AI has started changing hospitals across Europe. Finally. First documentation, then patient communication and medical knowledge.
Yet in the last twelve months, investor focus has materially shifted. No early-stage investor puts money into AI documentation anymore. The hot new topic seems to be billing, or in broader terms, revenue cycle management (RCM). In the US it’s been a hot market for years - now that’s reaching Europe.
Important note: I need to keep this article from becoming an encyclopedia. That’s why I’m going to strictly focus this chapter on RCM in German hospitals. That’s because a) I know this best, b) it’s the biggest health care market in Europe, with an est. €130-150b flowing through German hospitals annually, and c) consequently most European billing startups are targeting Germany. A lot has been written about US revenue cycle management already so I’m skipping that
Since it’s an incredibly complex market, I got high-profile support from two directions:
For one, I sparred with my friends from Bertelsmann Healthcare Investments, short BHI. They have been “long RCM” for quite some time, as they have helped scale Phare Health to an exit to R1 and are going to announce new RCM investments soon. I’ve added some of their takes as direct quotes throughout the article.
Secondly, I looped in Jörg Matheis, CEO of PVS Westfalen-Nord. They’re one of the most established service providers in the German RCM market and have lived the problem inside-out. With ordomediq they’re building a corporate startup in the space.
Together, let’s discuss why that market is attractive, what the value chain looks like and what startups need to solve.
Why it’s attractive (now)
You’ll probably agree that bringing AI to hospitals is a lucrative business opportunity. The big question is not if or when AI hits hospitals, but which startup wins among all this competition.
There’s no segmentation between AI use cases anymore. Every company tries to sell every AI use case imaginable to hospitals. Whatever the hospital wants. Even if a startup doesn’t pitch other AI use cases publicly, it’s certainly on their roadmap. A company that started with AI documentation certainly won’t stop there.
That’s where RCM comes in: It’s such a complex and high-priority part of each hospital’s business that most AI hospital startups in Europe haven’t touched it yet. It’s literally connected to everything in the hospital.
VCs have started loving this field for 3 reasons:
Sure, it’s difficult to convince hospitals to let AI anywhere near their business-critical processes. But at least the complexity shields you from your AI neighbors - the scribes, AI receptionists & Co
Since you sit directly at or over the hospital’s P&L, you can argue it allows higher value creation and thus value capture. In simple words: If you can make the hospital more money with this product, you’ll earn more yourself
Since all hospital processes are somehow tied to billing and revenue, these products could ultimately be used to steer hospital processes. They could grow well beyond the billing department into the entire organisation
What VCs underwrite when investing in these early-stage AI billing companies is another shot at goal of building a hospital software giant.
BHI comment: Why now? We believe Germany’s RCM landscape is due for a generational shift, driven by three converging forces. (1) Coding and regulatory complexity is compounding faster than hospitals and service providers can manage it. (2) Incumbent vendors are falling short on delivery, economics, quality and - as Unimed showed - security. (3) Most importantly, technology finally makes in-sourcing realistic: control over working capital, fewer third-party fees, less pressure on staff. The make-or-buy question is open again.
The amazingly complex value chain
Before we proceed we need to get to a joint understanding of the underlying use case. What is revenue cycle management even? What are the underlying tasks an AI could automate?
We’ll differentiate between GKV (statutory insurance) and PKV (private insurance), since they are two largely separate processes. The revenue cycle works something like this in German hospitals:
Patient access & pre-admission. Yep, this is where it starts. You need clean data to begin with. The hospital schedules the case and checks who’s paying. GKV patients get an electronic admission notice sent to their insurer, PKV patients sign a cost coverage confirmation and often extra upsell agreements. All of this is necessary but not directly revenue-impacting
Treatment & documentation. No billing without documentation! Doctors and nurses treat the patient and document it in the hospital information system, and many other data repositories (radiology, lab, …). This documentation is the raw material for everything that follows
The actual coding. Now it gets interesting. Hospitals need to translate the medical records into ICD-10 diagnoses and OPS procedure codes, while acknowledging the German Coding Guidelines (Deutsche Kodierrichtlinien, DKR). The codes might sound straightforward, but the DKR is where most ambiguity and mistakes happen. Often the coding is started by the clinicans and then refined by clinical coders in the medical controlling department. It really depends on the hospital how much is clinican vs coder responsibility. Once that’s done, a “grouper” software turns those codes into a “DRG”, the flat rate the hospital gets paid per case. Nursing costs are billed separately, and the current hospital reform is going to introduce further changes… it’s certainly not getting simpler over time.
