Health · News · 🇧🇷 Brazil
A pharmacist at a Roraima health post reviews prescriptions with AI and makes 600 clinical interventions
At the Caracaraí clinic, NoHarm's free platform reads prescriptions from the medical record and flags wrong doses and dangerous interactions in seconds, and the pharmacist decides what to raise with the doctor.
- Profession
- Pharmacist
- Tools
- NoHarm
- Source
- www1.folha.uol.com.br ↗
- Country
- 🇧🇷 Brazil
- Report date
- November 1, 2026
- Verified on
- November 2, 2026
What they did
Samuel Andrade, who coordinates pharmaceutical care in Caracaraí, Roraima, has used NoHarm's platform at the Manoel Luiz Rodrigues clinic since March 2025. The AI pulls prescriptions from the medical record and flags incompatible doses and drug interactions. The tool reached 341 clinics in 20 cities in Brazil's North and Northeast through BNDES's Juntos pela Saúde program, backed by Umane.
How it went
According to the pharmacist, more than 25,000 prescriptions went through the system, 11,600 got a closer review and there were at least 600 interventions to change clinical management. He mentions wrong doses of antiallergy and antiparasitic drugs for children, and says a check that would take over an hour reading package inserts now takes seconds.
Good to know
The AI flags the risk, but the decision stays with the pharmacist and the doctor. The figures come from the user himself, in an interview with Folha.
The prompt
I'm a pharmacist at [type of facility] and want a routine for reviewing prescriptions. For the prescriptions below [paste without patient identifiers, with age and weight], point out doses that may be out of range for the age, relevant interactions and duplicates, explain each alert and tell me what to confirm in the package insert or with the doctor before acting.
Replace what is in [brackets]. Prompt written by us from the account, not a literal copy of the source. Do not paste other people’s personal data (privacy laws such as LGPD and GDPR apply).
Summary in our own words. Results as stated in the original account; we have not tested or audited them. Read the source ↗ Found a mistake?
Similar cases
01Claude agent combines two years of WHOOP data with blood test results
An open project replaced Databricks' Genie agent with a Claude agent and measured both on 30 questions about personal health data.
02The GP form they kept putting off
A London user asked the agent to fill out a long form from their family doctor.
03Unexplained fatigue: agent suggested tests and specialists
Noah described his symptoms; the agent listed blood tests and doctors, and he booked everything.
04Disputing a hospital bill the insurer already paid
The agent found the insurer's statement for a surgery from a year earlier and sent the dispute with the document.
05Filling out insurance paperwork while caring for a newborn
Felix Rieseberg gave Claude the hospital PDFs and let the agent fill in the insurer's website.
06Charting your own blood tests from PDFs
Matt Stockton attached lab reports from different years and Claude built a tool that shows the trends and highlights abnormal results.
07Alexa routines to organize your day (with a therapist's tip)
A journalist set up mood, water and lunch reminders on Alexa based on advice from a psychotherapist.
08A pill dispenser that alerts the caregiver
A guide shows a Raspberry Pi with an agent that releases the dose, takes a photo and alerts the family if it isn't taken.