Stryker’s $3B Vocera Buy Signals AI Push in MedTech

A YC-Backed Startup Building the AI Layer Between Hospitals and Device Suppliers

A three-person team from Cologne is moving into the United States with an AI platform that is live in 17 European hospitals. The company is CONUS. Its product: an AI knowledge layer sitting between medical device manufacturers and the clinical teams who use their products. Its backers include Y Combinator's Fall 2026 batch. The expansion signals that vertical AI platforms, not manufacturer portals or generic search, are becoming the default interface between MedTech companies and hospitals.

CONUS launched six months ago. Since then it has signed agreements with Johnson & Johnson, Stryker, and Zimmer Biomet, three of the largest orthopedic manufacturers in the world, and gone live in hospitals across Germany and neighboring markets. The platform ingests manufacturer documentation, hospital-specific protocols, and peer-reviewed research, then lets surgeons and sales reps query it through a conversational interface called CONUS Intelligence. The founding team, Maximilian Becker, Vincent Paffrath, and Alexander Schulz, started the company in 2022 while finishing graduate degrees. They began with five chief orthopedic and trauma surgeons who told them the same thing: the information they needed during surgery was scattered across PDFs, vendor portals, and the memory of their sales representatives.

The Y Combinator connection came through a conversation at ETH Zurich. General Partner Tom Blomfield met the founders at a YC event and, by their account, told them they were thinking too small. That conversation pushed them to apply. They were accepted with Grey Baker as their primary partner. In September the three founders relocated to San Francisco for the three-month program. The move is deliberate: the US market is where the largest device manufacturers base their digital strategies, and where hospital systems face the most pressure to reduce avoidable errors tied to information gaps.

Traction in Europe gives the expansion credibility. Each deployment represents integration with existing hospital IT, manufacturer data feeds, and clinical workflows. The platform now handles product specifications, surgical technique guides, and hospital-standard operating procedures in a single search layer. For manufacturers, it replaces the fragmented portal strategy where each company maintains its own documentation site and hopes clinicians remember the login. For hospitals, it reduces dependency on sales reps as the primary knowledge conduit — a dependency that creates delays when reps are unavailable and blind spots when they are.

The long-term ambition is explicit: CONUS wants to become the transaction layer where hospitals select and order implants across manufacturers — what the founders describe as "Amazon for implants, embedded in the surgical workflow." That vision turns the knowledge layer into a procurement layer. The US expansion tests whether a European-built vertical AI platform can insert itself into the commercial relationship between American hospitals and the manufacturers that dominate the global orthopedic market. If it works, the sales rep's information monopoly breaks. If it doesn't, the manufacturers' own platform plays (Stryker's SmartHospital, Zimmer Biomet's ZBEdge) will set the terms.

Why Device Information Still Lives in PDFs and Sales Rep Pockets

Documentation error rates in the operating room are not marginal. A 2026 study at a U.S. children's hospital captured 1,240 timestamp events across 202 cases and found an overall error rate of 17 percent (roughly one in six entries), 2.6 times higher than the 0.1–0.5 percent benchmark for simple routine tasks in high-end manufacturing. When documentation was delayed more than 20 minutes, the error rate climbed to 38 percent. Delays occurred in more than half of all events. Each case generated an average of 253 clickable documentation tasks; the OR suite runs over 10,000 cases a year, producing roughly 2.5 million nurse charting tasks annually. Orthopedics posted the highest specialty error rate at 34 percent, followed by neurosurgery at 29 percent.

Perioperative nurses describe the friction: scanning a tray's barcode, the system returns no match, and they must decide: skip the entry, guess a proxy code, or leave the field blank and promise to fix it after the case. The item master, intended as the single source of truth for multiple hospital systems, is not maintained. Substitute products, increasingly common, are rarely catalogued. Item-number changes compound the problem. Barcodes on packaging are unreadable or the item isn't in the system. Charge-capture databases often fail to sync with documentation systems, creating siloed workflows that force OR staff and revenue-cycle teams to manually reconcile charts.

