Protocol AI
AI software from Swiss company Risklick AG that helps clinical teams design, optimise and write clinical trial protocols and study plans, drawing on a database of more than 800,000 trials, for both medicinal products and medical devices.
What is Protocol AI?
Protocol AI is clinical trial protocol software published by Risklick AG, a Swiss MedTech company based in Bern and a spin-off of the University of Bern. It is presented as an end-to-end solution that lets experts design and develop protocols and study plans quickly, so that sponsors shorten clinical development timelines. The company launched it for medicinal products at an event in April 2024, then released a version tailored to medical devices in January 2025.
The homepage frames the work in three stages. Explore analyses study design elements systematically; Design generates several scenarios so the design can be optimised; Develop drafts the documents using a customised generative AI. Underneath sits the product's main asset, a database of more than 800,000 clinical trials linked to publications, regulatory approvals and safety data. The vendor claims a 50% cut in protocol and study plan development time, 20% fewer protocol amendments and 40% less implementation time in EDC and eTMF systems, though it publishes no methodology for these figures.
The platform outputs a digital protocol built on the ICH M11 structure and the USDM standard, adaptable to any protocol template, with export to Word, Excel or JSON and integration with electronic trial systems. Inside the application the feature set is broad: document and template editors, a trial decision library, a study knowledge base with chat and synonym search, clinical guidelines and FDA approvals, regulatory parameters, adverse events, risk management, schedule of activities, amendment handling, audit trail, comments, a task manager and a Virtual PM. A connected tool, Study Engager, turns study documents into videos and questionnaires to gather patient feedback on the design.
Risklick backs the product with an advisory board of four named clinical experts, claims 13 peer-reviewed publications, and cites a completed proof-of-concept with Debiopharm in oncology plus a partnership with Citeline. Access is through a web application at protocolai.ch, reached only by requesting a demo. No price, tier or free trial is published anywhere on the site.
What it does
- Draft clinical trial protocols and study plans automatically with generative AI
- Explore study design elements systematically against a base of more than 800,000 clinical trials
- Generate and compare several design scenarios before committing to one
- Produce a digital protocol aligned with the ICH M11 structure and the USDM standard
- Run automated coherence, conflict and technical checks across study documents
- Export finished documents to Word, Excel or JSON for downstream systems
- Collect patient feedback on the study design through the connected Study Engager tool
When to use Protocol AI / When not to
A quick filter to help you decide if Protocol AI is the right fit.
When to use Protocol AI
- Clinical development teams in pharmaceutical and biotech companies who write trial protocols and study plans
- Medical device manufacturers preparing clinical investigations, including smaller firms with limited trial experience
- Medical writers and regulatory affairs staff who need structured authoring, consistency checks and change tracking
- Organisations moving towards ICH M11, USDM and Digital Data Flow interoperability for machine-readable protocols
- Academic clinical trial units and principal investigators designing investigator-initiated studies
When not to use Protocol AI
- Buyers who need to compare prices before talking to anyone: no rate, tier or free trial is published
- Teams wanting programmatic access: there is no product API and no public developer documentation
- Anyone expecting self-service sign-up, since the only public entry point is a demo request
- Non-English-speaking teams, as the interface and all published material are in English only
- Organisations looking for a system to run a trial and capture patient data, which is the job of an EDC rather than an authoring platform
How to use Protocol AI
A typical end-to-end flow, from setup to results.
