Xwits Healthcare — everything your clinic runs on, in one place

You already know this feeling

It is 6pm and you are not certain.

Not about the diagnosis. About everything around it.

—The patient you prescribed for this morning. Were they on something that interacts? You asked. They said no. You are fairly sure.

—Three people did not turn up today. Nobody rang them yesterday, because nobody knew which three to ring.

—The bill from Tuesday. It went out. Did it get paid? Somebody would have said something.

—NABH is in four months and the evidence lives in a register, two spreadsheets and one person's memory.

None of this is a clinical failure. It is a clinic running on attention, and attention runs out at about four in the afternoon.

A day

How a visit actually goes.

Nothing here is a separate system that somebody has to remember to update. It is one day, recorded once.

01They arrive. Registered at the desk, or they booked themselves. A booking is not the same as somebody in the room, and the record knows the difference.

02They wait. The queue shows who is next and how long they have been sitting there.

03They are seen. Notes, vitals and diagnostics on one record, with suggestions the doctor can accept or ignore.

04They are prescribed for. Checked against their allergies and their other medicines before it can be signed.

05They pay. A GST-correct bill, payable by a link sent to their phone.

06They come back. The follow-up is booked and chased — and never sent to somebody who asked not to be contacted.

Beyond the consulting room

For clinics that became hospitals.

Beds and wardsAdmissions, transfers and discharge, with occupancy you can see rather than guess.

Operating theatresScheduled by list, with the same guarantee against double-booking as a bed.

CredentialingEvery practitioner's registration and its expiry, flagged before it lapses rather than after.

InventoryStock, batches, expiry and replenishment, with procurement and supplier invoices behind it.

Revenue cycleClaims, the daily close, and the worklist of what has not been collected.

TelemedicineRemote consultations with the same record, the same prescription rules and the same consent.

Safety

The mistakes it will not let you make.

A bed or a theatre cannot be double-booked. Not because a screen prevents it — because the database rejects the second booking outright, so there is no race for two receptionists to lose.

Allergy checksRun before a prescription can be signed, against the allergies on the patient's record.

Consent, under DPDPARecorded, revocable, and honoured — a patient who opts out is not messaged again.

NABH recordsThe quality evidence an accreditation asks for, produced as you work rather than reconstructed.

An audit trailWritten whether or not anybody asks for it, because the day you need it is the day you cannot make it.

One clinic, one set of recordsNo clinic's data can reach another's. Enforced three times over, the last by Postgres itself.

Nothing messages a patient aloneAnything leaving the clinic waits for a person the first time it happens.

The part you are right to be careful about

What the software does, and what it will never do.

Software that helps in a clinic has to earn a specific kind of trust, and "AI-powered" earns none of it. So here is exactly what runs, and where it stops.

It never makes a clinical decision. The allergy check runs before a prescription can be signed — enforced, not advisory: prescribing something on the patient's recorded allergy list is refused until the clinician gives a reason, and that reason is recorded as theirs.

The no-show model is not a black box It is a calibrated logistic regression over how appointments have actually behaved at your clinic — booking lead time, hour of day, past attendance. You can read what it weighed.

It cannot learn to predict who is poor Religion, caste, gender and disability are not inputs. Not filtered afterwards — never collected as features. The two that could stand in for them are marked and reviewed. A model that quietly starts flagging one community is the failure this is built to prevent.

Its fairness is measured, not assumed Calibration, discrimination and a disparity ratio across groups, checked continuously — and it refuses to report any of them below a minimum sample, because a ratio computed over four appointments is noise with a decimal point.

The score never reaches the patient Nobody is told a machine thinks they will not turn up. Your front desk sees who to ring first. Your owner's view is counts by time of day, without names.

Forecasts say how they were made Next Tuesday's expected footfall is a day-of-week mean over your own history with a confidence band — and the method and the number of days behind it print beside the figure. A forecast built from two matching days announces itself as one.

Identifiers never leave the building Where a language model drafts text, Aadhaar, PAN and card-like numbers are stripped before the request and never restored. Phone and email are swapped for placeholders and put back afterwards, so the model works without ever holding them.

Nothing messages a patient on its own Anything leaving the clinic waits for a person the first time that kind of message is sent. Approve it once and it is yours; it does not quietly widen to something else.

Your records are yours No other clinic can reach them — enforced three times over, the last by the database itself, which no bug in our code can talk out of it. Consent is recorded and honoured under the DPDP Act.

If any of this sounds cautious, it is because a clinic carries the liability and the software does not. The correct posture for a machine in your consulting room is to be useful and to stay in its place.

Everything your clinic runs on, in one place.