Administration — HOS · Sri Venkateshwara Multi-Speciality Hospital

Hospital at a glance

Sri Venkateshwara Multi-Speciality Hospital

60 beds · NABH accredited · 4 departments live · FY 2026–27 tariff · Kukatpally, Hyderabad

142

active services

4

departments

10/12

seats used

Hospital profile

H

Sri Venkateshwara Multi-Speciality Hospital

Plot 14, Kukatpally Housing Board Road, Kukatpally, Hyderabad, Telangana — 500072

60 beds NABH accredited Reg. No. TS/HYD/HOSP/2014/0442 GSTIN 36AACCS1234F1Z5

Established

2009

Departments

4 active

Owner

Dr. G. Prakash

Plan

HOS Growth · 12 seats

Branding preview

Used on Rx, bills & letterheads

Primary colours

Full template set in the GST & Templates section.

✦ Setup copilot — I read your tariff Excel Not started

Upload your hospital's tariff sheet — the copilot reads every row, maps services to your master list, and flags anything ambiguous for you to confirm.

Found 142 services in tariff-2026.xlsx — 128 mapped automatically to your service master, 14 need your review before they go live in Billing.

142

rows read

128

auto-mapped

14

need review

Row 37 · "CONSLT GYN REVW"

→Review consultation — Gynecology · ₹300 · GST exempt

Matched on "GYN" + price pattern of your existing review consults

Row 61 · "XRAY KNEE AP/LAT B/L"

→X-Ray Knee (AP & Lateral, both) — Orthopedics · ₹800 · GST exempt

Ambiguous: could be single side ₹450 — suggested bilateral from "B/L"

Row 104 · "DRESSING LRG + MATRL"

→Wound dressing (large) — Nursing · ₹350 + consumables at 12% GST

Split suggested: procedure exempt, materials billed as pharmacy items

Source: tariff-2026.xlsx · uploaded by Anasuya, 09:12 AM · nothing goes live until you approve

Departments

Departments drive OPD queues, tariff grouping and doctor rosters — 4 active

General Medicine

1 doctor

Dr. K. RameshOPD Mon–Sat · 09:00–13:00

Tariff group SVC-01xx

Gynecology

1 doctor

Dr. Sunitha RaoOPD Tue/Thu/Sat

Tariff group SVC-02xx

Orthopedics

1 doctor

Dr. P. Anil KumarOPD Mon/Wed/Fri · OT Tue

Tariff group SVC-03xx

Pediatrics

1 doctor

Dr. Meena IyerOPD Mon–Sat · 04:00–07:00 PM

Tariff group SVC-04xx

Staff & roles

10 of 12 seats

Name	Role	Department	Status
GP Dr. G. Prakash	Owner	All departments	Active now
KR Dr. K. Ramesh	Doctor	General Medicine	Active now
SR Dr. Sunitha Rao	Doctor	Gynecology	Away
AK Dr. P. Anil Kumar	Doctor	Orthopedics	Away
MI Dr. Meena Iyer	Doctor	Pediatrics	Offline
Sw Swapna	Receptionist	Front desk	Active now
RT Ravi Teja	Pharmacist	Pharmacy	Active now
Pr Prasad	Lab tech	Laboratory	Active now
An Anasuya	Billing clerk	Billing	Active now
MG Mary Grace	Nurse	IPD Nursing Station	On duty

Invite staff by phone

New staff get only their workspaces — permissions × plan. A pharmacist never sees billing margins; a receptionist never sees the tariff master.

