# Medows — Full Reference

## What is Medows?

Medows is an AI-assisted clinical workspace for the doctor on rounds. It is a personal workspace — not an electronic health record (EHR) — that helps a single ward-based doctor hold their patient list, capture vitals between beds, ask a context-aware AI for clinical guidance grounded in the actual patient on screen, and hand the shift off to the next doctor over WhatsApp.

The product was designed alongside junior doctors and residents (Post-Graduate Trainees, or PGTs) in Indian hospitals — including AIIMS Bhubaneswar and RG Kar Medical College — where a single doctor on admission day routinely carries 20 to 30 new patients in twelve hours. The shape of the product reflects that environment: minimal taps, fast capture, and an AI that respects the doctor's time rather than producing textbook prose.

## Blog

Medows publishes essays on clinical AI, ward workflow, residency life, and case studies — written by the team and by partner doctors. The blog lives at https://www.medows.ai/blog.

### Recent posts
- [AMIE in Clinic: 98 Patients, 90% Match](https://www.medows.ai/blog/amie-in-clinic-98-patients-90-match) — A Lancet study put Google's AMIE chatbot in a real clinic with 98 patients. What the 90% match does and does not tell a doctor on rounds.
- [AI Scribes Cost Pre-Meds a Window Into Medicine](https://www.medows.ai/blog/ai-scribes-cost-pre-meds-a-window-into-medicine) — U.Va. Health cut student scribe shifts as ambient AI expanded. The note got easier. The apprenticeship got thinner. What that means for training.
- [AI Scribes: 69% Less Charting, One New Habit](https://www.medows.ai/blog/ai-scribes-69-less-charting-one-new-habit) — A Toronto study found AI scribes cut documentation time 69.1% in simulations. Some doctors added scratch pads to keep thinking.
- [AI Coding Tools: The $942M Insurer Claim](https://www.medows.ai/blog/ai-coding-tools-the-942m-insurer-claim) — Blue Cross says hospital AI coding added $942M with no change in care. Hospitals disagree. What it means for the doctor signing the chart.
- [Ambient AI Just Outgrew the Scribe](https://www.medows.ai/blog/ambient-ai-just-outgrew-the-scribe) — Heidi raised $340M to push ambient AI past note-taking into real clinical work. Here's what that shift means for the ward.
- [Rural Hospitals Get Help Vetting AI Tools](https://www.medows.ai/blog/rural-hospitals-get-help-vetting-ai-tools) — UNC Health and Duke Health launched a $4.4M network so small hospitals without data science teams can safely vet AI tools.
- [AI Discharge Notes: 32% More Editing Time](https://www.medows.ai/blog/ai-discharge-notes-32-more-editing-time) — A hospital study found AI-drafted discharge summaries took 32% longer to edit than notes written from scratch, though doctors felt faster.
- [AI Pushed Doctors Onto the Picket Line](https://www.medows.ai/blog/ai-pushed-doctors-onto-the-picket-line) — 150 Allina Health doctors struck partly over AI nudging billing codes and clinical calls without physician input.
- [AI Scribes Reach 986K VA Visits](https://www.medows.ai/blog/ai-scribes-reach-986k-va-visits) — The VA scaled ambient AI scribes to 986,000+ visits since October 2025, running two vendors at once, not one workspace.
- [Stroke AI: One More Catch Per 18 Scans](https://www.medows.ai/blog/stroke-ai-one-more-catch-per-18-scans) — A multinational study found AI support raised stroke-detection sensitivity from 46.6% to 63.7% on plain CT, catching one more occlusion per 18 scans.
- [Autonomous AI Wins Benchmarks, Not Wards](https://www.medows.ai/blog/autonomous-ai-wins-benchmarks-not-wards) — Dueling STAT essays debate if autonomous AI beats doctors. The benchmarks are real, but accountability on the ward isn't in the data.
- [AI Scribes Cut Notes, Not ER Wait Times](https://www.medows.ai/blog/ai-scribes-cut-notes-not-er-wait-times) — AI scribes save 1.6 minutes per note in the ER, but a new study found wait times and patient throughput unchanged. The bottleneck isn't documentation.
- [FDA TEMPO Pilot Skips AI Approval](https://www.medows.ai/blog/fda-tempo-pilot-skips-ai-approval) — Four AI tools now reach Medicare patients before FDA approval, under a pilot trading review for real-world data.
