Anyone who works in IT knows Moore's law: the number of transistors on a chip doubles roughly every two years. For decades, that was the pace of technological progress. It brought us everything, from the PC to the smartphone in your pocket.

AI is moving faster. The independent research institute METR measures the autonomous task length of an AI system: the length of a task an AI system can complete on its own, expressed in the time a human would need for it. Since 2024, that length has roughly doubled every hundred days. That is about seven times as fast as Moore's law. METR measures this on software and research tasks, so it is not a yardstick for healthcare. But the direction is clear: AI no longer just handles short jobs, it increasingly takes on longer and more complex tasks by itself.

We see this every day in our own field. Our developers also write code together with AI. For healthcare, the question is no longer whether AI will play a role, but how. And in healthcare in particular, that 'how' makes all the difference.

All In, with people in control

At Topicus Healthcare we have made a clear choice: we are going All In, for healthcare. For social impact, and with people in control. That may sound like a slogan, so we are happy to explain what it means in practice.

First of all: no AI for the sake of AI. Every application must demonstrably add value for the professional or the patient. A feature that mainly impresses in a demo but solves nothing in daily practice is something we do not build.

Second: reliability is our trade. Healthcare software has to meet strict requirements, such as NEN 7510, ISO 27001 and sometimes MDR. That is deeply embedded in how we work. Many innovative start-ups are currently entering the market with smart point solutions. That is great, because a lot of innovation comes from them and healthcare needs it. But getting such a solution to land reliably in healthcare, scalable and embedded in the bigger picture, takes knowledge and experience. That is where we want to make the difference.

Third, we deliberately choose independence. We work model-agnostic: we do not make ourselves dependent on a single AI supplier, including the big tech companies that currently dominate this field. That way we can choose what fits best for each application, and we stay in control ourselves.

And fourth: AI is above all a change in how people work. Handing out a tool and hoping for the best does not work. We train and support our people, share what works and what does not, and now see AI skills as a core competence. That is only possible in an open culture, where you are allowed to experiment and where we learn from each other's experiences.

What AI already delivers today

So much for the theory. What does it deliver in practice? Two examples from general practice that we offer in Topicus products today.

The S-line. Through an online questionnaire or e-consultation, a GP or practice nurse often receives a long, associative story from the patient. That is valuable information, but it takes time to read. AI turns that story into a compact S-line, the subjective part of the report, which is ready in the record as a proposal. The care provider sees at a glance what is going on. Meanwhile, the patient can simply explain in their own words what is wrong. This application is currently running as a pilot in Spreekuur.nl in combination with VIPLive Monitoring.

Sam, the AI assistant of Spreekuur.nl. Sam talks with the patient in plain language, without forms or menus. Scheduling an appointment, requesting repeat medication or, where possible, getting self-care advice straight away: it can all be done in one conversation. For the patient this is easy and available outside practice hours; for the practice it means fewer phone calls for standard matters such as scheduling an appointment or repeat medication. Sam is still very much in development. The next step is validated triage based on the NHG and NTS guidelines, with the care provider explicitly in the loop.

What we see for iQoutcome

With iQoutcome we see many opportunities for AI. At our recent inspiration day we presented two lines of thought. They have not been built yet: we are developing them further together with clients.

The first is an AI summary of outcome measurements on the dashboard. A clinician who wants to form a picture of the situation of a client who has been assigned several measurement trajectories, with several measurement instruments completed, needs to take some time to dig into it. That takes time, and with a full caseload there is a risk that a signal is missed. AI can summarize this into a short, clear text with the most relevant indicators. To support the clinician. The interpretation and the decision always remain with the clinician.

The second line of thought is personalized notifications. Non-response is a well-known problem with outcome measurements: when clients do not complete their questionnaire, the clinician has less insight into how they are doing. Not everyone responds at the same moment or to the same message. AI can learn when and how a client is best reached. That leads to more completed questionnaires, and with that better insight for clinician and client alike.

Enthusiastic, and rightly critical

Our clients' reactions at the inspiration day varied, and we found that valuable. Many clients see the potential of AI that supports the clinician and professional, and the ideas for iQoutcome were received positively. At the same time, some are hesitant about AI in the medical decision-making process. Will clinicians not be influenced too much? Will they keep thinking for themselves enough?

We recognize that question from our own field. In software development, too, the challenge is to let AI speed up the work without losing your own thinking. We do not have a complete answer to that in one go. But we do know how we want to approach it: step by step, explainably, and together with the people who work with it every day.