Tariff master

142 active services · consults GST-exempt · pharmacy 12% · room charges plain

FY 2026–27

Service	Department	Price	GST	Last updated
Consultation — first visit SVC-0101	General Medicine	₹300	Exempt	12 Jul 2026 · Dr. G. Prakash
Consultation — first visit SVC-0201	Gynecology	₹500	Exempt	12 Jul 2026 · Dr. G. Prakash
Consultation — first visit SVC-0401	Pediatrics	₹400	Exempt	03 Jun 2026 · Anasuya
Total Knee Replacement — package PKG-3102 · implant + 5-day stay + physio	Orthopedics	₹1,45,000	Exempt	28 Jun 2026 · Dr. P. Anil Kumar
Semi-private room — per day RM-1002 · 2 beds, AC	IPD	₹1,500	Plain	01 Apr 2026 · Dr. G. Prakash
Complete Blood Count (CBC) LAB-2001	Laboratory	₹350	Exempt	15 May 2026 · Prasad
Antenatal package — 3 visits + scans PKG-2210	Gynecology	₹6,800	Exempt	22 Jun 2026 · Dr. Sunitha Rao
Inj. Ceftriaxone 1g — administered PH-5044 · drug + administration	Pharmacy	₹185	12%	10 Jul 2026 · Ravi Teja

Showing 8 of 142 services

GST configuration

Consultations

Exempt

Procedures & surgeries

Exempt

Laboratory tests

Exempt

Room charges

Plain (no GST)

Pharmacy — Schedule items

12%

Pharmacy — consumables

12%

Default rates apply automatically when a new tariff-master row is created or mapped by the setup copilot. Per-item overrides are available from the Tariff Master section.

Rx pad

Doctor · Consultation · auto-filled with hospital profile

Rx pad

Doctor · Consultation

Bill format

Billing · GST invoice · GSTIN 36AACCS1234F1Z5

Bill header

Billing · GST invoice

Discharge letterhead

IPD · Discharge summary · attending physician signature block

Discharge letterhead

IPD · Discharge

The HOS AI Governance Framework

Five things HOS holds itself to before an AI draft ever reaches your staff — and, against each one, the exact mechanism on this page that proves it, not a claim in a slide deck. Published here because the alternative is asking you to take our word for it.

This is HOS's own published standard, independently reviewable on request — not a third-party certification. We don't hold a SOC 2, ISO 27001, or equivalent yet; that's separate work, not this.

1

Draft, never send, without a named human. Across the AI workforce this month, 96% of actions went out exactly as drafted and 0 were ever sent without a human's approval — not once, not since go-live (see AI Workforce). This page's own audit is stricter still: 81.9% of 4,812 actions were approved with zero edits at all, and every one of the other 2.7% is accounted for in "Where all 131 rejections went," below.

2

Refuse the categories that need a licence — don't just flag them. Three feature categories sit in the AI Regulatory Register as "not enabled — no licence" or "structurally blocked": deterioration forecasting, diagnostic imaging AI, triage that assigns clinical acuity. That's this establishment's copy; across the HOS product roughly forty such refusals exist, each naming the CDSCO SaMD class it would fall under. A register listing only what's switched on would imply nothing was deliberately withheld — this one doesn't.

3

A worker getting worse reports on itself. Continuous evaluation compares every draft against what the human finally approved. When edit distance rises — 1.2 → 3.6 edits over 14 days for the Discharge Drafter, above — the worker keeps drafting and the alert waits for you. Nothing is silently downgraded or hidden.

4

A correction becomes a fix, not a lost signal. Every human edit is logged against a cause and ranked by frequency × time lost, below. Two of this month's four are already fixed and measured; the largest remaining one is marked "no fix yet" rather than pretended away, because deciding which problem-list fields must carry forward is a clinician's call, not a configuration change.

5

The trail can't be quietly edited — by us or anyone. 48,201 blocks, each hash-chained to the one before it. Editing or deleting a single row breaks every hash after it, checked against an anchor published outside this hospital's own database at midnight. "Verify integrity now" in Audit Log proves that without asking you to take our word for it.

