kaiko
A clinical AI workspace that sits on top of a hospital's existing systems, assembling the full patient record into cited consult, meeting, handover and discharge outputs. Certified to ISO 27001 and NEN 7510, sold to hospitals only.
What is kaiko?
kaiko is a clinical AI workspace published by Kaiko B.V., an Amsterdam company with a second office in Zurich. Rather than another point solution or another chat assistant, two forms the company argues solve nothing, it installs as a layer on top of the hospital's existing systems and reasons across the whole patient record.
The problem it targets is familiar on any ward. Patient information sits scattered across systems, notes, scans and reports; hospital software is built to store data rather than analyse it; and clinicians rebuild the same overall picture by hand, under time pressure, often outside working hours.
The answer is a library of validated clinical workflows organised around four moments of care. Before the consult, kaiko produces a cited patient overview and answers guideline questions grounded in the department's protocols. In the multidisciplinary meeting, it builds per-patient summaries matched to the agenda and supplies comparable cases for a second opinion. At the diagnostic read, two workflows still marked emerging handle AI-assisted tumour segmentation and structured response assessment. After the visit, it drafts handover notes and discharge summaries, and extracts structured variables for cardiac, surgical, cancer and quality registries.
The design is agentic: the agent assembles, updates and reasons over context while the clinician steers and validates. kaiko launches in context from the EHR with the patient already loaded, reasons against local protocols rather than generic medical knowledge, and links every claim back to the passage it cited, so nothing has to be taken on trust.
Behind the product sits a research lab training foundation models across medical imaging, pathology, genomics and clinical text. Midnight, CoralBay and the eva evaluation framework are open-sourced on Hugging Face and GitHub, with papers on arXiv; Vortex, the flagship line of clinical frontier models, is announced as coming soon.
Deployments are named and dated. Patient Summary launched at the Antoni van Leeuwenhoek in May 2026, LUMC CaireLab joined in October 2025, and Vitalis moved to structural use in August 2026. Governance is certified, covering ISO 27001, NEN 7510, GDPR and the EU AI Act, but no kaiko feature is currently certified as a medical device.
What it does
- Assembles the full patient context from the record, with every claim linked back to its source
- Prepares the consult with a cited patient overview built from the whole file
- Answers clinical questions grounded in the department's own protocols rather than generic medical knowledge
- Builds per-patient summaries for multidisciplinary meetings, surfacing open questions and relevant guidelines
- Drafts shift handover notes and discharge summaries from the in-hospital record
- Extracts structured variables for cardiac, surgical, cancer and quality registries
- Runs AI-assisted tumour segmentation and structured response assessment over time, both emerging workflows
When to use kaiko / When not to
A quick filter to help you decide if kaiko is the right fit.
When to use kaiko
- Hospital clinical teams who rebuild the same patient picture by hand, across several systems, before every consult
- Multidisciplinary meeting chairs and coordinators preparing per-patient agendas, open questions and applicable protocols
- Oncology, radiology and pathology departments in European centres of the kind kaiko already works with
- Hospital CIOs and CMIOs who need ISO 27001, NEN 7510 and EU AI Act evidence before any AI reaches the ward
- Research groups looking for open clinical foundation models such as Midnight, CoralBay and the eva evaluation framework
When not to use kaiko
- Individual clinicians and private practices: there is no self-service sign-up and no published price
- Anyone who needs a certified medical device, since no kaiko feature holds that status and ISO 13485 and EU MDR are still in progress
- Teams expecting the tool to make the medical call: the website terms exclude clinical use and leave every decision with the user
- Buyers who must compare costs before speaking to sales, as no pricing page, free plan or trial is published
- Users expecting a mobile app, a browser extension or a documented public API, none of which exists
How to use kaiko
A typical end-to-end flow, from setup to results.
