Monday, 8:15 a.m. The practice phone rings nonstop, there's already a line in the waiting room and the inbox holds thirty requests for repeat prescriptions. You, who studied for years to care for people, spend the first hour of the day working as a switchboard operator. Can artificial intelligence change this picture? Yes, but not in the way many headlines suggest.
Ragusa, 1377: the cure that came from organization
Let's step back in time. It's 1377 in Ragusa, today's Dubrovnik, a maritime republic living off trade across the Mediterranean. The Black Death, which reached Europe thirty years earlier, keeps coming back. Nobody knows what causes it: germs would be discovered centuries later. Yet the city's government makes a decision that would set an example: anyone arriving from affected areas must spend a month in isolation outside the city before entering. A few decades later Venice would extend the period to forty days, and from those quaranta giorni comes the word quarantine. In 1423 Venice would also open its first lazaretto, on an island in the lagoon.
You may ask: Professor, what does the plague have to do with artificial intelligence? A great deal. In Ragusa nobody had a new cure. The great innovation was organizational: clear rules, separate flows, records of who came in and who went out. It got results without touching a single patient.
Today artificial intelligence enters the medical practice, and the temptation is to look straight at diagnosis. But the biggest and safest gains, once again, lie in organization. And the clinical decision stays with you.
The rules of the game, before the tools
Before talking about what to do, let's look at the boundaries.
- Health data is special-category data. The GDPR gives it reinforced protection (Article 9). That means reliable providers, data processing agreements, adequate security measures and, for the most sensitive processing, an impact assessment. In Italy, the reference authority is the Garante per la protezione dei dati personali.
- Never paste patient data into a general-purpose chatbot. Free consumer tools are not designed for health data: no names, tax codes, test results or medical histories.
- Software that diagnoses is a medical device. A program with a diagnostic or therapeutic purpose falls under the EU Medical Device Regulation and, if it uses AI, among the AI Act's high-risk systems. It's not a do-it-yourself job.
- Italian law is clear about the doctor's role. Italy's Law 132/2025 on artificial intelligence states that in healthcare AI is a support, that decisions remain with the physician and that patients have the right to be informed about the use of these technologies.
The message is not “forget it”. It's: start where the risk is low and the benefit is high. And that place, almost always, is far from diagnosis.
Where AI really helps a medical practice
1. A front desk that's never engaged
This is where most time is recovered. A single channel for non-urgent requests, with guided paths (“repeat prescription”, “certificate”, “appointment”, “information”), delivers every request already sorted into the right queue. Answers to administrative questions (hours, cover arrangements, documents to bring) can be automated. Appointment reminders with one-click cancellation reduce missed visits. We described this kind of project in the typical scenario of a group medical practice.
A golden rule: no automated system performs clinical triage. If a message contains signs of urgency, the reply must immediately tell the patient to call emergency services.
2. Documents that (almost) write themselves
Letters to colleagues, reports, instructions for preparing for a test, recurring forms: many texts follow similar structures. An assistant set up in an environment suitable for health data can prepare a first draft from the information you enter. You reread, correct and sign. For tools that transcribe or summarize consultations, be doubly careful: inform the patient, check where the data is processed and always review the final text.
3. Communicating better with patients
Explaining a treatment in plain words, preparing a clear instruction sheet, translating practical directions for a patient who doesn't speak the language well: these are tasks where AI is surprisingly good, as long as you check the content. A patient who understands is a patient who follows the treatment better.
4. Order in the practice's data
How many chronic patients haven't come in for a check-up in over a year? How many requests arrive on Mondays compared with Thursdays? A small dashboard on administrative data helps organize work and front-desk shifts.
What not to do
- Don't start from the tool. Starting from “let's buy AI software” instead of “where are we losing time?” leads to projects nobody uses.
- Don't forget less digital patients. The phone shouldn't disappear: it should become the exception, at set times, especially for older patients.
- Don't delegate judgment. A well-written draft is still a draft. Responsibility rests with whoever signs, today as always.
- Don't improvise on privacy. Risk assessment and contracts first, then go live.
How long it takes and what it costs
For a group practice, a digital front-desk project usually takes 6–8 weeks, plus the time for the impact assessment. Costs depend heavily on the tools already in use and the number of doctors: that's why we're wary of one-size-fits-all price lists and prefer to start by analyzing a typical week. You can get an idea of our work on the page for doctors and medical practices and the one on automation services.
What to do on Monday morning
- Count the calls. For one week, whoever answers the phone notes the reason for every call. You'll discover how administrative the workload really is.
- List your tools. Write down which software and apps the practice uses today and where patient data ends up. If a general-purpose chatbot is being used with real data, that's the first problem to fix.
- Pick a single process. Repeat prescriptions, reminders or certificates: start with the one that weighs most and leave the rest for later.
Want to think it through with someone who starts from people, not technology? Book a free intro call: we'll listen to how your practice works and tell you frankly whether and where AI can help.
The Professor at Penzai
