Carasent's Germany U-turn: the legacy product it planned to bury is now an AI project

A quiet Q3 on the surface — 14% organic ARR growth, a 300-user ambient scribe — but the call contains a genuine reversal on the German legacy asset management spent years planning to sunset.

CARA.ST · Earnings Call · 2026-10-15

A normal quarter with an abnormal footnote

Carasent's Q3 2026 numbers (reported October 15) read like steady execution: headline growth of 29%, 14% organic ARR growth, 12% organic revenue growth and an EBITDA margin of 20%. Total revenue grew from SEK 82m to SEK 108m. Nothing in that line-up demands a second look. What demands a second look is a strategic reversal the CEO tucked into a slide about Germany — and it reframes a company that has been telling the same "kill the legacy, migrate the base" story for several quarters.

Germany: from exit to retrofit

For years Carasent's German plan was arithmetic: shrink the inherited Data-AL base, push customers onto the new Webcur platform, let the old product die. That plan has a cost, and the CFO named it: the German business is "contributing a lot to that churn." Group churn was SEK 9m, or 2.8% — with the Nordics below 2% and Germany dragging the average up. Much of that is retiring doctors whose successors want a different system than Data-AL. So management is now testing the opposite. Instead of forcing a migration, it is using AI to modernise the legacy product in place. Daniel Öhman framed it as a real strategic option, not a throwaway slide:

I think there might be potential for long-term shift... It has nothing to do with our beliefs in Webcur. It's unchanged, but it's our belief that we can do so much more with our software now, and we know that it's so difficult to get customers to change. So if we can renew their present product, then that will really make it much easier to just keep the customer and start growing with them.
Daniel Öhman, Chief Executive Officer · 2026-10-15

This is the freshest idea in the dossier. The logic is that the hard part of software is not the code but the decades of accumulated customer requirements embedded in it — and AI can now port that accumulated knowledge across architectures in a way that was impossible before. If it works, Carasent could hold its churn-prone German base without forcing a disruptive swap. If it fails, it has spent development dollars on a product it is still trying to retire. Either way, this is a company that spent quarters on one message now openly entertaining two.

The ambient scribe is quietly becoming a business

Medsum, Carasent's embedded Medsum clinical AI (an ambient documentation assistant), moved from science project to revenue line: close to 400 active users, 300 paying, and SEK 3m of ARR. That is small against roughly 30,000–40,000 active users across the EHR base — but it is growing fast enough to be called out on the ARR bridge. The interesting part is how out of step Europe is with the United States. Öhman put the adoption gap starkly: in the U.S., “70%-80% of all outpatient visits in the U.S. use ambient scribes... In Europe, it's obviously moving way slower.” — Daniel Öhman, Chief Executive Officer · 2026-10-15 Carasent today sees only 5–10% of its users touching this technology in any form. That gap is the whole investment case: not a new market, but a catching-up market. Management is also deliberately not monetising the upside. The price stays at roughly SEK 750 per user per month, with the new agentic "action" features — auto-drafted sick-leave forms, referrals, revisit scheduling — bundled in free. Öhman's arithmetic on customer economics is compelling: “if they save an hour per day... that will give them easily time to have one extra patient... The payback is really easy. But healthcare is really slow to change, and that's the problem.” — Daniel Öhman, Chief Executive Officer · 2026-10-15 The payback is a month, not a year — and yet adoption still crawls because the buyer, a physician, is trained to fear change and litigation. The under-appreciated moat is regulatory. A MDR certification project (Class 2a, run through Sweden's Läkemedelsverket) will cost a few million SEK in Q4. Öhman's blunt payoff: “there will be fewer ambient scribes on the market because of this type of regulation.” — Daniel Öhman, Chief Executive Officer · 2026-10-15 Regulatory friction, usually a cost, becomes a competitive moat against a flood of point-solution ambient scribes.

Margins, seasonality and a foggy Q4

Before getting carried away, the CFO supplied the cold water. Q3 is seasonally the strongest margin quarter because holiday-lowered personnel cost lands on normal revenue; that reverses in Q4. Worse, the year-ago quarter was flattered by a one-off German consulting fee and a SkatteFUNN grant. “it was a very special quarter... a big one-off revenue in Germany, and the positive one-off effect on the cost side... these effects will not come back in Q4 this year.” — Svein-Martin Bjørnstad, Chief Financial Officer · 2026-10-15 Add the MDR spend, a sales-and-marketing step-up in Germany (SEK 3–4m incremental) and new hires, and the clean read doesn't resume until Q1 2027. Underneath, the operating leverage story is intact. The acquisitions — Falck Norge and Frelsesarmeen in the sales pipeline, and Infosolutions, which added over SEK 40m of ARR and carries ~70% gross margin — are being folded into a model that converts incremental revenue to profit efficiently. Ex-Germany, organic growth is 15% and the EBITDA margin is 25%, versus 12% and 20% for the group. That gap is precisely why the Germany decision matters more than its current revenue.

Confluence: applied AI in a year that turned against AI infrastructure

Broadly, AI was everywhere in this reporting window — HCLTech sold "agentic AI" offerings, Wipro pushed "physical AI," Marsh spoke of clients' AI transformation. Carasent sits adjacent but is a different animal: it is an applier of AI inside a narrow clinical workflow, not a builder of compute. That distinction has mattered in the tape. Over the last 90 days, the market's AI-infrastructure complex sold off hard — "AI data centers" was among the sharpest decliners, dragged by roughly 52 tickers, with high-bandwidth memory, PCIe Gen 6 and co-packaged optics all down. Applied, vertical AI that cuts a doctor's admin time is not the same trade, and may not share the same pain. Nobody in global coverage is selling clinical ambient documentation at scale; Medsum is a company-unique asset, not sector boilerplate. The other thread with a broader echo is public-sector digitisation. The e-referral rollout in Region Stockholm — finalised over the summer, going end-to-end — is deepening Carasent's grip on Scandinavia's largest healthcare market, with Aleris eye clinics onboarding on the back of it. Öhman also flagged, half tongue-in-cheek, that a less left-leaning regional government could help. None of that is a near-term revenue story, but it is a structural position no small-cap competitor can easily rebuild.

The takeaway

Strip away the seasonality and the Q4 comparison noise and Carasent is doing two things at once: ratcheting an embedded AI product from zero to a real ARR line while quietly abandoning the certainty of its old German playbook. The company never mentioned a price action rewrite — the tape is simply unavailable here — so the signal lives entirely in the transcript, which makes the reversal easy to miss and, for that reason, more interesting. If the AI-modernisation experiment works, Carasent stops paying the churn tax that has capped its group margins. If it doesn't, management has reintroduced strategic ambiguity exactly when investors were hoping for a clean migration story.