Discovery and enquiry
Find the right number among 30+, or fill a form and wait for a callback.
Enquiries go cold. Self-pay patients book elsewhere.
SHALBYCare AI
01Answers when
patients call
patients call
02Checks in every day
after surgery
after surgery
03Brings in a nurse
when something's wrong
when something's wrong
Care,
between visits.
An AI voice companion for every Shalby patient.
Digital Transformation · IIM Ahmedabad
Prof. Pankaj Setia · 30 October 2026
Arnav Dixit · Bhavya Jain
Shreya Gupta · Raima Chatterjee Ray
01 · Problem
Shalby's constraint isn't beds or doctors.
It's the conversation.
Every patient who drops out between enquiry, appointment, surgery and follow-up is someone whose recovery the hospital no longer sees.
0–0%bed occupancy, all of FY26
Management says existing hospitals can take 40–50% more patients without major capex.
Revenue growth, FY26
₹1,168 crore revenue. Growth has slowed while beds sit empty.
Who pays
33% self-pay37% insurance30% government
Self-pay patients compare hospitals and drift away if no one engages them quickly.
02 · Where engagement breaks
Five places patients slip away.
Triage and routing
Patients don't know which department or doctor fits their symptom.
Wrong-specialty visits, re-bookings, a poor first impression.
Appointment adherence
Reminders, if any, are one-way SMS. No way to confirm or reschedule.
No-shows and late cancellations leave slots empty.
Post-discharge recovery
Days 1–30 after surgery are largely unmonitored.
Complications caught late. Follow-ups missed.
Visit preparation
Clinicians start every follow-up from zero.
Longer consults, missed red flags, less confidence.
KeptMissedNever return
03 · Appointment no-shows
0 lakhbooked OPD visits a year
0missed. An estimated 25% no-show rate.
0never come back at all.
Reach is what matters. At University Hospital Zurich, phoning high-risk patients changed nothing (21.9% vs 22.0%), because 116 of 274 couldn't be reached.
An agent that calls at scale, at the right time and in the patient's language fixes exactly that.
04 · Follow-up after surgery
0 surgeries a year.
The next 30 days decide the outcome.
30%
miss their post-surgery follow-up today.
82.8% vs 69.7%
attend follow-up when the hospital phones them.
Daily.
One-off calls don't cut readmissions (7 trials, 7,075 patients). Care has to be daily, personal and focused on risk.
05 · Revenue leakage
₹0 crore
leaks out every year between visits. That's 3.4% of FY26 revenue (range ₹24–61 crore).
One OPD visit can become a joint replacement, at about ₹1.63 lakh per admission.
06 · Shalby's existing approach
Nine doors. Thirty numbers. One callback.
“Request an Appointment. Our Executive will contact you soon.”shalby.org booking page
Already strong
OTP-verified contacts · DPDP consent · 23 languages · cloud lab · a trusted orthopaedic brand
Missing
Real-time booking · a link to the hospital system · post-discharge monitoring · one shared journey
07 · Transformation purpose
Every Shalby patient is known, reached and cared for between visits, in their own language, so that no patient who needs us is lost to delay, forgetfulness or silence.
Existential
Shalby exists to restore mobility. Its outcomes are decided after the patient goes home.
Emotional
Patients and families want to feel known, safe and looked after between visits, not only during them.
08 · The pursuit of computational freedom
From bounded rationality to computational freedom.
Like a driverless car: intelligence comes from software plus data. Here it's the agent plus the patient record.
TodayWith Shalby Care AI
Patient
Remembers doses, tracks pain, finds the right number
Called daily, reminded, booked on the first call
Caregiver
Unsure what's normal, no one to ask after hours
A 24×7 voice in their language, escalation when needed
Nurse
Can't call every patient daily; signals are invisible
Works a ranked risk queue with plain-language reasons
Doctor
Reconstructs history at every visit
Reads a one-card pre-visit brief
Shalby
Engagement limited by callback headcount
Engages every patient continuously, at any volume
09 · DaWoGoMo assessment
Change all four together.
12345
Da Digital architecture
Wo Work
Go Governance
Mo Business model
Cu Culture
Maturity 1 (ad hoc) to 5 (adaptive). Grey: today. Lime: 12-month target. Work and the business model have the furthest to go.
Business-model options · valence × feasibility
Valence →Feasibility →
A Cost-out: replace staff with bots
B Wellness, engagement and care continuity · Lead option
C Partner care platform · year 2
D License to other hospitals · option
10 · The future-state journeys
Same agents. Same record. Staff act on exceptions.