- Request a demo from the homepage, which is the only public entry point to the product
- Discuss scope and terms with Risklick, since no self-service sign-up and no published pricing exist
- Receive access to the web application at protocolai.ch and sign in; nothing is installed locally
- Check that your infrastructure meets the vendor's minimum requirements, as the terms require
- Set up your organisation and assign the four roles: Organization Admin, Creator, Reviewer and Reader
- Import or build your protocol templates, which the platform is designed to adapt to
- Create a study, then explore design elements against the trial database and generate design scenarios
- Draft the protocol and study plans with the generative AI, working from the study knowledge you assembled
- Run the coherence, conflict and technical checks, then review as a team with comments and tasks
- Export the finished documents to Word, Excel or JSON and push them into your EDC or eTMF
Pros & Cons
Pros
- Verifiable academic grounding: a University of Bern spin-off with a named advisory board including a former director of the Bern clinical trials unit
- Dated, checkable industry references: a completed proof-of-concept with Debiopharm in oncology and a partnership with Citeline
- Built on open industry standards, ICH M11, USDM and Digital Data Flow, rather than a proprietary format only
- Covers both medicinal products and medical devices, with a version dedicated to the latter
- Scientific validation claimed through 13 peer-reviewed publications and a mention in a Lancet group journal
- Support advertised around the clock via ticketing and email, with a stated 24-hour maximum response time
- Short, readable legal documents in plain English, with GDPR compliance claimed and Swiss law applying
Cons
- No public pricing at all: no rate, no named tier and no pricing page anywhere on the site
- No free plan and no free trial announced; the only way in is a demo request
- No product API and no public developer documentation, so no programmatic integration is documented
- The headline gains of 50%, 20% and 40% come with no published methodology, and a launch release separately mentions reductions of up to 35%
- ISO 27001:2022 is presented as compliance, without a certificate number, certifying body or validity date
- Nothing is said about where data is hosted, no subprocessor list is published and no data processing agreement is mentioned
- The privacy policy has not been updated since October 2024, and nothing states whether customer documents feed model training
Pricing & Plans
Risklick publishes no pricing for Protocol AI. There is no pricing page, no rate card and no named plan on the site, and no amount or currency appears in any of its pages. No permanent free plan and no free trial are announced. The terms and conditions establish a subscription arrangement: fees are payable in advance and are non-refundable on termination, the agreement runs for an initial term and renews automatically unless cancelled with 30 days' notice, and usage limits such as the number of users, the duration and the available features are set by the plan subscribed to. Access may be suspended if payment is late, and the vendor's total liability is capped at the fees paid over the preceding 12 months. A figure can therefore only be obtained by requesting a demo and negotiating directly with the company.
Data, GDPR & hosting
A consolidated view of how Protocol AI handles your data.
GDPR overview
GDPR compliance is claimed explicitly in both the terms and the privacy policy, which is dated October 2024. The policy lists seven rights for users in the EEA: access, rectification, erasure, restriction, portability, objection and withdrawal of consent, all exercised by writing to support@risklick.ch. International transfers are acknowledged, with standard contractual clauses named as the safeguard, and the company states it does not sell or rent personal data. Security is said to follow ISO 27001:2022, and the minimum age is 16. Two gaps stand out for a Swiss company outside the EU: no representative is designated under Article 27, and no data protection officer is named. No subprocessor list is published, and no data processing agreement is mentioned anywhere.
Who owns the data?
The terms draw a clear line between the two sides. Customers keep ownership of what they upload, called Customer Materials, and must warrant that they hold the rights to it. In return they grant Risklick permission to use that content solely to provide and improve the Services. Risklick owns all rights to the Platform itself, including updates and improvements, and users agree not to copy, modify or reverse-engineer it. Both parties commit to mutual confidentiality over non-public business, technical and financial information. Notably, the terms say nothing about who owns the protocols and study documents the AI generates, which is left unaddressed.
Reuse rights
Customers retain ownership of the material they upload, so they may continue to use it freely without asking Risklick for permission. The terms impose no restriction on reusing, exporting or publishing the study documents produced on the platform, and the product is explicitly built to export to Word, Excel and JSON. What the terms do not settle is the status of AI-generated output: no clause assigns ownership of the drafted protocols and study plans either way. Conversely, the licence customers grant Risklick is limited in wording to providing and improving the Services, with no stated permission to sell, rent or publish customer content. The privacy policy separately confirms that personal data is neither sold nor rented. Use of the platform is restricted to the customer's authorised employees or representatives, and acting on behalf of third parties requires written consent.
