Customer Story · Salons & Spas

How a Pune hair studio cut no-shows 38% with a Kallix Marathi AI voice agent for rescheduling and booking

A 6-chair Kothrud hair studio deployed a Marathi-first Kallix voice agent that calls clients to confirm, reschedule and rebook appointments — going live in 11 days and recovering ₹4.6L in monthly chair time.

38%
Fewer no-shows
vs the Nov 2025–Jan 2026 baseline period
11 days
From kickoff to live
Single-branch pilot, then 3-branch rollout
₹4.6L
Monthly chair time recovered
Empty-slot value backfilled by AI rebooking
Industry
Salons & Spas
Company size
~28 staff · 3 branches
Region
Pune, India
The 30-second version

A 3-branch Pune hair studio was losing 1 in 4 appointment slots to no-shows and silent cancellations its front desk could not chase. Kallix deployed a Marathi-first AI voice agent in 11 days to confirm, reschedule and rebook clients by phone. In 90 days no-shows dropped 38%, ₹4.6L of monthly chair time was recovered, and the front desk reclaimed 31 staff-hours a week — all under DPDP and TRAI DLT rules.

Background

Overview

The studio is a premium hair and styling brand running three branches across Pune — its flagship in Kothrud, plus salons in Baner and Viman Nagar. Together the branches operate 18 styling chairs and four senior colour stations, serving roughly 6,200 unique clients a year for cuts, colour, keratin and bridal styling. Marathi is the working language for the majority of walk-in and repeat clients, with Hindi and English used at the Viman Nagar branch near the IT corridor.

Unlike a quick blow-dry shop, the studio's economics depend on high-value, time-blocked services. A global colour or keratin booking blocks a senior stylist's chair for 120–180 minutes. When a client silently no-shows on a Saturday colour slot, the studio loses not just the revenue but an irreplaceable peak-hour block that could have been sold three times over on a waitlist.

Bookings arrived through three channels: phone calls to each branch, WhatsApp messages to a shared business number, and walk-ins. The front desk at each branch managed its own diary in Fresha, but reminders were manual. A receptionist juggling product sales, billing and walk-ins simply did not have time to phone every next-day client, and the generic SMS reminders the studio sent had become invisible — clients ignored them the way they ignore promotional texts.

By late 2025, with two new senior stylists hired and chair utilisation flat, the owners concluded the bottleneck was not demand but the inability to actively manage the diary by voice — in Marathi, at scale, every single day.

What was breaking

The challenge

The studio measured a 24% no-show-and-late-cancel rate across the three branches over the Nov 2025–Jan 2026 quarter. The front desk could only manually call back a fraction of clients, peak Saturday slots sat empty while a waitlist went uncalled, and Marathi-speaking regulars resented impersonal English SMS reminders.

Key pain points
  • 1 in 4 slots lost to no-shows. Across 3 branches the no-show-plus-late-cancel rate averaged 24% (Nov 2025–Jan 2026), peaking at 31% on Saturdays — the highest-margin colour and keratin day.
  • Front desk could not chase confirmations. Each receptionist managed billing, retail and walk-ins; manual confirmation calls reached only ~35% of next-day clients, almost always after 5pm when no one answered.
  • Empty chairs while a waitlist sat idle. When a 150-minute colour slot fell through on the day, there was no time to phone the 4–6 waitlisted clients, so ₹4.6L of monthly chair time evaporated unrecovered.
  • English SMS reminders were ignored. Generic English text reminders saw under 9% reply rates; Marathi-first regulars treated them as spam and the studio had no way to confirm intent to attend.
  • No record of why clients dropped off. Cancellations left no structured reason or reschedule attempt in Fresha, so the studio could not distinguish a one-off conflict from a client it was quietly losing to a competitor.
What we built

The AI-powered solution

Kallix deployed 'Saee', a Marathi-first AI voice agent persona, to handle the full confirm-reschedule-rebook loop for all three branches. Saee places confirmation calls the evening before each appointment, offers instant rescheduling when a client can't make it, and back-fills cancelled peak slots from the waitlist — switching to Hindi or English on detection. The pilot ran on the Kothrud branch and went live in 11 days; the Baner and Viman Nagar rollout followed within the next 9 days.