AI Ethics Committee

Charter & standing scope — the body that gates AMBER-tier AI features and ratifies the RED-tier refusal list · position as at 18 Jul 2026

Composition — 5 voting seats, 1 non-voting

Chair · Medical Superintendent

Dr. G. Prakash

Clinical safety seat

Dr. K. Ramesh

Nursing & ward-safety seat

Mary Grace

Pharmacy & medication-safety seat

Ravi Teja

Patient-facing seat

Swapna

Independent seat

External legal counsel (retained per review — not an in-house role at 60 beds)

A HOS product/engineering liaison attends every meeting to answer technical questions, as a non-voting resource. The committee decides; the vendor explains — never the other way round.

Scope of review authority

No Reference-tier (AMBER) AI feature goes live for this hospital's staff until the committee has reviewed and approved it here first.
The Medical-device (RED) refusal list is re-ratified every quarter, confirming it is still current — not a one-time sign-off that goes stale.
Every open drift alert joins the committee's next agenda automatically, whether or not it has been fixed by the time the committee meets.
The committee can suspend any AI worker's live status pending review — a governance override distinct from the operational on/off switch in Staff & Roles.

Meets monthly, plus ad hoc within 5 business days of any new AMBER-tier request. Ratifying the RED-tier refusal list is not the same as clearing a feature for use — only a CDSCO licence does that; the committee's power is to keep the refusal list honest and to gate everything below it.

Committee review log

Every review this committee has actually done — approvals, ratifications and what's still open, in the order it happened

Date	Reviewed	Tier	Decision	Signed off by
03 Feb 2026	Lab Summarizer — pre-launch review	Reference	Approved — condition: no interpretive language	Dr. G. Prakash · Dr. K. Ramesh · Mary Grace
03 Feb 2026	Drug interaction & renal-dose alerts — pre-launch review	Reference	Approved	Dr. G. Prakash · Ravi Teja · Dr. K. Ramesh
10 Mar 2026	Early-warning score (ICU) — pre-launch review	Reference	Approved — condition: escalation stays human-initiated	Dr. G. Prakash · Mary Grace · Dr. P. Anil Kumar
01 Jul 2026	Quarterly ratification — RED-tier refusal list (deterioration forecast, diagnostic imaging AI, acuity-assigning triage)	Medical device	Ratified — all three remain not enabled	Dr. G. Prakash · Dr. P. Anil Kumar · external counsel
15 Jul 2026	Discharge Drafter drift alert — Orthopedics (see AI workers table, above)	Productivity	Escalated — re-grounding approval pending	Dr. P. Anil Kumar · review open

This log is the committee's own record, not a marketing summary — the open row above stays open here until the linked drift alert is actually resolved, not before.

Security & Compliance Posture

What's true about this architecture today, and what's still ahead — published so your IT/security reviewer doesn't have to take our word for it · position as at 18 Jul 2026

Control or certification	Status	Evidence / detail
Dedicated database per hospital ADR #2	True today	Every tenant listed in Superadmin runs its own isolated database. No code path in the product queries across two hospitals' clinical data — there is no shared table it could join against.
No clinical data in the control plane ADR #3	True today	The fleet console (Superadmin) sees tenant name, plan, device counts and uptime — never a patient record, note, or result. This is architectural, not a permissions setting someone could misconfigure.
AI regulatory tiering enforced structurally ADR #7	True today	~40 named RED-tier refusals across the product; a 48,201-block hash-chained audit log, independently re-verifiable (see "Verify integrity now" in Audit Log) rather than asserted on trust.
India DPDP Act 2023 obligations	True today	Per-purpose revocable consent, a patient-facing data-rights request tool (Patient Records), and an automated 72-hour breach-notification clock (see Superadmin) — live product mechanisms, not a policy document.
ISO 27001	In progress	Gap assessment stage. Not yet certified. No externally accredited audit has taken place — do not rely on this until a certificate is issued and its scope reviewed.
SOC 2 Type II	Not started	No engagement with an auditor has begun. Listed here so the gap is visible rather than silent.
GDPR-aligned data handling for international / Gulf-market deployments	Not started	Relevant once HOS operates in a market that requires it. Today's architecture (per-hospital data isolation, no cross-tenant sharing, no clinical data in the control plane) is a reasonable starting position for that work, not a substitute for it.

