
Since the last issue, two very different questions kept coming back to the same answer. Who assembles a regulatory document? And when do a handful of complaints become a signal? In both cases, the work that can be structured should be, so that the people who decide have the time and the evidence to do it.
Lead Story: less assembly, more judgment
In a regulatory document, the expensive part is rarely the judgment. It is the assembly. Since 2013 our consultants have written and reviewed the documents regulated companies produce again and again: PSURs, annual product reviews, clinical study reports, submission modules. More than 500 projects later, the pattern is familiar. The same data is collected from several systems, copied into the same template and reformatted for every cycle. Expert time goes into putting the document together instead of into deciding what it says.
AI can take over much of that assembly. In a regulated document it only helps if three things hold: the draft starts from the company's own template, every statement can be traced to its source, and the people who sign stay the people who decide. That is why we have partnered with AriaFlow.
AriaFlow is an AI-powered platform for preparing and reviewing regulatory documents. It brings information from multiple sources together with existing templates to produce structured drafts for expert review. In practice that comes down to four things:
- Start with your own template and processes or choose from the catalog. Turn an existing document template into a reusable workflow, or adapt a catalog template to your team's requirements.
- Verify what the AI produces. Follow citations back to the supporting evidence and see where information is missing or sources disagree.
- Get started in the Word environment your team already knows. Generate, guide, edit, and review drafts directly in Microsoft Word.
- Make each review count for future drafts. Expert edits can inform proposed changes to templates and prompts. Your team reviews and approves those changes before they are used in future drafts.
What a platform cannot bring is the regulatory side: knowing which document to start with, validating the use in a GxP environment, reviewing the drafts. That is our part. There are two ways to work with us. Our experts deliver the documents with AriaFlow as their engine, as a project, as a functional service or as part of your team. Or you license AriaFlow, and we select the use case, prepare supplier qualification and risk-based validation under GAMP 5, write the templates and review the drafts.
A first project starts small, with one recurring document. We agree the document type, your template, representative sources and who signs off. A pilot then produces real drafts, both sides review them, and the effort is compared with your previous cycle before further document types follow. Typical candidates are clinical evidence summaries, V&V summary reports, biocompatibility summaries, MDR technical documentation, PSURs and PBRERs, and clinical study reports.
Why now? AI is arriving in regulated work while the rules for it are still being written, from the EU AI Act to the draft GMP Annex 22. We want our clients to use it in a way that holds up in an inspection.
Request a live demo on one of your own document types
Field Notes: Barcelona, with AriaFlow at the table
On 5 and 6 November we were at the RAPS Europe AI in Regulatory Conference in Barcelona, together with AriaFlow, at a tabletop on the floor where the coffee breaks and lunch took place. In the World Café session on post-market surveillance we showed AriaFlow live: how a recurring regulatory document goes from your sources and templates to a structured first draft, and what stays with your experts.
No question came up more often at our table than this one: "Does this cover my specific case?" A demo of a finished document no longer impresses anyone. Teams want their own document, built their way. Two things changed the conversation. The first was the template catalog, and that a team can just as well start with its own template and processes. The second was the evidence: every statement in the draft can be followed back to its source, and the draft shows where information is missing or sources disagree. If we did not get to talk in Barcelona, the conversation is still open.
Capability Spotlight: when complaints become a trend
Twelve complaints a year, spread over 40 products. Hardly any per product, and still a signal. Alpenrose Medizintechnik is fictional, like Nordlicht and Orbit, the other two manufacturers in the case: 60 feedback reports a year on three rapid tests, and 9,000 complaints from 40 countries. All three have the same obligation to spot trends, and none of them can do it with the same method.
A signal is not automatically a problem. There is no single threshold that turns a pattern into a trend. What decides is whether risk acceptance tips over, not the size of the change. That logic has to be built into complaint handling: grouping what belongs together, finding signals in small numbers, and tying every trend to the risk assessment and post-market surveillance.
Quality Management & Compliance: this is what we set up with manufacturers, from complaint intake to a trending logic that spots recurring product problems in the data instead of reacting case by case. How we set it up
Regulatory Roundup
Japan: a pharmacy route with seven conditions. On 1 September 2026 the MHLW added starch derived absorbable topical hemostatic material to the list of specified insured materials that a pharmacy may supply without a separate license to sell highly controlled medical devices. The route only exists with a physician prescription, supply under the social insurance laws and a pharmacy that meets every specified requirement. It does not amend approval, does not create a new indication and does not change the reimbursement value. For a manufacturer, the work is channel work: which pharmacies can rely on the route, and who verifies, records, trains and routes complaints. The seven conditions in detail
US: FDA's surgical-robotics draft, open for comment until 24 November. The draft guidance on robotically-assisted surgical devices (25 September 2026, docket FDA-2026-N-9505) proposes an evidence framework for locally teleoperated surgical robots, and "local" is defined by the room: operator console and bedside subassembly in the same operating room. Remote teleoperation and robots that perform significant aspects of a procedure on their own are outside the draft; FDA takes up exactly those systems at a public workshop on 2 and 3 December. The umbrella and covered-procedure approach can reduce procedure-specific in vivo data, but FDA names no set list of suitable procedures, so the evidence has to be built around the specific claim. Diana Hohage's reading of the draft
Looking Ahead
Wednesday 25 November, 11:00 to 12:30 CET, online. Diana Hohage (Principal Consultant, Entourage) speaks in the DQS Med Xpert Lab series on complaint trending: from complaint to risk-based decision, with the three fictional manufacturers from above and the resolution we did not give away. It is held in German, and registration is on the DQS site. Webinar details
Friday 27 November, EUDAMED. Devices already on the market before 28 May 2026 that are still being placed on the market must be registered in EUDAMED no later than twelve months after the Commission's notice of 27 November 2025. The Commission's timeline names 28 November 2026, a Saturday, so Friday 27 November is the safe last day. What to check before then
Tell us which regulatory document your team prepares again and again. We show you on that document type where the assembly can be structured and what stays with your experts.