The workflow that produces those numbers has changed little in decades. Surgical reports are still largely dictated in free text, sent to a transcription office, corrected, and approved; no digital data-entry mask is involved. Physicians reuse old reports or personal text modules, applied individually without hospital-wide standardization. Even where structured reporting templates exist, most hospitals use partially structured forms that mix predefined patterns with free-text fields. The full potential of structured reporting (machine-readable, interoperable storage that can trigger downstream safety processes) is realized only when all information is captured through a digital interface. That interface is missing in most ORs.

Perioperative nurses bear the friction. Their documentation drives implant traceability, surgical billing, and patient safety, yet they describe it as the most frustrating part of their workflow. Structured reporting improves completeness (81 percent vs. 66 percent for free text) and surgeon satisfaction (visual analog scale 8.1 vs. 3.5), but adoption stalls. Transcription-based documentation remains common across Europe because strict data-protection laws and information-security requirements block cloud-based AI tools. Even where digital templates exist, they often lack the semantic data model and visual interfaces needed for true interoperability. AI scribes have expanded rapidly in outpatient settings (Mass General Brigham saw a 5.6-minute median EHR-time reduction per visit), but the OR is a different acoustic and operational beast. High-acuity environments are noisy, dynamic, and multi-speaker. Current ambient-scribe products cannot consistently distinguish simultaneous conversations among surgeons, anesthesiologists, nurses, and techs using a single recording device. Regulatory frameworks (EU MDR, FDA SaMD) have not clarified whether summarizing scribes count as medical devices, leaving liability for documentation errors unresolved.

The fragmentation is structural: device data lives in manufacturer PDFs, hospital item masters are incomplete, EHRs don't talk to charge-capture systems, transcription persists, and the only reliable real-time source is the sales rep standing at the surgeon's elbow. That is the problem a vertical AI knowledge layer is built to solve.

Inside the OR: A Knowledge Base Changes Day-to-Day Surgery

CONUS Intelligence sits on top of a corpus the company says exceeds 500,000 academic papers, 2,000 manufacturer documents, and official society guidelines. The platform ingests each hospital's SOPs, rehab protocols, and contact lists, then tags them to the relevant procedures. Clinic-specific implant variants (the exact catalog numbers a given OR stocks) live alongside the manufacturer's own documentation. Drug reference data sits in the same interface. The result is a single search surface that replaces the drive, the rep's phone, and the guideline app.

"Our standards are no longer scattered across the drive, but right where we operate. My team works from the same version, and I can rely on it."

The quote comes from Prof. Dr. C. W., chief physician of trauma surgery and orthopedics at a German hospital using CONUS under a site license. Internal clinic standards, implant-specific documents, and specialist contacts are gated behind that license, a deliberate design choice that keeps proprietary hospital workflows off the public web while still letting the AI reason across them.

OR nurses at CONUS sites report a parallel change. S. M., a surgical technologist in orthopaedics, writes: "In CONUS, I can see our procedure, the correct positioning, and the appropriate materials for the surgery at a glance. This gives me confidence and helps me work faster and more systematically in the OR." T. K., a physician assistant in trauma surgery, says: "Before a procedure, I use CONUS to review the workflow, implant systems, and specific considerations. This helps me prepare more thoroughly and support the team in the OR more effectively."

The platform surfaces three content layers in every answer: the hospital's own SOP, the manufacturer's IFU or technique guide, and the relevant guideline or paper. Each citation links to the source page. The company explicitly disclaims use for patient-specific diagnosis, dosing, or emergencies, a guardrail that keeps CONUS in the decision-support tier rather than the clinical-decision tier.

Manufacturers see a different workflow. Stryker and Zimmer Biomet, both early partners, upload structured product data (implant variants, surgical techniques, regulatory documents) directly into the platform. That data then becomes queryable by any licensed hospital using their devices. The sales rep still gets called, but for configuration questions or case-specific troubleshooting, not for "which tray do I need?"

Preparation time drops. The company reports significantly reducing preparation time across its European sites. Independent systematic reviews of AI in OR management, covering 22 studies from 2019 to 2023, find consistent efficiency gains when knowledge retrieval moves from manual search to structured query. The platform is version 1.19 as of the latest app-store release. Its knowledge base carries a validity stamp through July 2029. For the hospitals already live, the AI layer has become the default starting point — not a supplement, but the first place the team looks.

