What can AI do in healthcare?
In healthcare, AI can support administrative and organizational work that involves no clinical judgment. Examples are rewriting general patient information in plain language, summarizing internal documents and answering practical questions about the practice.
This article only covers those administrative and supporting use cases. AI for diagnostics, triage and treatment is outside its scope. Separate rules for medical devices apply to those applications.
The article describes possibilities and refers to what Dutch regulators have published, as of October 2026. It is general information. It does not give medical or legal advice. Discuss your own situation with your data protection officer or a lawyer.
Which administrative tasks in healthcare suit AI?
Tasks that need no patient data are the best first step with AI. Tasks with patient data raise extra questions about privacy, security and regulation.
| Task | What the AI does | Which data is needed | Where a person checks |
|---|---|---|---|
| Rewriting patient information | Turns an existing leaflet or web text into plain language | The existing text | A care professional checks the content |
| Summarizing internal documents | Summarizes a protocol or policy document for the team | The document itself | The document owner reads it through |
| Copy for the organization | Writes a draft for a job ad or internal newsletter | Job requirements, topic, tone | The sender does the final edit |
| Practical questions | Answers questions about opening hours, directions and procedures | The information on your website | The team decides which questions are passed on |
Tasks without patient data
A leaflet about an examination, the explanation on your website about requesting a repeat prescription, an internal protocol: these are texts without data about an individual patient. An AI assistant such as ChatGPT or Claude can rewrite, shorten, translate or summarize such texts.
The content remains the responsibility of the care professional or the organization. An AI model can add or drop a sentence in a way that changes the meaning. So have every text that goes to patients read by someone with subject knowledge.
Practical questions from patients
Many questions to a practice are practical: when consulting hours are, where to park, what to bring to an appointment. An AI agent, a digital worker that answers based on information you supply, can answer such questions. What that is, is explained in what is an AI agent.
The line is at questions about symptoms, medication and treatment. Those belong with a care professional. Set up an agent so that it always passes such questions on to the practice, and discuss beforehand which patient data is processed along the way.
Tasks with patient data
As examples of generative AI in healthcare, the Dutch Health and Youth Care Inspectorate (Inspectie Gezondheidszorg en Jeugd, IGJ) mentions summaries of a patient record and reports of a consultation between doctor and patient. In such applications the AI processes health data.
Whether and how you introduce such an application is a decision for the care provider, together with the data protection officer and the supplier. The next section describes what regulators have published about it.
Why does AI in healthcare call for extra care?
AI in healthcare calls for extra care because health data has extra protection and because errors in the output can have consequences for patients.
Health data is a special category of personal data. That is laid down in Article 9 of the GDPR. The Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) describes it as highly privacy-sensitive and specially protected.
The AP reported a data breach at a GP practice. On 6 August 2024 the AP wrote that an employee of a GP practice had, against the agreements, entered patients' medical data into an AI chatbot. According to the AP, most companies behind chatbots store the data entered. The regulator advises organizations to make clear agreements about which data employees may and may not enter. The notice is on the website of the Autoriteit Persoonsgegevens (in Dutch).
The IGJ calls for care. On 10 February 2025 the IGJ called on care providers to handle the introduction of generative AI carefully. According to the inspectorate these models can produce incorrect or incomplete information, with medication errors or an incorrect assessment of a patient as possible consequences. The news item is on the website of the IGJ (in Dutch).
An AI application can be a medical device. As an example the IGJ mentions an AI program that helps make diagnoses or supports a treatment. According to the inspectorate such applications must comply with the regulation for medical devices (MDR).
The EU AI Act sets its own requirements. The European AI regulation has stricter rules for high-risk AI. Which applications fall under them and from when the rules apply is covered in our EU AI Act explainer.
Which questions do you ask before introducing AI in your practice?
The IGJ advises care providers to be alert when purchasing, implementing and using generative AI. The inspectorate names four points: establishing the purpose of the application, determining which data is needed and validating it, checking results and assessing risks.
These questions help you have that conversation with your team and your supplier:
- What is the purpose of the application, and is that purpose administrative or medical?
- Which data goes in, and does it include patient data?
- Who checks the output before anyone works with it?
- Where is the data stored and who has access?
- Is there a data processing agreement with the provider?
- How does the application align with NEN 7510, the Dutch standard for information security in healthcare?
- Which agreements apply to employees who use an AI assistant themselves?
What the GDPR asks of you in general when using AI is covered in using AI in a GDPR-proof way.
How do you start with AI without patient data?
You start with a task that needs no patient data and first record the agreements with your team.
Step 1: make agreements. Record which AI tools employees may use and that no patient data goes in.
Step 2: choose one task. For example rewriting a leaflet in plain language or summarizing a policy document.
Step 3: have the output read. Someone with subject knowledge checks every text.
Step 4: evaluate. After a month, discuss what it delivers and where it went wrong.
Little experience with AI assistants so far? Then start with AI for beginners.
Frequently asked questions
What can you use AI for in healthcare? This article describes administrative and supporting tasks, such as rewriting general patient information, summarizing internal documents and answering practical questions. Applications for diagnostics and treatment fall under separate rules for medical devices.
Is health data a special category of personal data? Yes. Article 9 of the GDPR lists data concerning health among the special categories of personal data. The Dutch Data Protection Authority describes it as highly privacy-sensitive and specially protected.
What does the IGJ say about generative AI in healthcare? On 10 February 2025 the Dutch Health and Youth Care Inspectorate called on care providers to introduce generative AI carefully. It advises establishing the purpose, validating data, checking results and assessing risks.
Is an AI application a medical device? It can be. According to the IGJ an AI application can be a medical device, for example a program that helps make diagnoses. Such an application must comply with the regulation for medical devices (MDR).
What went wrong at the GP practice the AP mentions? An employee entered patients' medical data into an AI chatbot, against the agreements. The Dutch Data Protection Authority reported this data breach on 6 August 2024 and advises organizations to make clear agreements.
Next steps with AI in your organization
Learn to work responsibly with AI in a 1-on-1 AI training, or schedule a no-obligation call about the administrative tasks in your practice or institution.