This table is HOS's own transparency disclosure, not a compliance attestation. Rows marked "in progress" or "not started" are not claimed as complete anywhere else in the product — if you find one that is, that's a bug in our copy, not a second true status.

AI actions this month

4,812

▲ 18% vs last month

Approved without edit

81.9%

▲ 2.1 pts

Avg edits per draft

1.06

▼ from 2.1

Rejected outright

2.7%

131 drafts · every one traced

Drift alerts open

1

needs review

✦ Drift alert — Discharge Drafter Draft — awaiting approval

Edit distance on discharge summaries rose from 1.2 → 3.6 edits over the last 14 days, concentrated in Orthopedics. Root cause looks like the new TKR package pathway (added 09 Jul) — the model is still writing the old physiotherapy schedule, and Dr. P. Anil Kumar is correcting it every time.

Recommended: re-ground the Discharge Drafter on the 12 post-09-Jul orthopedic discharges. Estimated 4 min, no downtime. Until then the drafts stay usable — they just cost the surgeon an extra edit.

Detected by continuous evaluation · 80 summaries sampled — every one from the last 14 days · confidence high

AI workers — quality this month

Accuracy is measured against the human's final approved version, not against itself

AI worker	Drafts	Approved as-is	Avg edits	Rejected	14-day trend	Status
AI Scribe Consultation notes	1,842	84.1%	0.9	1.8%	▲ improving	Healthy
Billing Agent Charge capture	1,206	91.6%	0.4	0.7%	▲ improving	Healthy
WhatsApp Assistant Patient messages	974	76.2%	1.3	4.1%	— steady	Healthy
Lab Summarizer Plain-language result summaries	612	71.9%	1.7	5.6%	— steady	Watch
Discharge Drafter Discharge summaries	178	58.4%	3.6	9.0%	▼ degrading	Drift

A worker on Watch still drafts normally. A worker on Drift keeps working too — HOS never silently downgrades an AI behind your back, it tells you and waits.

What your staff keep correcting

The most-repeated human edits this month — two already fixed, two still waiting on you

✦ Ranked by frequency × time lost

Correction	Who	Times	Est. time lost
Discharge summary uses the old TKR physio schedule	Dr. P. Anil Kumar	12	38 min
Telugu summaries too formal — staff rewrite in spoken register	Prasad · Swapna	9	21 min	Fixed 08 Jul
Dressing charges billed as generic consumable instead of the tariff line	Anasuya	7	14 min	Fixed 12 Jul
Rx omits renal-dose note for elderly patients on Metformin	Dr. K. Ramesh	6	11 min

The flywheel

Your AI gets measurably better on your own data

Not a promise — a loop that already ran once this month, end to end, with a number on both sides of the fix.

✦ One loop, end to end — Telugu patient summaries Fixed · improvement measured

A refusal is only worth recording if something happens to it. This is the whole path for the largest group of rejections this month, and it is the same four steps for every row in the table below.

1

Refused by a human. 38 Telugu summaries rejected outright. Prasad's reason, stored as he typed it: “too formal — not how we speak to patients.” Swapna logged the same complaint on WhatsApp replies.

2

Grouped by cause. One cause behind all 38, and it was the register, not the translation — every rejected draft was accurate Telugu that no one speaks. It hit both Telugu-writing workers.

3

Fixed, and attributed. Re-grounded on the spoken-register phrasing in 24 of your own approved summaries — your staff's words, not a vendor style guide. Approved by Dr. G. Prakash on 08 Jul.

4

Measured afterwards. Telugu rejection rate 10.7% in the 8 days before the fix (31 of 289 drafts) against 1.7% in the 10 days after (7 of 402). Same metric, same workers, before and after.