Jörg’s comment: In the PKV system it’s even more complicated. Every single step may be billable: Rounds, laboratory, medication, ... In the GKV system, you’ll (mostly) get the same price for a procedure. In the PKV system, the “price” for a procedure can differ significantly - depending on how it was exactly done
Billing. For GKV patients, the billing department sends the invoice electronically to the Kasse via §301. The hospital bills the insurer directly besides negligible co-pays. For PKV patients, the process splits in two. The hospital usually bills the chunk of it (the DRG part) directly to the insurer. On top of that, the chief physician bills elective services separately under a centrally agreed fee schedule (GOÄ). That’s often done through an external billing provider and direct to the patient! Even if not, it’s likely done by the chief physician’s own office and not the hospital’s controlling department
Claims review & disputes. Afterwards, the staturoy insurers don’t just pay - of course not. They can trigger audits by the “Medizinischer Dienst”, which reviews whether the case was coded and billed correctly on a repeating basis. Those quotas are set annually, and they are going up with the current reform! This step is where hospitals actually lose money, and Medizincontrolling teams spend a massive share of their time on audit defense, appeals, and rebilling. On the PKV side, the insurer reviews GOÄ line items and multipliers instead - but from what I’ve heard it’s less picky than on the GKV side
Jörg’s comment: Besides the claims review: About 6% of all invoices create a contact with a patient after billing (payment issues, questions, wrong address etc.) - and dunning (Mahnwesen) is not even included
Payment & receivables. Finance departments book the incoming payments and chase open items. Statutory insurers pay within short statutory deadlines but claw money back after audits. PKV cash sometimes depends on the patient actually paying, so the bad-debt risk is with the hospital. Which is why it’s common to work with 3rd party factoring providers
Controlling & budget negotiation. At the end of the cycle, the hospital reports case mix and audit loss rates internally and has to deliver its case data to the governing body. Then they negotiate next year’s budget with the insurers based on which services the hospital will offer and at which planned volume. It’s not the step startups usually target, since it’s a very relationship- and expertise-driven process. Still good to know - because in this system, exceeding plans can mean that extra income is clawed back and re-distributed to other hospitals!
Important clarification: this whole workflow doesn’t directly apply to outpatient settings in Germany. It’s a mostly separate market with not necessarily less, but different complexity, much higher startup activity and faster decision-making. So as a startup, if you want to pick a slightly easier-to-enter market, choose outpatient clinics. If you want higher defensibility and larger potential upside, you pick hospitals - but make sure you raise enough funding
I think you understand why the space is so complex. You end up with three payers (GKV, PKV, patient), two different fee systems, and multiple different people or departments in charge of billing. It couldn’t be more confusing!
Which step is now the most valuable for startups to address?
Most founders start with coding, since that’s the most straightforward to address with AI. They usually promise the hospital better coding results and time saved
Many also want to cover the billing process for the PKV, since it’s an adjacent and easy-to-outsource step
In the medium term, some startups look at factoring because again it’s a typical outsourcing target. It also offers attractive margins. On the other hands you need excellent risk management, plus it’s cash-intensive and highly regulated
My general observation is that most startups stay in one lane at first: They either pick GKV/DRG coding, or the private side. I assume it’s to stay focused product-wise and because those can be different buying centers
Nonetheless the long-term vision usually is going upstream. Since most of the friction starts with the right documentation, that’s where you’d start!
BHI comment: In the US we’re seeing scribe/documentation players like Abridge now moving downstream into the RCM cycle. They are strategically well-positioned to train and deploy models on newly reached documentation standards. The European market is not near this development yet.
Excellent point from BHI here - why haven’t Heidi and Tandem launched this yet?! What happened to launching new products at godspeed? Well, I assume they’re aware of the challenges we will discuss next.
What startups need to solve
In theory, bringing AI to this process sounds easy: You connect to the EHR, pull the clinical documentation and match the procedures / diagnoses to the right codes with LLMs. You then group them by the given billing logic and digitally send out the result to the payer. Essentially one afternoon in Claude Code.