- Request a demo through the form on the site, or write to info@kaiko.ai; there is no self-service sign-up
- Scope the department and the workflows with kaiko, who configure the workspace around the systems already in place
- Connect kaiko to the hospital's EHR, which stays the system of record and is not replaced
- Set up SSO, group-level access controls and audit logging before clinical users are onboarded
- Validate the outputs against real clinical cases, in your own environment and with your own team
- Pilot on a single workflow and measure the effect, as Vitalis did between September 2025 and January 2026
- In daily use, launch kaiko in context from the EHR with the patient already loaded
- Tell kaiko what you are working on: a consult, a multidisciplinary meeting, a handover, a discharge letter
- Read the structured output and check each claim against the cited passage and the steps taken
- Edit, validate and act, remembering that the clinical decision stays with the clinician
Pros & Cons
Pros
- Measured and published time savings: in the Vitalis pilot, doctors saved 30 minutes per intake, physiotherapists 12 and occupational therapists 15, an estimated 632 hours a year
- Full traceability, with every claim linked to the passage it cited and the steps it took, so an output can be checked rather than believed
- ISO 27001 and NEN 7510 certifications verified by external audit, with the BSI certificates linked from every page
- Patient data stays under the hospital's control, is not pooled into a shared model, and is claimed to sit on European infrastructure
- Installs above the systems already running: the EHR is not replaced, and workflows arrive built and validated rather than developed in-house
- Named, checkable deployments at the Antoni van Leeuwenhoek, LUMC CaireLab and Vitalis
- A genuinely open research output, with Midnight, CoralBay and eva on Hugging Face and GitHub and papers on arXiv
Cons
- No published price and no pricing page at all: the site's full sitemap contains none, and every route ends at a demo form
- No free plan and no trial announced, and an institutional-only sale that puts the tool out of reach of an individual practitioner
- No kaiko feature is certified as a medical device, and ISO 13485 and EU MDR are still described as in progress with the notified body
- The website terms rule out clinical use of the workspace while the product pages sell support for complex clinical decisions, and the gap is never explained
- The trust center is unreachable without a browser, leaving the subprocessor list, any customer DPA and the exact certification status unverifiable
- Two diagnostic workflows are labelled emerging, and Vortex, the flagship model line, is still announced as coming soon
- No interface language, named integration, mobile app, documented public API or published support channel
Pricing & Plans
No price is published. The site carries no pricing page at all, its full sitemap contains none, and neither a permanent free plan nor a free trial is announced. Every call to action leads to the Request demo form or to info@kaiko.ai, with terms agreed at institution level. A prospective buyer therefore cannot compare costs without contacting the vendor first.
Data, GDPR & hosting
A consolidated view of how kaiko handles your data.
GDPR overview
Implementation is concrete and documented. The privacy notice, last updated on 16 December 2025, claims compliance with the GDPR, the Dutch UAVG and the Swiss Federal Data Protection Act. A Data Protection Officer is designated, reachable at privacy@kaiko.ai or at Louwesweg 6, 1066 EC Amsterdam. Sections 10 and 11 set out access and copy, rectification, erasure, objection, restriction and portability, with a one-month response deadline extendable to three months. Both supervisory authorities are named with full addresses: the Autoriteit Persoonsgegevens in The Hague and the Swiss Federal Data Protection and Information Commissioner in Bern. Security follows the ISO 27001 standard, and the product is externally audited against ISO 27001 and NEN 7510, with SSO, group-level access controls and full audit logging. No Article 27 representative is designated, and none is required: Kaiko B.V. is established in the Netherlands.
Who owns the data?
kaiko states that patient data stays yours: it remains under the hospital's control and is not pooled into a shared model, the workspace improving through the company's research and clinical partnerships rather than by moving customer data elsewhere. For its own commercial activity, covering clients, recruitment and the website, Kaiko B.V. acts as the controller of the personal data it processes. Website content, including text, graphics, software and other materials, is owned or licensed by kaiko under section 3 of the terms. One gap matters: the published privacy notice explicitly excludes research involving patient data, which falls under a separate Privacy Notice for Participating Patients that the site does not publish.
Reuse rights
Reuse is tightly framed. Under section 4 of the terms, you may view and download material from the site for personal or internal business use only. Modifying, copying, reproducing, distributing or creating derivative works of that content is forbidden, commercial use requires kaiko's prior written consent, and proprietary notices may not be removed or altered. Anything you submit through a contact form or by email may be used by kaiko for business purposes, including answering your enquiry. On the personal-data side, processing rests on contract performance, legal obligation and legitimate interest, with consent sought only where the law requires it, and kaiko states that it undertakes no automated decision-making or profiling. Personal data may be shared with the Swiss branch, with IT and cloud subcontractors, with external advisers and in the event of a business transfer, though none of those third parties is named. The only transfer outside the EEA foreseen is to Switzerland, which the notice covers with a European Commission adequacy decision.
