Healthcare
Pharmacy
The problem
AI gives one interaction level for three drugs at once or cites a section of the product information that does not exist. A wrong dose, a schedule that does not match the prescription or a missed pharmacovigilance report passes because the text looks professional.
31 courses · 6 specific to this profession
A process from your work
How often does a label's dosing direction come back to you for correction?
- Step 1From the prescription, the tool proposes the dosing direction in plain words for the patient.
- Step 2You compare the proposal with the prescription before the label is printed.
- Step 3The decision rests with the pharmacist who verifies the prescription: a direction that does not match goes back for correction.
- Where the tool goes wrongThe label says “twice a day” and doubles the prescription's daily amount.
- The verification stepCompare the daily amount on prescription and draft; if they differ, the draft is not printed.
More: what was measured and the full check
What changed, as measured
After the AI suggestions came into use, near misses in medication directions (errors the pharmacist caught and sent back) fell by 33%, and edits to accepted suggestions fell by 44.3%.
Who was measured: The prescription workflow of a large online pharmacy in the United States: the technicians who enter directions and the pharmacists who verify them, before and after the AI system went into use.
The caveat: A before-and-after comparison at one company, written by the company; it measures errors and rework, not minutes.
The check, step by step
- Put the draft next to the prescription and compare the fields one by one: amount per dose, times per day, time of day or relation to food, duration.
- Work out the daily amount on both: tablets per dose multiplied by doses per day. If the two numbers differ, the draft is not printed.
- Delete from the draft any instruction that is not in the prescription or in the medicine's package leaflet.
- Before sticking the label on, read it aloud to the patient and ask her to tell you herself how many times a day she will take the medicine.
The course covers prescription verification with AI, through practical cases.
We can help you with this process: “Pharmacy AI: Prescription verification and drug interactions”, 8 hours.
A documented failure pattern
“Take 1 tablet twice a day”
A pharmacist asks an AI tool for a clearly worded supplementary label for an elderly patient. The label goes on the medicine box as part of the information on the medicine's use, which Law no. 105/2014 (Article 3, point 33) lists among a pharmacy's activities.
Source: Nature Medicine, 2024, and 3 more.
Why it is easy to miss
The number the eye checks, "1 tablet", matches the prescription. The error is in the words "twice a day", which are not a number and so do not get compared. The line "Do not exceed the prescribed dose" makes the label look careful, even though the label itself doubles the daily amount. The tool does not make the error every time. The same request can give a correct label today and a wrong one tomorrow, so it is not enough that the tool was right before. In a 2025 study, three of four experts judged ChatGPT's patient instructions factually correct. Only the fourth noticed that some versions had the wrong dosing frequency.
Where the failure pattern is documented: Peer-reviewed studies: Nature Medicine (Amazon Pharmacy, April 2024), Frontiers in Artificial Intelligence (King Saud University, April 2025), British Journal of Clinical Pharmacology (August 2024); World Health Organization guidance (January 2024).Sources: Nature Medicine, 2024; Frontiers in Artificial Intelligence, 2025; British Journal of Clinical Pharmacology, 2024; World Health Organization, 2024.The case above is ours, written for work in Albania; the laws it names were checked against the official text.
What it looks like
The prescription1 tablet, once a day.
The tool's draft"Take 1 tablet twice a day. Do not exceed the prescribed dose."
Translated from the Albanian original.
Specialist courses for this profession
Pharmacy AI (introductory)
Pharmacy AI: Operations foundation
Pharmacy AI: Prescription verification and drug interactions
Pharmacy AI: Regulatory compliance and pharmacovigilance
Pharmacy AI: Patient data and records
Pharmacy AI: Preparation for assessment
What you leave with
The interaction sheet splits the prescription into drug pairs; for each pair it records the pharmacy reference sources checked, at least two, any mismatch, and the prescriber it was passed to. It is filled in before dispensing.
Pharmacy
The catalogue lists all 31 together, with filters for format and time.