- [OpenEvidence Hits 100% on MedQA. So What?](https://www.medows.ai/blog/openevidence-hits-100-on-medqa-so-what) — OpenEvidence's new AI models score a perfect 100% on MedQA. That solves a test, not the doctor's real problem on rounds.
- [Discharge Delays: AI Just Bought the Fix](https://www.medows.ai/blog/discharge-delays-ai-just-bought-the-fix) — LeanTaaS bought Aidin to close the gap between a ready-to-leave patient and an empty bed. The math behind hospital discharge delays.
- [Hospitals Deploy AI Faster Than They Test It](https://www.medows.ai/blog/hospitals-deploy-ai-faster-than-they-test-it) — 92% of health systems deployed AI. Under half have a real test environment for it. A new UPMC/KLAS report maps the gap.
- [When Hospital AI Overrules the Nurse](https://www.medows.ai/blog/when-hospital-ai-overrules-the-nurse) — Nurses protested Palantir's hospital AI over staffing and care calls, as a California bill would keep the final say with clinicians.
- [Sodium is a neurological observation](https://www.medows.ai/blog/sodium-is-a-neurological-observation) — Sodium climbed for a full day after correction started. That's the point where it stopped being a fluids problem.
- [Ambient AI Scaled. Consent Didn't.](https://www.medows.ai/blog/ambient-ai-scaled-consent-didn-t) — Ambient AI scribes went from pilot to standard practice. Patient consent and data governance did not scale with them.
- [FDA Cleared 1,357 AI Devices. Only 3 Proven.](https://www.medows.ai/blog/fda-cleared-1-357-ai-devices-only-3-proven) — FDA cleared 1,357 AI medical devices. A new study found only 3 were tested on whether patients actually got better.
- [The AI Scribe Rollout That Actually Stuck](https://www.medows.ai/blog/the-ai-scribe-rollout-that-actually-stuck) — Cleveland Clinic put an AI scribe in front of 4,000+ clinicians in 16 weeks. The lesson wasn't the model. It was the rollout.
- [Diagnostic AI's 95% Accuracy Problem](https://www.medows.ai/blog/diagnostic-ai-s-95-accuracy-problem) — New research: data leakage inflates AI schizophrenia-detection accuracy by up to 30%, exposing why near-perfect diagnostic AI claims deserve real scrutiny.
- [The CSF that did not fit](https://www.medows.ai/blog/the-csf-that-did-not-fit) — A CSF report with 920 cells and 92% polymorphs and a glucose and protein that both read normal. What antibiotics can and can't explain before the next lumbar puncture.
- [FDA Will Grade AI Like It Grades Doctors](https://www.medows.ai/blog/fda-will-grade-ai-like-it-grades-doctors) — The FDA wants feedback on grading generative AI medical devices the way it evaluates physicians: by competency, not just code.
- [Diagnostic AI Finally Gets a Rulebook](https://www.medows.ai/blog/diagnostic-ai-finally-gets-a-rulebook) — Twelve US health systems and Aidoc formed a consortium to govern diagnostic AI, betting that shared standards beat point solutions.
- [Clinical AI's Real Danger: What It Skips](https://www.medows.ai/blog/clinical-ai-s-real-danger-what-it-skips) — The largest clinical AI safety study yet found most severe errors are omissions, not wrong answers, and purpose-built tools beat general chatbots.
- [Hospital AI Pilots Work. Scaling Doesn't.](https://www.medows.ai/blog/hospital-ai-pilots-work-scaling-doesn-t) — A new survey finds EHR integration, not clinician trust, is the real reason clinical AI pilots stall before they scale.
- [Before AI Hits the Ward, Doctors Test It](https://www.medows.ai/blog/before-ai-hits-the-ward-doctors-test-it) — UnityPoint has 60 physicians pressure-test AI tools before rollout, not after. What that says about verifiable clinical AI.
- [Clinical AI Explanations Can Mislead Beginners](https://www.medows.ai/blog/clinical-ai-explanations-can-mislead-beginners) — A Nature Medicine study finds AI diagnostic explanations boost novice confidence even when wrong, while trained clinicians barely benefit.
- [AI Scribe or Medical Device? The UK Decides](https://www.medows.ai/blog/ai-scribe-or-medical-device-the-uk-decides) — The MHRA just clarified when an AI scribe counts as a medical device in the NHS, and what it means for the doctors relying on one.
