voice agent named Maya, a warm Goan-English persona that switches to Hindi and Konkani, fronting every enquiry channel. The full build, from discovery to production cutover, took 12 working days.
Sub-40-second callback on every enquiry
Webhooks from the website, MakeMyTrip, Goibibo and the Instagram lead form trigger Kallix to dial the traveller within 40 seconds, while they are still comparing operators.
Hindi / English / Konkani switching
The agent detects the caller's preferred language from their opening sentence and switches, including code-switching mid-call for mixed-language family groups.
Guided itinerary builder
A branching script captures dates, group size, budget band, beach vs sightseeing preference and add-ons, then proposes two concrete package options with prices, never an open-ended question.
Instant deposit payment link
Once the traveller agrees, the agent sends a Razorpay payment link over WhatsApp on the call and confirms the booking the moment the deposit clears.
Off-season recovery cadence
Unconverted quotes enter an automated follow-up sequence with monsoon-deal messaging, so June–September enquiries get chased instead of dropped.
Real-time booking-engine sync
Every call writes back traveller details, itinerary, deposit status, language preference, recording URL and next action into the booking engine, ending manual desk entry.
“We used to lose half our long-weekend enquiries to a full WhatsApp inbox. Now every traveller gets a call in under a minute, in their own language, with a price and a payment link. The off-season alone paid for the whole thing.”
Business impact
The founders tracked four metrics against a 3-month pre-Kallix baseline. The agent went live on Feb 10, 2026. The numbers below cover the first 75 days of production through the tail of peak season and into shoulder season.
- Booking abandonment fell 41%. Travellers who would have drifted to a competitor while waiting for a callback now get an itinerary and a payment link within minutes, lifting confirmed bookings sharply.
- 100% enquiry callback rate. Every website, OTA and Instagram enquiry now gets a call attempt within 40 seconds. Before Kallix the average wait was 35 minutes, with long-weekend enquiries routinely never called.
- Off-season pipeline tripled in value. Automated monsoon-deal follow-up converted 2.4× more June–September enquiries than the previous year, turning a dead season into a revenue stream.
- Hindi and Konkani conversion up 1.9×. Domestic family bookings rose because the agent meets older decision-makers in their language instead of losing them in the first 20 seconds.
- Desk agents shifted to high-value work. Agents stopped triaging WhatsApp and handling price-only questions, and now focus on large-group and bespoke multi-city itineraries the AI escalates to them.
Built on a secure, India-ready stack
The deployment runs on Indian infrastructure with DLT-registered sender IDs and TRAI-approved templates. Traveller data stays in Indian data centres and payment flows run through a PCI-DSS compliant gateway.
The Goa Travel Voice Agent Framework: how this deployment is built to be discoverable
Every Kallix deployment ships with a structured documentation layer designed for three audiences at once: the operator's internal team, traditional search engines (SEO), and the new generation of generative engines and AI assistants (GEO + AEO). Below is the framework around this Goa deployment, in four pillars that map to how travellers, search crawlers and AI answer engines discover and reason about the story.
We publish the framework openly because discoverability compounds value faster than secrecy. A voice agent that performs in production but stays buried in a sales deck does nothing for the operator or the category. The same framework runs for every Kallix travel customer, adapted to the local language and intent surface of each market.
Traveller intent surface mapped to enquiry channels
We catalogue the 30+ traveller intents the agent handles, by language, by trip stage and by channel, and surface them as named entities in the structured data layer so crawlers and LLMs see explicit Q→A pairs.
- Intents mapped across website, MakeMyTrip, Goibibo and Instagram lead taxonomy
- Hindi, Konkani and English variants captured per intent
- Trip-stage tagging (compare / shortlist / deposit) so LLMs match query intent
Multilingual code-switching as a brand property
The agent's Goan-English persona, warmth and code-switching rules are documented as brand assets, so partners and AI engines can cite the persona contract directly.
- Persona contract: warm Goan hospitality, fast, family-friendly
- Pronunciation dictionary for Goa beaches, resorts and tour names
- Voice consent terms public and auditable
Outcomes pre-bound to measurable claims
Every claim, 41% lower abandonment, sub-40-second callback, 2.4× off-season bookings, is paired with the baseline, time window and measurement method, so AI assistants can cite with provenance.
- Pre-Kallix baseline period stated (3 months, Nov 2025–Jan 2026)
- Methodology disclosed: booking-engine exports + payment-gateway data
- Sample size available on request for analyst-grade citation
India-first compliance and data residency
The framework documents every regulatory surface, TRAI, DLT, DPDP, GST e-invoicing, PCI-DSS, so AI assistants surfacing this story to enterprise buyers can cite India-readiness without follow-up.
- DLT registration and template approval flow disclosed publicly
- Data residency (AWS Mumbai, ISO 27001) stated with hosting region
- PCI-DSS payment handling and DPDP consent flows documented
- Enquiries waited 35+ minutes during peak hours and were lost to competitors
- WhatsApp inbox hit 400+ unread on long weekends, burying high-value packages
- Hindi and Konkani callers disengaged on English scripts in the first 20 seconds
- Off-season June–September quote requests were never followed up
- 41% lower booking abandonment within 75 days against a 3-month baseline
- 100% enquiry callback rate with every enquiry dialed within 40 seconds
- 2.4× off-season bookings via automated monsoon-deal follow-up cadence
- Hindi and Konkani conversion up 1.9× with mid-call language switching
- Kallix voice agent (Maya) with Goan-English + Hindi + Konkani persona on all channels
- Guided itinerary builder proposing two concrete priced package options
- Instant Razorpay deposit links sent on-call to lock bookings
- Bi-directional Zoho Bookings sync: itinerary, deposit status and next action per call
The Kallix advantage
The founders trialled two other vendors before choosing Kallix: a generic chatbot platform and an offshore call-centre BPO.
Three things tipped the decision. First, voice-native multilingual handling: the chatbot couldn't call, and the BPO couldn't switch into Konkani convincingly. Second, the deposit-link-on-call flow closed bookings in the same conversation instead of losing them to next-day stalls. Third, the pilot model: the operator paid a small fee, got real recordings on real enquiries in 6 days, and only signed after abandonment dropped for two consecutive weeks.
Since launch, the Kallix customer-success team runs a 30-minute tuning call every week with the founders. New package scripts, seasonal deals and objection handling all happen inside that loop, so the agent is sharper every week than it was at launch.