When I was working in workforce administration at a mid-size hospital network, I spent a fair amount of time with ward coordinators during their morning routines. What struck me was not the complexity of the scheduling problem. It was the simplicity of the tool they were using to solve it. A printed contact list, a phone, and a tracking sheet for who had been reached. The same process that would have been recognizable fifteen years earlier.
This was not a technology-averse environment. The same ward managed patient records digitally, tracked medication administration through a software system, and ran rostering software for planned schedules. But the moment a planned schedule broke down, whether a sick call, a sudden demand spike, or a last-minute double booking, the response was entirely manual. You picked up the phone and started calling.
The specific frustration that led here
I had been building workforce scheduling tools for hospital networks for several years. The tools were good at what they did: they helped coordinators build and visualize rosters, track certifications across a nursing team, and project staffing costs. What they did not do was help when reality diverged from the plan.
I watched a coordinator spend nearly three hours filling an overnight gap in a surgical ward on a Thursday evening. She had 23 contacts on her list. By the time she reached confirmed, she had made 17 calls. Eleven went to voicemail, two nurses were already working elsewhere, one had moved and was outside the city, two declined, and one confirmed. The confirmed nurse had been the fourth person she called, but their callback did not come through until an hour and a half into the process.
The problem was not the list. The problem was that the list was static and the matching problem was dynamic. The nurse pool's real availability was changing continuously through other shift confirmations, personal schedules, and sleep patterns around night shifts, and the only way to discover the current state was to call each person individually.
What Markus and I concluded after digging into it
I brought this framing to Markus Eigenmann, who had spent years building matching infrastructure for service industry labor markets. His reading of the problem was immediate: this is not a contact management problem, it is a real-time two-sided availability problem. The data to match a nurse to a shift exists in most cases. What does not exist is a mechanism to read both sides simultaneously and surface the right matches without sequential manual effort.
Two-sided availability matching at speed is a solvable engineering problem. The specific challenge in healthcare is that credentials and scope of practice add constraints that generic gig platforms do not handle well. A nurse who is available and willing to work cannot necessarily take a specific shift at a specific ward. Swiss healthcare has both federal and cantonal layers of certification requirements, and individual wards often add specialty requirements on top of that. Generic availability matching that ignores credential depth produces high false positive rates: matches that look good until the coordinator checks whether the nurse actually qualifies for this particular ward.
We spent time with Sophie Meier, who at that point was running nurse recruitment and partnership building for a hospital group, to understand the credential landscape from the clinical hiring side. Her view was clear: platforms that glossed over credential specificity were the ones that broke trust with clinical teams. A coordinator who confirmed a nurse through a platform, then had to send them home because a certification was incorrect, did not come back to that platform.
What we are actually trying to build
Carewell is not a staffing agency and we are not trying to replace one. We are a matching platform. The clinics own the shifts and their own standards. The nurses own their credentials and their availability. Our job is to read both sides accurately and surface the right options fast enough that the manual phone process is no longer the faster or more reliable path.
We care about credential depth because credential errors destroy clinical trust faster than slow matching. We built the SRK/ASI verification layer before we built any of the matching speed features, because a fast match that is wrong is worse than a slow match that is right.
We are also honest about scope. We are starting with Swiss German-speaking Switzerland, where Sophie's network gave us the clinic relationships and nurse community access to seed the pool properly. Expanding faster than the nurse pool depth supports would produce a poor experience. The under-10-minute gap fill we see in early-access pilots only holds because the relevant nurse pool is adequately populated. Thin pools produce slow matches; we are not going to paper over that with speed claims that do not reflect real supply conditions.
Why this moment specifically
Swiss nursing vacancy rates have been climbing for several years. The Swiss Nursing Association publishes figures on this regularly. The structural dynamic is not a short-term gap in an otherwise stable system; it reflects demographic pressure, training capacity constraints, and an ongoing tension between nurses' scheduling preferences and how hospitals have historically structured work. Nurses are increasingly seeking schedule control. Clinics are increasingly struggling to fill short-notice gaps as nursing populations skew older and working patterns fragment.
These two trends point toward a marketplace model. Nurses with the experience and credentials to work independently want a reliable way to pick up shifts on their own terms. Clinics with reliable, frequent gaps need a pool of qualified nurses who have already been vetted and are ready to confirm. The connection between those two needs was not being made efficiently because the infrastructure to make it did not exist in the Swiss market.
We are not the first to notice this. There are gig staffing platforms in other European markets, some with healthcare-adjacent features. But the Swiss-specific credential structure, the language regions, and the cantonal variation in certification requirements make a purpose-built Swiss approach more defensible than adapting a foreign platform for local use. That is the bet we are making.
What matters to us as we grow
We want ward coordinators to genuinely trust the matches we surface. Not because we told them to, but because the nurses we confirmed actually showed up, were qualified for the ward, and worked the shift well. That trust is built shift by shift. The speed metrics matter and the cost metrics matter, but they are second-order. First-order is: does the nurse who confirmed actually arrive and do the job?
We also want nurses to feel that Carewell respects their professional standing. Swiss nurses are qualified healthcare professionals with real credentials and real licensing obligations. A platform that treats them as interchangeable hourly labor is wrong about what they are. We built the Verified badge and the in-platform credential record partly for the practical reason that it speeds matching, and partly because nurses who feel their credentials are properly recognized use the platform differently than nurses who feel like a commodity.
That is the version of Carewell we are building. Not perfect yet, nowhere near the pool depth we want to reach, but pointed in the right direction.