Computed in your database from your own approvals · 01–18 Jul 2026 · this runs continuously and reports to you, not to us

Where all 131 rejections went

Every draft a human refused this month, grouped by the reason they gave

✦ 131 rejections · 2.7% of 4,812 actions

Reason the human recorded	Rejections	Worker	What was done	Effect on accuracy
Telugu summary too formal for a patient	38	Lab Summarizer 22 · WhatsApp 16	Re-grounded 08 Jul · Dr. G. Prakash	10.7% → 1.7%
Note omitted a comorbidity already on the problem list	33	AI Scribe 32 · Discharge Drafter 1	Open — needs a clinician to decide which problem-list fields must always carry forward	No fix yet
Slot offered outside the doctor's OPD session	24	WhatsApp Assistant	Session times added to the booking check, 14 Jul · Dr. G. Prakash	Too early to measure
Discharge summary carried the old TKR physio schedule	11	Discharge Drafter	Bundled into the open drift alert above — awaiting your approval	Not started
Dressing billed as a generic consumable, not the tariff line	9	Billing Agent 8 · AI Scribe 1	Tariff mapping corrected to SVC-2210, 12 Jul · Anasuya	0 in the 6 days since
One-off — no repeated pattern across the month	16	Lab Summarizer 12 · Discharge Drafter 4	Logged, no fix queued — a single refusal is not yet a cause	—

Rejections are grouped by the reason the human recorded, kept in their words. HOS does not rule on whether a refusal was clinically right — the person who refused the draft is the authority. A fix changes how an AI worker drafts; it never alters a clinical record, and it never reverses anybody's rejection. Where a fix is too recent to have produced a measurable change, this table says so rather than claiming a win.

🔒 Tamper-evident log Chain intact

Every entry is hash-chained to the one before it, so removing or editing a single row breaks every hash after it. Not even we can rewrite your history — and you can prove that to an auditor without taking our word for it.

Chain head · block #48,201 · a7f3c9e1b28d40f6…c15e

Audit log

Time	User	Action	Detail
10:14 AM	anasuya	Billing	Billed #INV-2214 · ₹1,240 · Lakshmi Devi
10:02 AM	dr.ramesh	Rx	Approved Rx #RX-8871 · AI draft, 2 edits · Venkatesh Naidu
09:42 AM	swapna	Registration	Registered Mohd. Irfan · token T-14
09:30 AM	dr.prakash	Price change	Updated SVC-3102 Total Knee Replacement package · ₹1,40,000 → ₹1,45,000
09:18 AM	ravi.teja	Billing	Dispensed against #RX-8865 · Tab. Metformin 500mg × 30
09:05 AM	prasad	Lab	Uploaded result #LAB-4471 · CBC · B. Srinu
08:54 AM	dr.prakash	Login	Signed in · new device · Hyderabad, IN
08:40 AM	anasuya	Billing	Refunded #INV-2201 · ₹200 · duplicate lab charge, approved by owner
08:22 AM	swapna	Registration	Registered Padma Sree · token T-06 · antenatal visit
08:05 AM	dr.prakash	Price change	Added service SVC-4402 Nebulization — Pediatrics · ₹150 · Exempt

Every price change, refund and AI approval is logged with who and when. Nothing is deletable — export as CSV for your auditor.

NABH Digital Health maturity

Computed from your live configuration — not a checklist someone maintains by hand

◆ Silver — achieved Gold is 4 objective elements away

Objective elements met71 / 84

Standards covered33 / 38

Chapters complete4 / 8

13 objective elements are unmet across 4 chapters. 4 of those 13 are what Gold requires — the rest are Platinum-tier.

Software certification prerequisites: ABDM M1 ✓ M2 ✓ · M3 pending · WASA ✓ · ISO 27001 in progress, not yet certified

What Gold needs

4 items

Gap	Chapter
Closed-loop medication administration (eMAR barcode scan)	Medication mgmt	R2
Critical result alerting with recorded acknowledgement	Diagnostics	R2
WHO Surgical Safety Checklist, digitally signed	Clinical process	R2
ABDM M3 certification (health-information exchange)	Interoperability	In progress

Every gap here is a HOS roadmap item, so your tier rises as the product ships — you don't run a separate accreditation project. The ABDM Digital Health Incentive Scheme can subsidise this.