Data retention & training
Hosting summary
kaiko states that patient data stays under the hospital's control and sits on European infrastructure, and the site describes the product as made in the EU. No hosting country and no cloud provider is ever named, so the claim cannot be checked from the pages themselves. The privacy notice foresees a single transfer outside the EEA: to Kaiko's own branch in Zurich, which the document justifies with the European Commission's adequacy decision for Switzerland, treating such transfers as equivalent to intra-EEA ones. Security follows the ISO 27001 standard, with encryption of sensitive data, regular reviews of access rights and confidentiality agreements signed by staff, and the product itself is externally audited against ISO 27001 and NEN 7510, the BSI certificates being linked from every page. One caveat: the public website resolves to an anycast Amazon node in the United States, which describes where the marketing site is served and says nothing about where clinical data lives.
Things to keep in mind
Risks and trade-offs to weigh before adopting kaiko.
- The website terms declare the workspace not intended for clinical use, yet the product pages promote support for complex clinical decisions; the contradiction is never resolved, and it leaves the legal exposure with the user
- Outputs are probabilistic and may be incomplete, inaccurate or misleading by kaiko's own admission, and each one must be independently reviewed by an appropriately qualified professional
- Automation complacency: a fluent, well-cited summary invites trust, and the busier the shift, the less likely anyone is to open the source passage the tool helpfully links
- Skill erosion, as clinicians who stop assembling the patient picture themselves may lose the habit of noticing what a summary quietly leaves out
- The published privacy notice does not cover patient data used in research, which falls under a separate document the site does not publish
- European hosting is claimed without naming a country or a provider, and the subprocessor list lives on a trust center that cannot be reached
- No feature is certified as a medical device, so treating a segmentation or a response assessment as a diagnostic result goes beyond what the vendor itself claims
Setup & Integrations
Technical difficulty
Not a self-service tool: setup is an institutional IT project rather than an installation. It starts with a demo and a contract, then an integration with the hospital's EHR, in-context launch, SSO, group-level access controls and audit logging. kaiko says the workflows arrive built and validated and that it configures the workspace around the systems already running, with nothing for the customer's team to build or maintain. Outputs are then validated on real clinical cases, in the customer's own environment. Vitalis piloted from September 2025 to January 2026 before generalising, which is a realistic timescale to plan for.
Deployment
Behind kaiko
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Do we have to replace our EHR to use kaiko?
What happens to our patient data?
Is kaiko certified and compliant?
How much does kaiko cost?
Who makes the clinical decision?
Which hospitals already use it?
What time savings have actually been measured?
What is kaiko Research?
Is there a mobile app or a public API?
Where is the data hosted?
Should you pick kaiko?
kaiko is unusually convincing on the ward and unusually opaque at the negotiating table. Three named European institutions run it, the Antoni van Leeuwenhoek, LUMC CaireLab and Vitalis, and the published pilot figures are specific enough to be argued with: 30 minutes saved per intake for doctors, 12 for physiotherapists, 15 for occupational therapists, roughly 632 hours a year. Behind the product sits a genuine research lab, with open foundation models on Hugging Face, papers on arXiv and a CVPR 2026 challenge win alongside the Netherlands Cancer Institute.
The governance story is equally serious on paper: ISO 27001 and NEN 7510 verified by external audit, GDPR compliance, EU AI Act conformity, SSO, group-level access controls and full audit logging. It is also, in practice, hard to check. The trust center that would carry the subprocessor list, any customer DPA and the exact certification status is unreachable without a browser.
Two things deserve a buyer's attention. The commercial side is entirely closed: no price, no plan, no trial, no self-service, and a sitemap that confirms no pricing page exists. And the site contradicts itself. The website terms state that the workspace is not intended for clinical use and must not be used to make medical decisions, while the product pages sell a second-opinion workflow to support complex decisions in any specialty. No feature is certified as a medical device, and ISO 13485 and EU MDR remain in progress.
For a European hospital with a real fragmentation problem, an integration team and a compliance function, kaiko is a credible candidate that arrives with workflows already validated. For anyone hoping to trial it alone, compare its cost, or lean on it as a certified diagnostic aid, it is the wrong tool today.
- Choosing a selection results in a full page refresh.
- Opens in a new window.