The Sales Rep Is No Longer the Only Door

For decades the medical device sales rep owned the information pipeline. A surgeon with a question about implant sizing, instrument compatibility, or a revised surgical technique called the rep, often the same person who had steered the original purchase. That dynamic made the rep indispensable but also created a bottleneck: clinical teams waited for answers, and manufacturers had no direct line to the end user once the contract was signed.

CONUS and similar AI knowledge layers break that bottleneck. By ingesting product catalogs, hospital protocols, and published research into a single searchable platform, they give clinicians instant answers without a phone call to the rep. The platform already works with Stryker and Zimmer Biomet across its European hospitals, and its US expansion will test whether American systems adopt the same direct-access model.

Survey data shows the sales force is still early in its own AI adoption. Only 36 percent of 150 US medical device reps polled by AcuityMD in mid-2026 said they use AI regularly or occasionally; 92 percent of those users reported saving at least four hours a week, but the work automated so far is routine: email drafting, task organization, meeting summaries. Few reps use AI for strategic account research or commercial insight, and most rely on general-purpose tools rather than healthcare-specific platforms. That gap is where a vertical knowledge layer matters: it turns proprietary device data into a queryable asset that both the hospital and the manufacturer can trust.

The rep's role is shifting rather than disappearing. Tasks that are purely informational (product demonstration (28 percent automatable per one analysis), clinical outcomes data presentation (55 percent), contract negotiation support (38 percent)) are moving into the platform. What remains irreducible are clinical relationships, in-OR technical support, and hands-on training (rated 20 percent automatable). Companies are responding: enterprise sales quotas and on-target earnings rose 14 percent as AI absorbs volume work, while sales leadership headcount grew 22 percent. Firms are spending more on fewer, higher-skilled professionals who can interpret platform analytics and translate them into account strategy.

Manufacturers gain something they never had before — a direct feedback loop. When clinicians search CONUS for a specific implant revision or complication rate, that query signal flows back to the product team without a rep filtering it. Predictive analytics on the platform can forecast demand, optimize inventory, and surface leads from prescription histories and clinical trial data. Chatbots and recommendation engines already change how reps engage accounts, but the deeper shift is structural: the knowledge layer becomes the default interface, and the rep becomes a high-leverage interpreter of that layer rather than its sole custodian.

The winners will be reps who treat the platform as a teammate, not a threat. Organizations that embed AI into structured sales processes, not just as a productivity add-on, report stronger commercial outcomes. The door is still there; it just opens onto a wider room.

Countermoves: Big MedTech, Rivals, and Regulation

Stryker's response to the knowledge-layer shift has been the most aggressive in the industry. The company's Vocera acquisition in 2022 laid a hospital-wide communication backbone that integrates with more than 150 clinical systems. The care.ai purchase in 2024 added ambient intelligence: sensors that monitor patient rooms, detect falls, and automate documentation. Together they give Stryker a data flywheel no competitor yet matches: procedural data from 1.5 million Mako robot cases annually, plus real-time recovery data from hospital wards. The company's 2023 R&D spend hit, and its AI patent grant share reached 26 percent in Q1 2024. Mako is already moving beyond knees and hips into spine and shoulder applications.

Zimmer Biomet has taken a different path. Its 2022 partnership with Hospital for Special Surgery created a dedicated AI innovation center for robotic joint replacement. The Monogram Technologies acquisition in October 2025 positions Zimmer to field the first fully autonomous orthopedic surgical robot, with FDA 510(k) clearance in March 2025 and commercialization expected early 2027. OrthoGrid Systems, acquired in August 2024, adds AI-powered fluoroscopy guidance for hip arthroplasty. The ZBEdge suite now links ROSA robotics intra-op data, Persona IQ smart-implant sensors, and the mymobility app tracking 250,000 patients post-discharge. WalkAI, launched March 2022, predicts 90-day gait recovery from daily mobility patterns. Independent studies show 15-minute OR time savings and 95 percent cup-placement accuracy with these AI systems.

The competitive set is not standing still. Intuitive Surgical's da Vinci 5 delivers 10,000 times the compute of its predecessor with true force feedback and an AR roadmap, but its scope remains the surgical event itself — no ward-level sensing, no post-discharge loop. Medtronic's Hugo RAS platform bets on modularity and an open console; its Touch Surgery AI analyzes operative video for training. Johnson & Johnson's VELYS system differentiates on a CT-free, imageless workflow for knee replacement and is building an AI case-management tool for ambulatory surgery centers. Each rival improves the robot. None has assembled Stryker's hospital-wide nervous system.