AI regulatory register

Which AI features are cleared to run on your patients, and which are not — under the CDSCO Software-as-a-Medical-Device framework

AI feature	Class	Basis	Status here
AI Scribe — consultation notes	Productivity	Transcribes and structures what was said. Not diagnostic — outside the SaMD scope.	Live
Discharge Drafter, Billing Agent, claims & appeals	Productivity	Administrative automation over records that already exist.	Live
Lab Summarizer — plain-language result summaries	Reference	States how each recorded value compares with its reference range and with this patient's own earlier results. It does not interpret the finding, name a condition, or suggest a drug or dose.	Live · human-approved
Contradiction Sentinel — order vs result consistency	Reference	Flags an inconsistency inside your own record. Never phrased as a diagnosis.	Live · human-approved
Drug interaction & renal-dose alerts	Reference	Checks against a licensed drug database. Alerts only — HOS never adjusts a dose itself.	Live · human-approved
Early-warning score (ICU) — NEWS/MEWS from charted vitals	Reference	Displays a published score computed from values your nurses recorded. Presenting recorded data is not prediction.	Live · escalation by a human
Deterioration forecast — predicting a future event	Medical device	Predicting what will happen next is a Class C SaMD. Needs a CDSCO licence, ICMR ethics review and separate patient consent.	Not enabled — roadmap
Diagnostic imaging AI	Medical device	Class C/D. HOS will integrate a CDSCO-cleared vendor rather than build one.	Not enabled — partner
Acuity-assigning triage	Medical device	Assigning clinical acuity is a regulated decision. HOS orders a queue; it does not rank clinical urgency for you.	Not enabled

We publish this because the alternative is you finding out from a regulator. Anything marked Medical device stays off until it is licensed — no hidden beta, no "clinical decision support" euphemism. Every Reference feature keeps a human decisive and is recorded in the tamper-evident audit log.

DPDP consent & data rights

Digital Personal Data Protection Act 2023 — consent, erasure and breach obligations

No open requests

Obligation	How HOS meets it	Status
Specific, revocable consent	Per-purpose consent artefacts; ABDM Consent Manager compatible; patient can revoke from their own app	Met
Right to erasure	Patient data-request workflow; medico-legal retention honoured and explained rather than silently overriding the request	Met
72-hour breach notification	Automated detection → owner + DPO alert → templated Board notification	Met
Purpose limitation & minimisation	Workspace permissions bound to role; break-glass access logged separately and reviewably	Met
Significant Data Fiduciary duties (DPO, DPIA, audit)	Applies above scale thresholds — HOS supplies the audit evidence; appointing a DPO is the hospital's duty	Your action

Start a training session

Pick a role — the trainee sees exactly that role's workspaces, on fake patients

Role

Trainee name

Sandbox contents

Synthetic

Synthetic patients40

Beds in fake ward54

Tariff lines (yours)142

Scripted scenarios18

Scenarios include a STAT lab order, an ICU deterioration, a disputed bill, a drug substitution and an expired-batch catch — the situations that actually go wrong on a real floor.

Staff training record

Who has been signed off on which workspace — the sheet an NABH assessor asks for

Staff	Role	Scenarios passed	Score	Signed off	Status
Swapna	Reception / OPD	18 / 18	96%	02 Jul 2026 · Dr. G. Prakash	Cleared
Mary Grace	Nurse · Ward & ICU	18 / 18	92%	05 Jul 2026 · Dr. G. Prakash	Cleared
Anasuya	Billing	16 / 18	88%	08 Jul 2026 · Dr. G. Prakash	Cleared
Ravi Teja	Pharmacist	14 / 18	79%	—	4 scenarios left
Lavanya joins 22 Jul	Reception / OPD	0 / 18	—	—	Not started