- [Which AI Can a Doctor Actually Trust?](https://www.medows.ai/blog/which-ai-can-a-doctor-actually-trust) — A new leaderboard scores 18 AI models on medical safety. The general chatbots doctors reach for aren't the safest ones in the room.
- [AI Finds Bias in Half of Pregnancy Notes](https://www.medows.ai/blog/ai-finds-bias-in-half-of-pregnancy-notes) — A new AI analysis of 640,000+ obstetric notes found stigmatizing language in 47% of pregnancies, worse for Black and less-educated patients.
- [AI fixed the clinical note. Not the ward round.](https://www.medows.ai/blog/ai-fixed-the-clinical-note-not-the-ward-round) — Ambient AI scribes cut documentation time and burnout, and that is real. But the note was never the hard part of a shift. The ward round is the frontier.
- [India moved to ban paraquat. The casualty doctor's problem does not end today.](https://www.medows.ai/blog/india-moved-to-ban-paraquat-the-casualty-doctor-s-problem-does-not-end-today) — India issued a draft order to ban paraquat on 13 July. It is not final yet, and the patient who walks into casualty tonight still has no antidote.
- [Medows is launching on Product Hunt](https://www.medows.ai/blog/medows-is-launching-on-product-hunt) — Medows is launching on Product Hunt on July 16: the first AI clinical workspace built for the individual doctor. Walk a real shift in our no-sign-up demo.
- [Why Medows Is the First AI Clinical Workspace](https://www.medows.ai/blog/why-medows-is-the-first-ai-clinical-workspace) — Most clinical AI solves one task. Medows is an AI clinical workspace built for the doctor on rounds — holds your list, answers grounded in the patient on screen, handover over WhatsApp.
- [Why handover defaults to WhatsApp, not email](https://www.medows.ai/blog/why-handover-defaults-to-whatsapp-not-email) — We send handover over WhatsApp because the doctor on the receiving end already has it open. The whole story what it costs us, what we considered, and what would change our mind.
- [Why Doctors Need a Workspace of Their Own](https://www.medows.ai/blog/why-doctors-need-a-workspace-of-their-own) — Five studies on how junior doctors actually spend their time, what handover failure costs in serious adverse events, and why the ward is the only part of modern healthcare still running on paper and memory.
- [Five Cases on the Ward: What an AI Workspace Looks Like in Real Use](https://www.medows.ai/blog/five-cases-on-the-ward-what-an-ai-workspace-looks-like-in-real-use) — Hyperkalemia at 2 a.m., quiet sepsis on a Sunday morning, a drug interaction caught at order time, and a handover that didn't drop the pending lab - four composite cases from Indian wards showing how Medows actually works.
- [The Slow Goodbye: Why Paraquat Poisoning Kills Long After the Bottle Is Empty](https://www.medows.ai/blog/the-slow-goodbye-why-paraquat-poisoning-kills-long-after-the-bottle-is-empty) — A weed-killer with no antidote, a patient who walks in talking, and a death that arrives weeks later as the lungs quietly turn to scar.
- [Why we built Medows as a workspace, not an EHR](https://www.medows.ai/blog/welcome) — Most clinical AI products land in the EHR. We didn't. Here's why the doctor on rounds needs a workspace of their own, and what we learned building one alongside residents at AIIMS Bhubaneswar and RG Kar.
- [What "Ask Medows" Actually Means](https://www.medows.ai/blog/what-ask-medows-actually-means) — Three things "Ask Medows" is — context retrieval, source-grounded generation, in-patient interface. Three things it explicitly is not.
- [The Eighteen-Minute Handover](https://www.medows.ai/blog/the-eighteen-minute-handover) — Forty seconds per patient is enough for the urgent thing. Everything else gets dropped. What it means to design a handover that respects that math.
- [Why Doctors Don't Trust AI (Yet)](https://www.medows.ai/blog/why-doctors-do-not-trust-ai-yet) — Three reasons doctors don't trust clinical AI yet, in order of how often I hear them. None of them are about the AI being smart enough.