Journey A · a new patient calls · 9:40 pm
दो हफ़्ते से कमर का दर्द पैर तक जा रहा है।
I'm Shalby's AI assistant. Is it okay if I ask two quick questions?
Safety net checked every sentence · no red flags
I'd suggest the spine surgeon. Tomorrow at 10:30 or 4:15?
10:30, please.
Booked after a clear yes · reminders 48 h and 3 h ahead · doctor gets a brief
Journey B · recovering from a knee replacement
- DischargePatient and caregiver consent to daily calls in Gujarati, Hindi or English.
- Every dayHow are you feeling? Medicines, pain vs yesterday, swelling, fever, sleep, mood, exercise.
- Missed blood thinnerGoes straight to the nurse's queue.
- Calf swelling or chest painCall stops, safety advice given, nurse and on-call doctor alerted.
- Day 14Follow-up, refills and tests booked after a clear yes.
11 · Live demo
Let's see it live.
Risk-ranked call queueA daily care callThe appointment lineA pre-visit brief
1 · ChannelsPatient phoneWhatsAppWebsite / appStaff console
2 · Voice & languageSpeech-to-textTurn detectionText-to-speechLanguage ID
3 · Safety & policyRed-flag netNegation parserConsent + identityPII redaction
4 · Agent engineCare CompanionAppointment lineTier-rule gateBrief writer + ranker
5 · Data & integrationIntegration hubCare recordABDMAnalytics / KPI
6 · Models (swappable)Enterprise LLMIndic speechRisk scoreEvaluation
12 · AI architecture
The model proposes.
The rules dispose.
Every sentence passes safety nets before the model sees it. Every action passes tier rules in code before it touches the record.
BuyTelephony, Indic speech, foundation models
PartnerThe voice-agent runtime
Build & ownSafety nets, tier rules, risk score, scripts, care record
13 · Safe and explainable by design
Four autonomy tiers, enforced in code.
1
Auto
Logs a dose, saves the check-in, routes to a doctor.
2
Confirmed
Books a follow-up, refill or test only after a clear yes.
3
Human
Missed blood thinner or a dose question goes to a nurse.
!
Escalation
Calf swelling, chest pain: stop, give safety advice, alert.
Explainable risk score · points per signal
Medium at 22, High at 45. The AI may reorder the list with a one-line reason, but never replaces the score.
14 · Impact and KPIs
What changes every year.
0
more booked visits kept
No-shows 25% → 18%0
more bookings from enquiries
24×7 first-call booking0
more patients followed up
Attendance 70% → 82%0 lakh
wellness check-ins
~10 calls per surgical patientSafety KPIs are gates. Any breach pauses that capability.
100%emergency redirect≥99%red-flag recall0wrong-information incidents≥95%high-risk patients contacted same day≤1.5 sreply time
15 · Business Model Canvas
Engagement becomes a wellness and retention engine.
Key partnersVoice-agent platform · Indian telco · enterprise LLM (India region) · HIS / LIS vendor · NHA (ABDM) · insurers · home care, pharmacy, physio
Key activitiesDaily wellness calls · intake, routing, booking · risk queue and briefs · safety review
Key resources30 years of orthopaedic outcomes · clinician-signed rules · care record · agent engine
Value propositionPatients: one warm voice in their language, 24×7, never alone in recovery.
Doctors: a pre-visit brief and fuller clinics.
Shalby: loyal patients and early warning.
Doctors: a pre-visit brief and fuller clinics.
Shalby: loyal patients and early warning.
RelationshipsAutomated but personal · human by exception · always says it's an AI
ChannelsPhone in and out · WhatsApp read-backs · web call · staff console
Customer segments≈27,600 surgical patients and caregivers · new OPD callers across 13 units · international · scheme and insured
Cost structureVoice ₹1.14 cr · team ₹1.60 cr · security & DPDP audit ₹0.30 cr · LLM for briefs ₹0.15 cr a year · one-time setup ₹2.25 cr
Value capture~17,200 more visits kept · ~3,300 more follow-ups · ~2.8 lakh check-ins · retention into rehab, home care and second-joint surgery
Shalby Care AI
Never alone between visits.
≈17,000 more visits and 3,300 more recoveries kept on track every year.
Arnav Dixit · Bhavya Jain · Shreya Gupta · Raima Chatterjee Ray
Sources: Shalby FY26 results and earnings call · shalby.org · Boggan et al., Annals of Internal Medicine 2025 · University Hospital Zurich no-show study · Int J Community Med Public Health · IIM Ahmedabad, Digital Transformation (Prof. Pankaj Setia)