Or how an AI solution became a lesson in business model innovation
There is something you hear surprisingly often during your time in the Netherlands: countless casual remarks about a certain physiotherapy appointment. Whether someone cannot work on a group project because they have to be there, someone recovering from a visible injury, or simply someone sharing a bad experience. Where I come from, it is much less common to hear about physiotherapy unless someone has a particularly serious problem. So when two of my teammates came up with an ambitious idea to improve an industry that apparently affects so many people, I was immediately excited.
From our initial idea and research, we seemed to be facing several clear problems: limited treatment capacity, expensive sessions, growing patient demand, and pressure on physiotherapists. We immediately felt that we had hit the nail on the head. Here was an industry where we could potentially offer something better to almost every stakeholder. It was exciting to imagine introducing something close to a miracle solution.
We identified the main opportunity in improving the quality and supervision of exercises performed independently at home, something physiotherapists have traditionally had limited visibility into. By doing so, we hoped to improve recovery and relieve some of the pressure on the industry.
My own work began with investigating the market, the industry, and the problem itself more closely. Our initial idea still contained many unknowns that needed validation.
A solution that sounds wonderful in theory
The concept sounded simple enough. A patient has an application on their phone that reminds them when it is time to perform their daily exercises. They open it, see the prescribed exercises, how to perform them correctly, and how many repetitions they should complete.
Using computer vision, the application could analyse live or uploaded recordings and provide feedback on their movements. Afterwards, patients would answer a short questionnaire about their symptoms and recovery.
Meanwhile, the physiotherapist could see how the patient was progressing at home, receiving information about completed exercises, movement quality, and reported recovery. This could also allow treatment adjustments without necessarily requiring another appointment.
Importantly, the physiotherapist would remain the decision-maker. Our intention was never to replace medical expertise with AI, but to give professionals better information and make their work more efficient.
In theory, the solution seemed wonderful. Patients could receive more continuous guidance, while physiotherapists could spend less time repeatedly explaining familiar exercises and more time helping patients who actually needed their attention. This could potentially reduce unnecessary appointments and create capacity for new patients, without sacrificing treatment quality.
And considering the pressure on the healthcare workforce, this also seemed like a more sustainable way to organise care.
There was just one catch.
When improving efficiency means losing money
During my investigation of the industry and its business models, I encountered a systemic problem that challenged our entire value proposition.
Dutch physiotherapy predominantly operates under a fee-for-service model, meaning practices are generally paid for individual treatment sessions.
Suddenly, the supposed advantage of requiring fewer appointments became a disadvantage for the very practices expected to purchase our product. More efficient treatment could mean less revenue.
This was probably my biggest realisation throughout the project: creating value and capturing value are two very different things.
We had been approaching the problem mainly from the perspective of improving the service. But even if patients benefit, and even if the technology works, why would a practice pay for something that could reduce its income?
What made this particularly interesting was that the industry had not simply overlooked digital innovation. Our research uncovered existing applications, blended-care experiments, and studies investigating alternatives to traditional treatment. One Dutch study even examined combining fewer face-to-face sessions with a digital exercise programme, finding broadly comparable clinical outcomes and lower intervention costs.
The technology, therefore, was not necessarily the missing piece. The greater difficulty was creating a business model that rewarded its benefits.
We had encountered a classic business model innovation problem. Industries can be transformed by rethinking not only what they offer, but how they create, deliver, and capture value. Yet in healthcare, changing the payment structure is much more complicated than simply launching a new product.
Bundled payments, for example, could allow practices to receive compensation for an entire treatment pathway rather than individual sessions. Under such a model, improving efficiency would become much more directly profitable. However, changing reimbursement arrangements requires the involvement of insurers and other stakeholders, making immediate widespread adoption unrealistic.
So rather than abandoning our solution or depending on an industry-wide transformation, we started exploring how it could create value within the existing system.
Finding the missing link in our business model
One possibility was improving adherence and reducing premature dropout. With reminders, feedback, and better monitoring, patients might be more likely to continue their prescribed rehabilitation. For practices, this could mean more patients completing appropriate treatment rather than abandoning it halfway through.
Another advantage was faster information exchange. Instead of spending the beginning of each appointment reconstructing what happened since the last visit, therapists could already have a structured overview of the patient’s progress. This could free time for more important clinical work.
A third possibility was better utilisation of capacity. If some patients required fewer routine check-ins, practices could potentially accommodate additional patients, provided there was sufficient demand to fill those appointments.
These benefits were not guaranteed, but they gave us several possible ways to justify the product economically without immediately changing the reimbursement system.
Our proposed subscription of €15 per active patient per month made the challenge surprisingly concrete: the application would need to generate approximately twelve minutes of physiotherapist value each month to justify its price.
What I found particularly interesting was the possibility of decoupling certain activities traditionally delivered together during a physical appointment. Routine exercise monitoring, progress reporting, and information gathering could take place independently, while professional assessment and clinical decisions would remain with the physiotherapist.
We were not removing the professional from the service. We were reorganising the activities around their expertise.
The bigger opportunity lies beyond the application
However, I believe the greatest potential of the solution lies in the future business models that such a system could enable.
Once practices have access to reliable information about patients’ exercises and progress, entirely different ways of organising treatment become possible.
For example, patients with similar conditions and compatible rehabilitation programmes might be offered supervised group sessions where clinically appropriate. This could potentially reduce the cost per patient while allowing therapists to serve more people.
Alternatively, certain treatment pathways could be offered as more affordable packages, with a greater proportion of supervised home rehabilitation. This might make physiotherapy accessible to price-sensitive patients who would otherwise postpone or avoid treatment.
The application could also maintain a connection between patients and practices after treatment. If symptoms returned, patients could more easily reconnect with their physiotherapist, who would already have relevant information about their previous rehabilitation.
These possibilities made me realise that collecting information is not valuable simply because it produces data. Its real value comes from the decisions and new service arrangements it makes possible.