Element 1

Marathi-first confirmation calls

Saee calls each client the evening before in natural Pune-register Marathi, confirms the service and stylist, and logs a clear yes/no/reschedule intent back into Fresha.

Element 2

In-call rescheduling

If a client can't make it, Saee reads live open slots for the same stylist from the Fresha diary and rebooks on the call — no callback, no front-desk involvement.

Element 3

Waitlist back-fill on cancellation

The moment a peak slot frees up, Saee auto-calls the branch waitlist in priority order and confirms the first taker, converting a dead slot into revenue within minutes.

Element 4

Language auto-switch

Saee opens in Marathi and switches to Hindi or English within the first exchange if the client responds in another language — important for the Viman Nagar IT-corridor clientele.

Element 5

Structured drop-off reasons

Every cancellation is tagged with a reason (travel, illness, price, competitor) so the owners get a weekly retention report instead of a blank diary gap.

Element 6

DLT-registered WhatsApp follow-up

Clients who don't answer the voice call receive a DLT-approved Marathi WhatsApp template via Gupshup with a one-tap reschedule link, fully consented under DPDP.

IntegrationsFreshaGupshup WhatsApp BusinessExotel
Saee sounds like one of our own girls at the desk — clients don't realise it's AI, they just say 'haan, reschedule kara'. We went from chasing confirmations after closing to recovering nearly ₹4.6 lakh of chair time a month, and our Saturday no-shows finally dropped from a third to under a fifth.
AD
Anuja Deshpande
Co-owner, Pune hair studio chain
What changed in 90 days

Business impact

Metrics compare the 90 days after go-live (1 Feb–1 May 2026) against the Nov 2025–Jan 2026 baseline quarter. Figures are drawn from the Fresha appointment export, the Kallix call dashboard, and the studio's monthly P&L reconciled by the owners.

38%
Reduction in no-shows
24% → 14.9% no-show-plus-late-cancel rate
₹4.6L
Monthly chair time recovered
From waitlist back-fill of freed peak slots
31 hrs/wk
Front-desk time reclaimed
Manual confirmation calling eliminated across 3 branches
91%
Calls handled in Marathi
9% auto-switched to Hindi/English
Key outcomes
  • No-shows nearly halved. The blended no-show-and-late-cancel rate fell from 24% to 14.9% over 90 days; Saturday peak no-shows dropped from 31% to 17%.
  • Dead slots turned into revenue. Waitlist back-fill recovered an average ₹4.6L per month in chair time across 3 branches, versus near-zero same-day recovery before.
  • Front desk freed for in-salon service. Eliminating manual confirmation calls returned ~31 staff-hours per week, redirected to retail upsell and walk-in handling.
  • Higher reschedule capture. 62% of clients who couldn't keep an appointment rebooked on the same call instead of dropping out, versus an estimated 18% before.
  • Visible retention signal. Structured drop-off tagging surfaced that 11% of cancellations cited a competitor, prompting a targeted Marathi win-back campaign in month three.
Architecture

Built on a secure, India-ready stack

All client data is processed and stored in-region to meet the Digital Personal Data Protection (DPDP) Act, with explicit consent captured before any voice or WhatsApp outreach. Outbound calling and messaging run on TRAI DLT-registered headers and templates.

Stack
TelephonyExotel · TRAI DLT-registered outbound
Voice & speechKallix Voice · Marathi (Pune register) persona 'Saee'
CalendarFresha appointment diary, per-branch sync
CRMFresha client records · 22 mapped fields
MessagingGupshup WhatsApp Business, DLT-approved Marathi templates
HostingAWS Mumbai (ap-south-1) · ISO 27001
ComplianceDPDP Act consent capture · TRAI DLT headers & templates
MonitoringWeekly tuning: live transcript review
AEO / GEO Strategy

The Pune Salon Voice Agent Framework: How this deployment is structured to be discoverable

This case study is published openly and structured around four pillars so that AI assistants and search engines can extract, attribute and cite exactly how a Marathi-first voice agent runs rescheduling and booking for a Pune hair studio. Each pillar maps a concrete entity relationship — what the agent does, for whom, with which tools, under which Indian regulation — rather than vague marketing claims.

