I spend most of my time building AI products. But one of the projects closest to my heart has very little to do with chatbots or agents. It is about nurses who drive from home to home, every day, to care for people who cannot come to them.
When I started Maison des Infirmiers, a turnkey network that helps home-care nurses in France launch their own private practice, I had a strong opinion about technology. I thought it could fix most of what makes the job hard. I still believe technology can help a lot. I am just much clearer now about where it stops.
This post is my attempt to draw that line honestly.
A profession that is growing, and carrying more weight
Independent nursing in France is not a niche. According to the DREES (the statistics department of the French health ministry), around 98,600 nurses worked in private practice in 2021. The same study notes that the number of independent nurses grew by 28.5% between 2013 and 2021, compared with 6.9% for salaried nurses.
That growth tells you two things. First, more nurses want autonomy over how they work. Second, more care is moving into people's homes, which means more responsibility rests on professionals who often work alone.
An independent nurse is a clinician, but also a small business owner. They manage schedules, billing, regulatory rules, replacements during holidays, relationships with doctors and pharmacies, and their own social contributions. None of that was in their training. Much of it happens late at night, after the last visit.
That is exactly where technology belongs.
What technology does well: the work around the care
The best use of software in home-care nursing is not touching the care itself. It is removing friction from everything around it.
Finding the right people. A nurse who wants to take a week off needs a replacement they can trust. A patient leaving hospital needs a nurse nearby. These are matching problems, and matching is something software does well. This is why we built Soinvite, a directory with more than 100,000 nurses listed, and RemplaIDEL, which helps nurses find a replacement within 48 hours. Without such tools, these searches mostly happen through phone calls, word of mouth and social media groups. That works, until it does not.
Seeing the money clearly. Many independent nurses only understand their real income at the end of the year, when contributions and taxes arrive. That uncertainty creates stress and bad decisions. Previzo was built to forecast contributions and help nurses optimise their income. A simple, honest projection of what you will owe is not glamorous technology. It is just useful.
Reducing administrative load. Paperwork, scheduling, reminders, document preparation: this is where AI assistants are starting to make a real difference. A well designed assistant can draft, summarise, remind and organise. In our AI projects at ThalerTech, we often see that the biggest gains come from these unglamorous, repetitive tasks, not from spectacular demos.
Launching a practice with less guesswork. Opening a private practice involves many steps that every nurse has to rediscover alone. A structured network with shared templates, processes and tools shortens that path. Technology here is not the product. It is the backbone that makes support scale.
What technology cannot do: the care itself
Now the other side.
It cannot replace presence. A home visit is not only a clinical act. The nurse is often the only professional who sees how a patient actually lives: whether they eat, whether they are lonely, whether something has changed since last week. No sensor or app captures that judgment the way a trained person in the room does.
It cannot carry responsibility. When a nurse decides something is wrong and calls a doctor, they own that decision. AI can surface information, but in healthcare the accountability must stay with a human professional. I am very cautious about any product that blurs this line.
It cannot fix structural problems alone. Reimbursement rules, territorial shortages, long driving distances in rural areas: these are policy and organisation questions. Software can make a broken system slightly more bearable, but it should not be sold as the cure.
It cannot build trust from nothing. Patients and families trust a nurse, not a platform. Nurses trust colleagues they know. Any tool that ignores these human networks will fail, however well it is engineered.
How I think about building for nurses
A few principles guide how we build in this space.
Start from the field, not from the technology. The most useful features come from listening to nurses describe their week, not from asking what AI could theoretically do. If a feature does not save time or reduce worry on a real Tuesday afternoon, it is not a priority.
Respect the time of the user. Nurses are on the road, between patients, often on a phone. A tool that needs long training or many clicks will not be used. Simplicity is not a design choice here. It is survival.
Keep humans in the loop. Whenever a decision touches patient care, the product should inform, never decide. This is a line I hold even when the technology seems capable of more.
Be honest about limits. It is tempting in healthtech to promise transformation. I prefer to promise less and deliver what we promise. As I often say: "Ideas, no matter how brilliant, are mere potentialities. It's the execution that brings them to life." In healthcare, sloppy execution has real consequences.
I wrote more about why I keep starting companies like this one in Why I keep building companies. The short version: I am drawn to problems where good execution changes daily life for real people.
Where this is going
I expect AI assistants to become a normal part of an independent nurse's toolkit within a few years, mostly for administration, organisation and communication. I also expect the best tools to be almost invisible: they will run in the background, inside apps nurses already use, and simply make the day lighter.
What I do not expect, and do not want, is technology that tries to stand between the nurse and the patient. The value of home care is human. Our job as builders is to protect that, by taking everything else off the nurse's shoulders.
Key takeaways
- Independent nursing in France is growing fast, and nurses carry both clinical and business responsibilities.
- Technology helps most with the work around care: matching, finances, administration and launching a practice.
- Presence, judgment, accountability and trust cannot be automated.
- Build from real field needs, keep it simple, and keep humans in charge of care decisions.