The ChallengeTrusted Care, Built Entirely on Paper
An oral surgery practice in a tier-3 Indian city has spent the last decade earning the trust of more than 5,000 rural patients, one consultation at a time. Patients in this market often travel hours for specialist care, or skip it entirely. The practice earned real trust. But it ran its entire workflow on paper.
The gap showed up right after the consultation. Without digital records, staff couldn't check a patient's treatment history, couldn't see who'd stopped treatment midway, and couldn't tell who was overdue for a follow-up. Instead of a system, the clinic relied on phone calls someone remembered to make, and WhatsApp broadcasts it sent to every patient at once, whether the message applied to them or not. Care reached patients late. Retention depended on chance. And the practice lost revenue it never even knew it was missing.
Engineers at Madspek asked the obvious question first: why not just use standard hospital EMR software? The practice had already answered that question. Enterprise-grade systems cost too much and needed more staff than a single practice could spare. Cheaper alternatives raised doubts about usability, reliability, and patient data handling. This effectively became the brief for us!
Building Around the Constraint, Not Despite ItA Digital Foundation, Then a Conversational Layer
The foundation had to come first. Using Claude Code, Madspek's engineers built a secure, DPDP-compliant system that moved the clinic's core workflow, patient onboarding, treatment history, and prescriptions, off paper and into a structured database. They then added Google's Gemini LLM as a conversational layer, so the doctor and approved nurses could pull up a record or log an update just by asking for it, alongside the regular digital forms.
That solved record-keeping. It didn't solve the harder problem: reaching the right patient with the right message at the right time, without adding work for the staff.
Three Agents, One LifecycleSpecialized Agents on Gemini Enterprise Agent Platform
To close that gap, Madspek deployed three specialized agents on Google Cloud's Gemini Enterprise Agent Platform. Each agent handles one link in the same chain. Individually, each agent automates one step. Together, they run a continuous patient lifecycle engine, something the practice never had the staff or systems to run itself.
| Agent | Operational Role |
|---|---|
| The Database Analyst | Reviews every patient's treatment history and flags who's overdue for a checkup, who needs a follow-up, and who left treatment incomplete. Those patterns matter for the patient's health and the clinic's revenue. |
| The Vernacular Copywriter | Takes those flags and drafts a personalized, empathetic reminder in Hindi for each patient, based on their specific clinical context, not a generic template. |
| The Outbound Manager | Takes over once the doctor or staff approves a batch of messages. Schedules and triggers the voice calls or texts through a third-party communications provider. |
The ImpactEarly Numbers Already Point in a Clear Direction
The platform is now live, and the early results already point in a clear direction.
Structured patient outreach has roughly tripled.
The clinic projects a 15% revenue uplift on a current ARR of about ₹1 crore (~$105,000).
Staff are recovering an estimated 20% of clinic hours per week that used to go into manual follow-up.
Proactive outreach is projected to save patients more than ₹10 lakh a year (~$10,500) in costs they'd otherwise pay for delayed care.
None of this asked the clinic to become a different kind of business. It just needed a system built around the workflow it already had, not a scaled-down version of hospital software. That's usually where the real opportunity sits for small, non-digitized practices: not adopting AI in the abstract, but building it narrowly enough to fit how the business actually runs.
Ready to Digitize Your Practice?
Have a workflow like this one, running on paper or spreadsheets instead of a system? Talk to Madspek about an AI Readiness Audit.