- [The Casualty at 11 p.m.](https://www.medows.ai/blog/the-casualty-at-eleven-pm) — Forty significant clinical decisions in an hour. Six hours of sleep. The pitch every AI clinical tool gets wrong.
- [The CCF Case Nobody Escalated](https://www.medows.ai/blog/the-ccf-case-nobody-escalated) — A 62-year-old man in decompensated CCF by 10 p.m. The 6 p.m. obs were the inflection point. Nobody saw them as a slope.
- [What Changed When We Standardised Handover](https://www.medows.ai/blog/what-changed-when-we-standardised-handover) — I-PASS cut preventable adverse events by 30%. The follow-up showed adherence to the bundle decayed by year-end. What that tells us about tool design.
- [The Burnout Number Indian Residency Does Not Talk About](https://www.medows.ai/blog/the-burnout-number-indian-residency-does-not-talk-about) — 40-60% burnout in published Indian resident cohorts. The hours are the visible part. The cognitive-load-per-hour is the part we don't talk about.
- [Context-Aware AI vs. ChatGPT for Clinical Reasoning](https://www.medows.ai/blog/context-aware-ai-vs-chatgpt-for-clinical-reasoning) — A junior doctor typing into ChatGPT on his phone gets a plausible answer. Why that answer is slightly useless for the patient three feet away.
- [The Morning Round in 90 Minutes](https://www.medows.ai/blog/the-morning-round-in-ninety-minutes) — 90 minutes on paper. 1.8 minutes per patient in practice. What that actually means for the round you think you're doing.
- [Hyperthyroid Storm — The Case I Almost Missed](https://www.medows.ai/blog/hyperthyroid-storm-the-case-i-almost-missed) — A 28-year-old with fever, tachycardia, and hypotension. Sepsis was the obvious differential. The TSH was the answer.
- [What AIIMS Taught Me About Residency Software](https://www.medows.ai/blog/what-aiims-taught-me-about-residency-software) — Three things I didn't expect from a week shadowing a PGT at AIIMS Bhubaneswar. The paper list, WhatsApp, and the most stressful eight minutes of the day.
- [The DKA That Survived Because of One Note](https://www.medows.ai/blog/the-dka-that-survived-because-of-one-note) — A 45-year-old man, DKA, an 11 p.m. ABG, and a handover line that meant the receiving doctor caught a worsening at 3 a.m. instead of 8.
- [The Stroke Time Window Under Heavy Load](https://www.medows.ai/blog/the-stroke-time-window-under-heavy-load) — A 93-minute window. 140 minutes to needle. About 1.9 million neurons lost per minute. Where the time actually went.
- [What 5.9 Hours of EHR Actually Looks Like](https://www.medows.ai/blog/what-five-point-nine-hours-of-ehr-actually-looks-like) — The Arndt 2017 EHR-event-log study, broken down by activity. Half an hour a day goes to navigation alone.
- [The Polypharmacy Trap in the Geriatric Ward](https://www.medows.ai/blog/the-polypharmacy-trap-in-the-geriatric-ward) — An elderly man on warfarin, a new antibiotic, and the kind of interaction every pharmacy software catches — but only after the order is dispensed.
- [Pediatric Dosing in the Busy OPD](https://www.medows.ai/blog/pediatric-dosing-in-the-busy-opd) — A 14 kg child, the wrong dose, and the dose-checking literature that says it's the system, not the resident.
- [Why We Still Hand Over on Paper](https://www.medows.ai/blog/why-we-still-hand-over-on-paper) — The largest published reduction in preventable adverse events in residency literature came from a structured handover. Most of us still do it on a folded sheet.
- [The Twenty-Eight Patient List Problem](https://www.medows.ai/blog/the-twenty-eight-patient-list-problem) — Miller's 1956 paper put working memory at 7 ± 2 items. The average Indian PGT carries 28 on their ward list. What that gap actually costs.
- [Sepsis That Walks In Talking](https://www.medows.ai/blog/sepsis-that-walks-in-talking) — She walked into casualty post-op, day 2, with vitals that looked fine on every check. The slope was the diagnosis.
- [The Interrupted Shift](https://www.medows.ai/blog/the-interrupted-shift) — A K+ of 5.9 that should have been caught at the bedside, the 9-minute interruption interval published in 2010, and why every ward tool we have assumes the doctor's working memory is intact.

