Skip to content
PenzaıHUMAN TECH

The GuideItaly

Transcribing Medical Visits with AI: Faster Notes, Patient Data Kept Safe (and the Secret of the Pecia)

AI visit transcription for doctors: how it works, what Italian law 132/2025 says, which privacy guarantees to demand and how to adopt it safely.

The Professor at PenzaiReviewed by Agostino Cacciola5 min read

Medico in camice davanti a un computer nello studio mentre scrive il referto dopo una visita

Can you use AI to transcribe visits and draft the medical note? Yes, under three conditions: the doctor rereads and signs every text, the patient is informed, and the vendor handles health data with written guarantees (contract, EU servers, no training on your data). If one of the three is missing, wait.

A common scenario: a practice with one doctor, six visits before lunch, and an hour of notes to write in the evening. Transcription tools listen to the conversation, turn it into text and propose a draft note. The time saving is real. The risk is that nobody rereads.

What it can do (and where it gets it wrong)

FunctionWhat it doesWhere it can fail
DictationThe doctor speaks, text appearsDrug names, dosages, abbreviations
Visit transcriptionListens to the conversation and writes itOverlapping voices, dialect, noise
Structured summarySplits into history, exam, planMay omit or misattribute a detail
Draft reportPrepares text in your formatMay "complete" what nobody said

The last point is the sneakiest: a generative system tends to write plausible sentences, and a plausible sentence in a patient record is not a true sentence.

The secret of the pecia: why you always reread

Medieval universities invented a system that looks modern. Students needed texts, but copying a whole volume took months. So the stationarii, booksellers licensed by the university, kept an official exemplar divided into numbered sections, the pecie, and rented them out piece by piece: several scribes worked on the same text in parallel. According to the Hill Museum & Manuscript Library's description, the system began in the thirteenth century and relied on approved exemplars. The best-known evidence places it in major universities such as Bologna, with a documented trace as early as 1228 in a contract from Vercelli.

And you will say: what about doctors? The point is trust. Speed was not enough: the copy had to be faithful to the exemplar, which is why the exemplar was approved by the university and those who broke the rules risked penalties. The faster the copy ran, the more control mattered.

Well, your practice today is in that same workshop. AI is a very fast scribe, and a fast scribe who makes mistakes makes them faster. The approved exemplar is you: the draft is worth something once you have reread it.

What the law says

Italian law 132/2025, article 7, sets three useful points for a medical practice:

  • The decision stays with the doctor. AI supports prevention, diagnosis and treatment, but the decision always rests with the medical professional.
  • Patients have the right to be informed when AI technologies are used.
  • Systems must be reliable and checked over time to reduce the risk of errors.

On privacy, health data is a special category (art. 9 of the GDPR). The Italian data protection authority's ruling of 7 March 2019 clarifies that for processing necessary for care the doctor does not ask for consent, while anything beyond that needs a different legal basis. An audio recording sent to an external service is additional processing: assess it with your privacy consultant, including whether an impact assessment (art. 35 GDPR) is needed.

What to ask the vendor before signing

  1. A data processor contract (art. 28 GDPR), with the list of sub-processors.
  2. Where the data is: EU servers or adequate safeguards for transfers.
  3. No training of models on your recordings, written in the contract.
  4. Deletion of the audio after transcription, with stated timeframes.
  5. Integration with your practice software, so you don't copy and paste by hand.
  6. Costs: ask for a written quote and compare at least two offers; price lists change often and there is no "standard" price.

How to introduce it in the practice

  1. Start with report dictation, which is less invasive than recording the conversation.
  2. Test with non-sensitive visits or dummy data for a week.
  3. Inform patients with a sign and a sentence at the start of the visit; anyone who doesn't want the recording must be able to say no without consequences.
  4. Reread every draft and note the errors: after two weeks you will know where the system errs most.
  5. Write a one-page internal rule: what is recorded, who can use it, where it is stored.

Mistakes to avoid

  • Using a generic chatbot, with no data processing agreement, to summarize notes with names.
  • Signing a draft read in a hurry.
  • Recording without telling the patient.
  • Leaving audio files on the doctor's phone.

What to do on Monday morning

  1. Measure the time you spend on notes and reports for a week, to know what a tool is really worth.
  2. Ask two vendors for the contract, data location and training policy before watching a demo.
  3. Write the information sign for the waiting room and rehearse the sentence to say at the start of the visit.

Further reading

This guide connects to the medical practice front desk with AI and the dossier on AI for professional practices. To understand who answers when something goes wrong, read liability for AI errors in business.

If you want to work out which tool makes sense in your practice and how to set it up prudently, see how we work with doctors and book a free introductory call.

Was this useful?
Buy us a coffee

Penzai Magazine

The day's stories, every evening in your inbox

Briefings, guides and analyses from the Penzai Professor: a preview and link for each article, one email a day. Free gift: the AI checklist for SMEs. Unsubscribe anytime.

Prefer Telegram? Get every article as soon as it is out Follow on Telegram

Comments

No comments yet. Be the first.

All articles
Portrait of the smiling Penzai mascot

Let's talk about your next project.

A focused consultation to understand how technology can really help you. First we understand your work, then we find the right technology.