Applied AI for doctors, one clinical task at a time.
FOR STATES, MEDICAL COLLEGES, HOSPITALS & CONFERENCES
This is no longer a question of whether AI enters your wards, your classrooms and your community health programmes. It is already there, arriving through personal phones rather than through any policy you wrote.
What has not arrived is the habit of verifying what it returns — and the evidence below shows that habit is not optional, is not conferred by seniority, and does not appear on its own.
GreyBrain trains the verification habit, at the scale you need it, on a curriculum built the way medicine is taught. Free to start; quoted at cohort scale. Everything on this page is sourced, and every source carries its own limitation.
WHAT THE EVIDENCE SAYS HAPPENS
These are measured effects from controlled studies, not predictions. Each was produced with a model that was sometimes wrong — which is the only kind that exists.
Clinician diagnostic accuracy fell this far when the AI they were shown was systematically biased. Showing them the model's explanation did not repair it.
Jabbour et al., JAMA — randomised clinical vignette study, 457 clinicians, 2023
Limitation: Clinical vignettes, not live patients; US hospitalists, nurse practitioners and physician assistants.
Physicians who had already completed a 20-hour AI-literacy course still lost this much diagnostic accuracy when the model's advice contained errors.
Qazi et al. — single-blind randomised trial, 44 physicians, Pakistan, 2025
Limitation: 44 participants, clinical vignettes, one country. It shows training alone is not a cure.
Mammography accuracy among the most experienced radiologists when the AI suggested the wrong category. Inexperienced readers fell from 80% to 20%.
Dratsch et al., Radiology — prospective reader study, 27 radiologists, 2023
Limitation: A purported AI in an experimental setting, 50 mammograms, one modality.
Endoscopists detected fewer adenomas without AI after a period of routine AI use than they had before it — the skill itself eroded.
Heudel et al., ESMO Real World Data and Digital Oncology — scoping review, 2026
Limitation: A narrative review; the authors state the empirical base for deskilling is still small.
WHO WE BUILD THIS FOR
A Health Secretary and a conference convenor are not buying the same thing. Find the one that describes your week.
MEDICAL & NURSING COLLEGESYour students are already using it for assignments and your residents for case presentations. Nobody has taught either group how to check what it gives back, and nothing in the curriculum requires it.
AI literacy is absent from Indian undergraduate medical curricula, while CBME asks for digital preparedness.
STATE & NATIONAL HEALTH DEPARTMENTSASHAs, ANMs and community health officers are being handed digital tools faster than anyone is being trained to use them, and the competence gap is measurable before it is visible.
A framework aligned to ABDM and Mission Karmayogi measured ASHA digital competency at 39.9%, with no ASHA scoring above 60%.
HOSPITALS & HEALTH SYSTEMSClinicians in your hospital are pasting into chatbots today. There is no policy, no record of who was taught what, and no shared standard — and the accountability for what follows sits with the institution.
ICMR's 2023 guidelines already name institutions and ethics committees among those accountable for AI use.
CONFERENCES & PROFESSIONAL BODIESYou have one slot and a full hall, and you want the delegates to leave having done something rather than having watched a slide deck about the future of medicine.
Asked how they want to learn this, 63.1% of Indian medical students chose hands-on workshops.
Illustrations are AI-generated and do not depict real people, institutions or programme participants.
THE SIZE OF THE GAP
28
of 194 countries
offer pre-service digital health education to their health workforce; 29 offer in-service training.
WHO, Competency frameworks and standards for digital health: a landscape analysis, 2025
Limitation: Counts whether a programme exists, not whether it is good or how many it reaches.
1 million
ASHAs in India
the largest community health workforce in the world, and one whose reach is limited by knowledge and skills rather than numbers.
Panda et al., Human Resources for Health, 2024
Limitation: The paper evaluates one telementoring programme; the workforce figure is context, not its finding.
91.2%
of medical students
had received no prior AI training, while 87.1% wanted AI in the curriculum and 63.1% asked specifically for hands-on workshops.
Sorte et al., Journal of Education and Health Promotion, 2025
Limitation: One institution (AIIMS Nagpur), 217 students, 36.6% response rate, self-reported.
WHAT TRAINING ACHIEVES — AND WHERE THE EVIDENCE STOPS
The systematic reviews are clear about how far the evidence goes, and we would rather you read that from us than discover it later.
AI training for health workers reliably improves reactions, attitudes, knowledge and behaviour — levels 1 to 3 of the Kirkpatrick-Barr hierarchy. Across 27 studies, none reached level 4: organisational change or demonstrated patient benefit.
Awareness of AI among nursing groups pools at 73% and positive attitudes at 71%, but the review finds a lack of institutional readiness: programmes without standardised curricula, faculty training or infrastructure. Individual enthusiasm has outrun institutional capacity.
Structured digital training consistently raises community health workers' competence. Sustaining it requires institutional embedding, tiered training and governance — without those, initiatives remain fragmented pilots.
A single digital capacity-building programme trained 42,192 medical officers, community health officers and frontline workers across nine Indian states, with statistically significant knowledge, attitude and practice gains in multiple cadres.
WHAT YOU ARE ALREADY ANSWERABLE TO
India's first. Ten principles including accountability, data privacy, validity and fairness — and it names clinicians, institutions and ethics committees among the stakeholders responsible.
Read the source ↗Sets out uses across research, education and administration alongside risks from inaccurate output and automation bias.
Read the source ↗GreyBrain is not a regulator and confers no regulatory status. These documents are cited because they already apply to you, not because they endorse us — they do not.
HOW IT IS DELIVERED
Put a batch, a department or a district through the free foundations before you commit anything. Two rungs carry a published price; above them, scope, cohort size and assessment shape the cost, so we quote rather than list.
F0–F6, the whole textbook, free to read. Put a batch, a department or a district through it before you commit a rupee.
Recorded lessons with worked examples and exercises, self-paced, for anyone who wants to go further on their own.
One talk-length slot for a full hall. Everyone leaves with the practice for their own role.
One day, hands on. Every participant makes something in their own subject, checks it against a source, and defends what they changed.
Foundations for everyone, scoped capstone work, and an agreed assessment and review on the IIHMR Bangalore certificate pathway.
WHO TEACHES IT
7 named faculty — practising physicians, biomedical engineers and machine-learning researchers, including faculty from IIHMR Bangalore and IISc Bangalore. Every credential is checkable, and the names are on the front page.
Tell us the shape you have in mind — a session, a workshop, a batch, a district — and we come back with scope, dates and a written quotation. Scope, cohort size and assessment are agreed before a programme begins, and those are what shape the cost, so there is no standard price to publish above the entry course.
Certificate, capstone and review requirements depend on the programme and are agreed with your institution in advance. A course certificate recognises learning; it does not confer a clinical licence, specialist accreditation, NMC recognition or CME credit, and no improvement in examination results or patient outcomes is promised.