## How Medows is different from other clinical AI tools

Most clinical AI products in the market fall into one of three categories:

1. **EHR-integrated documentation tools** (Abridge, Nuance DAX, Suki) — these reduce documentation burden during outpatient consultations, mainly in the US, by transcribing the visit and writing the SOAP note. They require an EHR integration and are sold to hospitals.
2. **Open-web clinical search engines** (OpenEvidence, Glass Health, UpToDate) — these answer general clinical questions but have no patient context. The doctor must restate the case each time.
3. **Hospital-floor workspaces** — historically this has been paper and WhatsApp groups. There is no purpose-built workspace for the doctor walking a ward.

Medows is in the third category. It is a personal workspace that lives on the doctor's phone (or tablet), holds the patient list for the shift, and uses retrieval-augmented generation (RAG) so the AI's answer is grounded in the specific patient the doctor is looking at — vitals, recent labs, active orders, tasks completed — without the doctor having to re-state context.

OpenEvidence withdrew from the UK and EU on 30 April 2026, leaving global doctors without a context-aware AI workspace. Medows is built global-first.

## Product capabilities in detail

### Ward list (Track)
- Every patient on one card with vitals, complaints, active orders, and lab values
- Color-coded by acuity (red = critical, amber = watching, green = stable)
- Drag-and-drop reordering and acuity sorting
- Single-tap to open a patient

### Ask Medows (Consult)
- An AI consult anchored to the patient currently on screen
- The AI sees vitals, recent labs, completed tasks, and active orders before the doctor asks
- Returns differentials, dosing, and next steps — not textbook prose
- Citations to clinical guidelines (AIIMS, ICMR, NICE, UpToDate-equivalent sources)
- Designed to be verifiable — every claim sits next to the patient context it used

### Vitals tracking
- Quick capture between beds (BP, HR, SpO₂, temp, RR)
- Trend visualisation over the last 6 hours per patient
- Out-of-range values flagged in the ward list

### Handover (Hand off)
- Two lines per patient — what matters now, what's pending
- Auto-generated from the shift's context
- Sent to the next doctor over WhatsApp in one tap, or copied as plain text

## Who Medows is for

**Primary users:**
- Junior doctors and residents (PGTs) on ward duty
- Clinicians covering night shifts and admission days
- Doctors in government and busy private hospitals where patient load per doctor is high
- Solo practitioners covering multiple beds simultaneously

**Geography:**
- Built global-first
- Initial pilots in India (Indian government medical colleges)
- Expanding to UK, EU, and global English-speaking markets where the OpenEvidence withdrawal left a gap

## What Medows is not

- Not an EHR and not a substitute for hospital documentation
- Not a multi-tenant hospital admin tool — there are no shared workspaces, admin dashboards, or role-based access controls
- Not a billing or scheduling product
- Not a diagnostic medical device — it is decision support, and every suggestion is verifiable against the source guideline

## Data and privacy

- Patient data is scoped to a single doctor's sign-in via row-level security on Postgres (Supabase)
- No shared workspaces across doctors
- No data is sold or shared with third parties
- Doctors are responsible for entering only the patient information they are clinically authorised to handle

See https://www.medows.ai/privacy for the full privacy policy.

## Pricing

Free for individual clinicians while the product is still finding its shape. There is no paid tier yet. Future paid tiers, if any, will be for team or hospital-grade features — the personal workspace will remain accessible to individual doctors.

## Technical context

- Web application, mobile-first responsive UI
- Built on Next.js, Supabase (Postgres), and large language models for the AI consult
- Sign-in via Google
- No native mobile app yet — runs in mobile browser

## Team

Medows is built by doctors and engineers — initially in India — with active design partnership from junior doctors at AIIMS Bhubaneswar, RG Kar Medical College, and other Indian government medical institutions.

## Common questions

**Is Medows an EHR?**
No. Medows is a personal workspace for the doctor on rounds. It does not replace hospital documentation.

**Is Medows free?**
Yes, for individual clinicians. There is no paywall.

**Where can I use Medows?**
Medows is web-based and works globally on mobile and desktop browsers. Sign in at https://www.medows.ai.

**How is Medows different from OpenEvidence?**
OpenEvidence is an open-web clinical search engine without patient context. Medows is a personal workspace that holds the patient list and answers questions grounded in the specific patient on screen. OpenEvidence also withdrew from the UK and EU in April 2026.

**Does Medows integrate with hospital EHRs?**
No. Medows is intentionally a personal workspace and does not require EHR integration.

**Is Medows a medical device?**
No. Medows is decision support for clinicians. Every AI suggestion is verifiable against the source guideline cited.

**Who built Medows?**
Medows was built by doctors and engineers, with design partnership from junior doctors at AIIMS Bhubaneswar and RG Kar Medical College.

**How do I contact Medows?**
Founder and partnership enquiries: alapan@medows.ai or alapanx@gmail.com. Website: https://www.medows.ai.