Data retention & training
Hosting summary
Risklick names no hosting country and no hosting region. The privacy policy states only that data is kept on secure servers, potentially outside the user's own jurisdiction, and that by using the services the customer consents to transfers to countries whose data protection laws may differ. Standard contractual clauses are cited as the safeguard for those transfers. Security measures are described as industry standard and said to comply with ISO 27001:2022 for information security management, although no certificate reference is given. Third parties are mentioned only by category, hosting providers and analytics services, and none is named. The policy adds the usual caveat that no transmission over the internet is completely secure. For a Swiss publisher whose customers handle clinical trial documentation, the absence of any named jurisdiction is the notable point, and it should be clarified contractually before any sensitive material is uploaded.
Things to keep in mind
Risks and trade-offs to weigh before adopting Protocol AI.
- Automation bias: a protocol drafted in minutes still carries regulatory and patient-safety consequences, and reviewers may scrutinise generated text less closely than text they wrote themselves
- Skill erosion: junior medical writers and clinical scientists who start from an AI draft may never build the design reasoning that the work is supposed to teach
- The terms let Risklick use customer materials to provide and improve the services, and nothing says whether that includes training models on your study documents
- No hosting country is disclosed and data may be processed outside your jurisdiction, which matters for study documents and any personal data they contain
- No subprocessor list and no data processing agreement are published, leaving a gap in the due diligence most sponsors are required to perform
- The performance figures on the homepage are unsourced vendor claims, and a launch release quotes a different number, so budgeting on them is risky
- The product lives on a second domain, protocolai.ch, while the legal documents sit on risklick.com, so check which entity and which terms govern your contract
Setup & Integrations
Technical difficulty
Technically light, organisationally heavier. The product is a web application, so there is nothing to install and access is through a browser. The real effort sits upstream and around it: entry is only by demo request rather than sign-up, the terms require your infrastructure to meet unpublished minimum requirements, and you must configure the organisation, the four user roles and per-study user groups. Protocol templates then have to be imported or built, as the Debiopharm pilot illustrates. End-user documentation exists but is not public. With no API, there is no integration work to plan.
Deployment
Integrations
Supported languages
Behind Protocol AI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does Protocol AI do?
Who publishes Protocol AI?
Does it cover medical devices as well as medicines?
What data is the tool built on?
Which standards and formats does it support?
How much does it cost?
Is there a free trial?
Is there an API?
Where is the data hosted?
What is the minimum age and how do I reach support?
Should you pick Protocol AI?
Protocol AI is a narrow, serious tool aimed at a real bottleneck: the months of expert work that go into a clinical trial protocol before a study can start. Its credibility rests on things that can be checked. Risklick AG is a University of Bern spin-off, its advisory board names four clinical experts, and its customer references are dated and specific, notably a completed proof-of-concept with Debiopharm on oncology protocols and a partnership with Citeline. Building on ICH M11 and USDM rather than a closed format is a sound bet for teams that care about downstream interoperability with EDC and eTMF systems.
What the site does not offer is any way to evaluate the commercial proposition on your own. There is no price, no named plan, no trial and no self-service entry: everything runs through a demo request. That is normal for enterprise software sold into pharma, but budget holders cannot size the purchase before engaging. The headline claims of 50% faster protocol development and 20% fewer amendments are the vendor's own and come with no published methodology, so treat them as a hypothesis to test during the demo, not a benchmark.
The transparency gaps around data deserve attention given what the tool handles. No hosting country is named, no subprocessor list is published, no data processing agreement is mentioned, no Article 27 representative is designated even though the publisher sits outside the EU, and the privacy policy has not moved since October 2024. The ISO 27001:2022 claim is worded as compliance and carries no certificate reference. None of this is disqualifying, but all of it belongs on the checklist for the first call: ask for the certificate, the subprocessor list, a DPA, and a written answer on whether study documents are ever used to train models.
- Choosing a selection results in a full page refresh.
- Opens in a new window.