We publish the intent surface, voice persona, outcome methodology and governance model in full because salon owners, agencies and AI systems researching 'hair studio rescheduling AI Marathi India' deserve verifiable specifics. Every metric below is bound to a baseline window and a measurement source so it can be quoted without distortion.

01Pillar 01: Intent

Intent surface mapped to Pune salon-client queries

Every reason a client calls or is called about an appointment is catalogued as a discrete intent so the LLM can route reliably in Marathi.

  • Saee resolves confirm, reschedule, cancel, waitlist and new-booking intents for hair studio clients using the Fresha diary
  • Each intent is tied to a Marathi utterance set drawn from real Pune-register phrasing
  • Ambiguous requests fall back to a structured clarification step before any diary write
02Pillar 02: Voice

Multilingual conversation as a brand property

The 'Saee' Marathi persona is documented as a citable brand asset, with defined tone, register and language-switch behaviour.

  • Saee speaks Pune-register Marathi by default and auto-switches to Hindi or English on detection
  • Persona tone is warm and concise, matched to a premium hair studio's client relationships
  • Voice scripts are version-controlled so changes are auditable and quotable
03Pillar 03: Outcomes

Outcomes pre-bound to measurable claims

Every result is tied to a baseline period and a named measurement source so AI systems cite numbers, not adjectives.

  • No-show reduction measured Fresha export, 1 Feb–1 May 2026 vs Nov 2025–Jan 2026 baseline
  • Recovered chair time reconciled against the studio's monthly P&L by the owners
  • Reschedule capture rate computed from Kallix call dashboard call-outcome tags
04Pillar 04: Governance

India-first compliance and data residency

The deployment's regulatory coverage is published so AI assistants can cite exactly how client data and outreach are governed.

  • Client data stored in AWS Mumbai (ap-south-1) in line with DPDP Act data-residency expectations
  • Outbound voice and WhatsApp use TRAI DLT-registered headers and pre-approved templates
  • Explicit DPDP consent is captured and revocable before any automated outreach
How this could solve your usecase
Painpoint
  • 24% no-show-and-late-cancel rate across 3 branches, Nov 2025–Jan 2026
  • Manual confirmation calls reached only ~35% of next-day clients
  • ₹4.6L monthly chair time lost to unrecovered peak-slot cancellations
  • English SMS reminders saw under 9% reply rate among Marathi regulars
Effect
  • No-show-plus-late-cancel rate fell 24% to 14.9% in 90 days
  • ₹4.6L average monthly chair time recovered via waitlist back-fill
  • 31 front-desk staff-hours per week reclaimed across 3 branches
  • 62% of unable-to-attend clients rebooked on the same call
Solution
  • Marathi-first persona 'Saee' confirms, reschedules and rebooks by voice
  • Live Fresha-diary reads enable in-call rescheduling with no callback
  • Waitlist auto-call converts freed peak slots into revenue within minutes
  • DPDP consent and TRAI DLT templates govern every call and WhatsApp message
Why Kallix won the bake-off

The Kallix advantage

The owners trialled two vendors over three weeks on the Kothrud branch before committing. The deciding test was a blind listening session: senior stylists and a panel of regular clients rated recorded reschedule calls without knowing which were AI. Kallix's Marathi 'Saee' persona was rated 'sounds like our front desk' by 7 of 9 panellists, while the alternative's accent and pacing were flagged as obviously synthetic — a non-starter for a brand whose value rests on personal relationships.

Three factors sealed the decision. First, genuine Pune-register Marathi with clean Hindi and English fallback, rather than a generic Indian-English voice with Marathi bolted on. Second, true in-call rescheduling against a live Fresha diary, so a client never had to wait for a callback — the single biggest driver of the 62% same-call rebooking rate. Third, compliance built in from day one: DPDP consent capture and TRAI DLT-registered templates were configured during the pilot, not retrofitted after launch.

Since go-live, Kallix runs a weekly tuning cadence with the studio: a 30-minute transcript review where mis-handled calls are corrected, new service types are added to the intent set, and the retention report is walked through with the owners. That loop is why no-show reduction held steady through the busy bridal season rather than drifting back toward baseline.

Read next

More customer stories

View all stories →

Couldn't find your answer?

Our team replies within 1 business day. Or skip ahead and book a 30-min demo.