European regulation is accelerating the shift. The EU AI Act, with its August 2026 compliance deadline for high-risk medical AI, forces manufacturers to prove continuous post-market surveillance, data governance, and model monitoring, capabilities that standalone device portals lack. The May 2026 omnibus amendment clarified product-safety exemptions and revised timelines, but the core obligation remains: AI-enabled devices must demonstrate lifecycle traceability from training data to clinical outcome. Fines reach €35 million or 7 percent of global turnover. A platform that centralizes protocols, device data, and research — and can audit every answer back to source — becomes a compliance asset, not just a workflow tool.

The Talent Race Behind the Knowledge Layer

CONUS lists three employees on its Y Combinator page (three founders) and its job posts read like a map of where the vertical AI knowledge layer actually breaks. The company says it is growing its engineering team across backend, AI infrastructure and product engineering. The workatastartup posting adds frontend product engineering to the mix and specifies the stack: strong Python skills and experience building production backend systems, ideally with Django, plus sound judgment around API design, relational data modeling, permissions and application architecture.

That challenge — grounding LLM output in traceable, regulated source material — is the hiring signal. The EU AI Act, which entered into force in 2024 with high-risk obligations operational from August 2026, turns it into a compliance requirement. The regulation creates roles that didn't exist two years ago: AI Compliance Auditor, AI Risk Assessment Specialist, Technical Documentation Lead, AI Notified Body Assessor, Fundamental Rights Impact Assessor. Lawyers, policy professionals, compliance officers, and risk managers with AI knowledge are now the most sought-after profiles. An estimated 60,000-plus companies operating in the EU need to comply by 2026; fewer than 5,000 qualified compliance professionals exist in Europe.

The salary data reflects the scarcity. Engineers with EU AI Act compliance knowledge command an estimated 20–30 percent premium over comparable profiles, particularly when they also bring medical imaging, clinical workflow or medical device experience. Regulatory Affairs Managers with Act expertise see some of the fastest salary growth in HealthTech. AI Compliance Officers are entering at Director or VP level. Roles linked to Act readiness take 30–50 percent longer to fill than pre-Act equivalents. Many HealthTech companies launched compliance programmes 12–18 months before the August 2026 deadline; those that waited face higher costs, fewer candidates, and compressed timelines.

Cologne, where CONUS is based, illustrates how the talent map is reshaping around this intersection. The city hosts 14 healthtech startups (led by Cannamedical Pharma, DISCO Pharmaceuticals, and CompuGroup Medical) making up 6 percent of its 800-plus startup ecosystem. Germany's broader AI market sits at with a average senior AI engineer salary (Western Europe average); Switzerland tops the region at. The EU targets in AI investment by 2030. Across Europe, 1.9 million AI-related job openings were recorded in 2025; Germany alone faces a shortage of 97,000 AI specialists.

Immigration policy is adapting. The EU Blue Card now covers AI and tech professionals with a relevant degree and a job offer paying 1.5 times the average local salary (typically), granting work rights across all member states. Germany's Chancenkarte (Opportunity Card) goes further: AI professionals can enter to job-hunt without a prior offer for up to one year. For a vertical AI platform like CONUS, building the knowledge layer between manufacturers and hospitals, these pathways matter. The company needs engineers who can ship Django-backed APIs, design retrieval pipelines that cite source documents, and navigate MDR/IVDR and AI Act requirements simultaneously. That hybrid profile — backend fluency, RAG architecture, regulatory literacy — is what the market is pricing.

The broader European backend market shows 728 open roles as of October 2026, with 71 new postings in the prior week; AI engineer vacancies sit at 283. But the MedTech vertical narrows the funnel further. CONUS's own postings signal the blend: work with product and frontend engineering to turn customer requirements into maintainable software, with clear architecture and engineers who take responsibility for both implementation and the user experience. The knowledge layer doesn't hire pure researchers; it hires product engineers who understand that a hallucinated specification in a surgical protocol is a patient safety event, not a model quality metric.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.