Running today at
Your hospital runs its cataract pathway on the phone: conversion callbacks, pre-op prep, post-op follow-up, recalls. NexaNeura automates those workflows end to end with AI voice agents that speak the patient's language, and puts every outcome back in front of your team.
Live at two eye hospitals in India · Built in Bengaluru
The full cataract journey as your hospital actually runs it. Pick any stage to see what the agent does and hear it on a real call. Each workflow switches on independently, and nothing gets re-plumbed when you add the next one.
NexaNeura is not a chatbot or a reminder tool. It is the phone workflows your hospital already runs, executed by AI voice agents instead of an overstretched telecalling desk.
A nightly export of who was advised surgery, who is booked, who was operated, and who is due back. Read-only, and we build the connector for whatever hospital system you run.
Calls in the patient's language, on the right day, with the right script for that stage. Retries on schedule, calls the attendant when the family decides, and stops when asked.
Bookings in the counsellor queue, red flags with the duty doctor, confirmations on the OT list. Every call recorded, transcribed, and reported.
Hindi, Kannada, Telugu, Tamil, Marathi, Bengali. And it calls the family member who actually decides, not just the patient.
A nightly list from the system you already run is enough to start. No new software for your staff to learn.
Patient data, recordings, and transcripts stay in your own cloud tenancy, in India, under your retention policy.
Our Head of Medical authors every symptom list and escalation rule, countersigned by your surgeon. The agent never diagnoses and never prescribes.
Post-operative calling is the one workflow that touches clinical risk, so the architecture is published rather than buried. Severity is assigned by a fixed rule set an ophthalmologist wrote. The model never decides how urgent something is.
No workflow that touches a symptom goes live until Dr. R. Poornika, our Head of Medical, has authored the protocol and your own ophthalmologist has countersigned the escalation table, including who holds the duty phone at night and on Sunday.
Your team does not log into a new system to work. They log in to check. Three views: where patients sit in the pathway, what the agents did today, and every clinical escalation with the time it took to reach a human.
Every stage is clickable in the live console down to the individual patient, the calls placed, and what the agent was told.
Placed against connected, by day. Connect rate is the number your telecalling team will care about most, and it is reported whether it moves up or down.
| Severity | Reported symptom | Patient | Routed to | Time to human | Outcome |
|---|---|---|---|---|---|
| Tier 1 | Sudden vision drop, day 3 | R. Devi, 68 | Dr Poornika | 4 min | Seen same day |
| Tier 1 | Pain with discharge, day 5 | M. Iqbal, 71 | Dr Rao | 7 min | Seen same day |
| Tier 2 | Drops finished early, day 6 | S. Lakshmi, 64 | Counsellor | 31 min | Refill arranged |
| Tier 2 | Redness increasing, day 2 | K. Prasad, 59 | Dr Rao | 18 min | Review booked |
| Tier 3 | Night glare, day 7 | A. Fatima, 66 | Script only | — | Reassured, on schedule |
Patient names shown are placeholders. In the live console this table is the medico-legal record: who reported what, when it reached a clinician, and what was decided.
Most healthcare voice AI answers the phone. That is a receptionist problem. Ours is a clinical operations problem: knowing that the day-3 call matters more than the day-1 call, that the escort is the barrier and not the price, and that the second eye is worth more than the first enquiry.
Every recovered patient below already chose your hospital, saw your surgeon, and was told they need an operation. The acquisition cost is spent. The workflows bring them the rest of the way. We label what is modelled and what is measured, because the two are not the same thing.
Eight workflows from referral to second eye. Closes once, at the second eye, then returns the patient for lifelong review.
Injection cycles every four to twelve weeks, indefinitely. Around half of Indian patients drop out of the cycle, and missing a dose costs vision permanently.
Decades of follow-up and daily drops with no end point. Adherence and review attendance are both poor, and the vision lost is never recovered.
Annual screening recall on the hospital's own diabetic base, where the relationship already exists and the recall is the entire intervention.
Health tech built without a doctor on the inside gets the workflow wrong. Our Head of Medical is a practising ophthalmologist who authors and signs every clinical protocol the agents run.
7+ years in HealthTech. Ex-Tricog Health, where he built AI ECG systems processing 30M+ ECGs. 4 research papers, 114+ citations. Co-founded and exited Aeonix Robotics.
Ophthalmologist with 8 years of experience, practising in Kolathur, Chennai. Authors and signs every clinical protocol, symptom screen, and escalation rule NexaNeura runs.
Fullstack architect leading platform development. Expert in scalable cloud systems, healthcare data standards, and production-grade microservices.
Specialises in CNN-based image processing and deep learning models. Leads NexaNeura's cataract grading and co-morbidity detection algorithms.
Ex-NASA researcher with deep expertise in AI applications in medicine. Guides model validation, clinical research design, and trials for NexaNeura.
22+ years in Quality and Regulatory, with expertise in NABL and ISO consulting and clinical research. Skilled in QMS, risk assessment, and audits.
Nobody is replaced. Your counsellors and telecallers keep doing what they are best at, which is closing the patient in front of them. NexaNeura works the list they never reach: the patient who left without booking, the day-60 callback, the second eye, the uncollected spectacles. It hands warm, confirmed intent back to your team with full context, so the same headcount converts more. In practice your telecallers stop dialling cold lists and start working qualified ones.
Yes, and it does not require an integration project to begin. A scheduled nightly export is enough: patient name, phone, preferred language, attendant contact, what the surgeon advised and for which eye, booking status, and surgery and review dates. Read-only. We build the connector against whatever system you run. Write-back for slots and notes comes later, when you want it, never as a precondition.
One scheduled export, typically half a day of work for whoever maintains your hospital system, and a single call with us to agree the fields. There is no software for your staff to install or learn. Outcomes arrive in your existing counsellor queue and as a daily report.
Hindi and Kannada today, with Telugu, Tamil, Marathi, and Bengali configured per site. The agent asks the patient's preference on the first call and remembers it. Scripts are written for older patients on ordinary handsets, not smartphone users, and the speech models are tuned for that population rather than for clean studio audio.
The post-operative call runs a fixed, tiered symptom list authored and signed by our Head of Medical and countersigned by your own surgeon. A red-flag answer ends the screen immediately and reaches your duty doctor within an agreed time target. An ambiguous answer is treated as the more serious tier. A call that drops mid-screen, or a patient we cannot reach inside the day-3 window, triggers a human callback the same day. The agent never tells a patient a symptom is harmless outside your approved script.
Inside your own cloud tenancy, in India. Your hospital remains the data fiduciary under the DPDP Act and we operate as your processor under a signed data processing agreement. Recordings and transcripts follow your retention policy, consent notices are delivered in the patient's language, and opt-outs propagate immediately and permanently across every workflow.
We start with the two workflows that carry no clinical risk, so you see a real number before committing to anything further.