In practice this breaks on so many ends. Brace yourself:
A large chunk of the documentation might not be available in digital form. You might need to scan in paper or just live with the data being analog. Even if you have digital data, it might be in the form of PDFs and you’d need huge OCR volumes to process it. A lot of data processing just for very few snippets that actually inform the billing result. That impacts your margins
Even worse than analog data is never-documented data. Controlling / billing departments in hospitals spend a lot of time calling clinicians and asking for missing information. Very often important clues are not documented or not in an ideal way. Not trivial to solve with LLMs!
You will likely need integration into the hospital’s EHR, which is a notorious challenge. Perhaps you can work with data exports in the beginning, but for a truly great product EHR integration remains a must. Unfortunately, the billing process (under the GKV at least) is considered a “core functionality” of every EHR. These systems may be slow. They may be bad (today). But they are already in place! They have full access to all data. They are in a position to restrict access to the hospitals data for newcomers. Yep, in theory that shouldn’t be the case - but I can assure you, it is
Often, the EHR is not enough. The preferred way for RCM players to get data access is to connect to a clinical data repository (CDR). That’s essentially a uniform, accessible, aggregated data lake across all systems of record. All hospitals want that. BUT… they plan 5 year-projects for that transformation, and most have not even started. Building those CDRs btw creates another potential source of competition. Those companies set up a CDR, provide numerous services based on that data - and can finally also enter billing
While trying to solve the documentation gap, you run into hardcore change management. Will the doctors really love it when an AI tells them how to treat and document their patients? Imagine you’re finally leaving an 8-hour surgery just get a “want me to code this case? here’s the catch: it’s not a simple fracture—it’s a polytrauma. And that matters.”
Another fun challenge in Germany is data protection. We’re the absolute overlords of data privacy. For example, you’re not allowed to just send an invoice to patients via email, hell no! It technically contains health data, so you need end-to-end encryption. Emails don’t clear that bar, so your best guess is a link to a secure portal in the email with 2-factor-authentification... just wow
Then, there’s the determinism problem of AI. Using LLMs means statistics. If you bill an item as it has been done “probably”, you are committing fraud... and the willingness of patients and health insurers to use the f-word (at least as a threat) to force future “defensive billing” is high
If you target the private system, know this: We’ll have a GOÄ reform with a new set of codes by 2028. Do you just focus on that new rule set? That means you have no product to show until then. Or do you opt for old GOÄ and new GOÄ at the same time?
Jörg’s comment: Besides all these RCM-specific topics, you still deal with the long decision timelines in German hospitals. It doesn’t help that recently, a huge data breach made the rounds on national German media. Unimed, a billing service provider, and several University hospitals were involved. There goes the risk appetite on the hospital side…
Sounds horrible, right? Maybe not a venture opportunity?
I actually believe all these points are challenge,s but not a showstopper long term. Smart and driven founders will be able to solve it. My main point here is that it’s not smooth sailing like you’d hope for. Startups in this space need a serious strategy to navigate all the fragmentation, the hidden agendas and the data minefield.
In terms of moat and stickiness of contracts, this is of course awesome! Once you’re the AI revenue cycle management provider or have helped to bring the process inhouse, chances the hospital kicks you out are a conservative 0.0%. And as teasered in the beginning: RCM is close to the brain of hospital operations & admin. You can expand in every direction from there - in the end, most things revolve around money, right?
Which startups to watch?
There have been some cool funding rounds recently, not all of them public yet. In Germany, the publicly visible players are probably calliora, which focuses on RCM as part of their AI platform, and perhaps Apylon (based in Denmark but close). In France, BeParallel raised a large round with a similar entry point, leveraging computer-use agents to run coding in “parallel systems” - an interesting way to circumvent cumbersome EHR integration to get to ROI. In Eastern Europe, Mediqcode is making waves. And we’ll hear about more names soon, I’m sure!
BHI comment: Winning teams will pair AI fluency and deep technical backgrounds with real domain expertise and access (which is a rare combination), a security-and-customer-first mindset, and a commitment to selling real outcomes rather than additional software. More to come!
I’m betting on this to become the new booming category in European healthcare. It’s a slow start, it will take funding, but the reward is huge. If it just remotely develops toward the US market - with top players making billions in revenue and huge PE bets - we’ll see great VC outcomes.
We’ll talk more about how startups solve this in another edition!
Speak soon,
Lucas
P.S.: If you’re building in this space, let’s talk. Immediately