Lapsi Health embeds its clinical AI into medical records via Redox: the plumbing matters more than the algorithm

By Zendoric · July 27, 2026.

Lapsi Health, a clinical artificial intelligence developer with Dutch and US roots, has closed a deal with Redox, one of the leading providers of healthcare data interoperability, to natively connect its Keikku platform to hospitals' electronic health record (EHR) systems. The agreement allows healthcare centers to deploy Lapsi's AI without building a custom integration for each one: they tap directly into the network of connectors Redox already has in place with more than 12,000 healthcare organizations and over 100 EHR systems — among them Epic, Oracle Health/Cerner and MEDITECH — according to Redox's own figures.

Keikku combines three functions in a single interface inside the EHR itself: ambient documentation (it records the conversation between doctor and patient and drafts the clinical note automatically), real-time contextual clinical reference and diagnostic support. The platform can run on software alone or be paired with Lapsi's own hardware, capable of capturing high-fidelity ambient audio and bodily acoustic signals — digital auscultation — for AI-assisted analysis. "Healthcare does not need more AI tools disconnected from one another; teams need solutions that fit naturally into the clinical workflows that already exist," sums up Rodrigo Alvez, chief technology officer at Lapsi Health.

The problem this deal seeks to solve has a name of its own in the healthcare sector: "point-solution fatigue," the overload hospitals suffer when they have to evaluate, buy and integrate dozens of AI tools separately — one to transcribe consultations, another for triage, another for clinical coding — each with its own process for connecting to the record. Redox acts here as an intermediate layer: instead of Lapsi building a different interface for each hospital, it writes once against Redox's network, which translates, normalizes and feeds the clinical notes back into each client's EHR. According to the company, that cuts implementation timelines from months to weeks.

Our reading is that this news says less about how smart Lapsi's AI is and more about where value is shifting in clinical AI: from the quality of the model to the plumbing that connects it to the doctor's real work. It is the same pattern we already see in agentic AI in other sectors — whoever controls integration and distribution wins, not necessarily whoever has the most capable model — and healthcare, with dozens of mutually incompatible EHRs, is fertile ground for infrastructure to win the game before the algorithm does. Redox does not diagnose or draft notes; it sells the pipe through which all of that flows, and that middleman position makes it almost indispensable for any clinical AI provider looking to scale fast.

It is worth separating two promises the announcement blends together. Ambient documentation — recording the consultation and generating the note automatically — is already a proven category in the market, with well-documented benefits: it frees doctors from the most administrative and tedious part of their work, the part that demands the least clinical judgment. As sector context, it is the same logic we have been observing in other professional fields: AI takes the paperwork first, not expert judgment. AI-assisted digital auscultation for diagnostic support is a more ambitious promise, and this announcement provides no independent clinical validation data — no sensitivity, no specificity, no comparisons against a human specialist. The hardware is real; diagnostic reliability at scale, for now, remains a product aspiration and not a demonstrated clinical result.

In the short term, integrations like this ease the administrative strain that drives clinical burnout, and they fit a healthcare AI market that, by the industry's own account, has become hospitals' top technology investment priority for 2026. In the long term, if AI-assisted auscultation tools come to be validated as solidly as ambient transcription already is, they point in the direction we champion at Zendoric: earlier, more accessible diagnoses that depend less on a specialist's availability, a small but real step toward medicine that detects before disease advances.