Regulatory evidence for AI medical devices
From first idea to FDA submission, every step read from the record
Precedium is a consultancy for AI medical-device teams. Each engagement sets your device beside every comparable device the FDA has cleared — what they claimed, what they showed, with the page — and hands you a document you can check line by line.
167
product codes with cleared AI devices, each read
3,236
predicate citations traced between cleared devices
13
evidence elements read in every readable summary, with pages
0
statements about what the FDA will decide
From the Device Evidence corpus of FDA clearance summaries for AI-enabled devices, measured September 2026.
The journey
Four engagements, one for each stage of the path to the FDA
Start where you are. Each engagement is fixed in scope, delivered as a document, and sold on its own — most teams begin with S1 and return at the next stage.
- S1
Idea / early design
Regulatory pathway brief
Where does our device fit, and what has the FDA already cleared that looks like it?
- S2
Planning validation
Evidence plan
What will our validation have to show, and what should we ask the FDA before we run it?
- S3
Validation done
Validation gap review
Do our reports state what cleared devices in our code stated, and where are the gaps?
- S4
Ready to file
Submission drafting
Can the submission content be assembled without losing three weeks to tables and checklists?
Regulatory pathway brief
Which product code, which predicate, which pathway — read from every cleared device that came before yours.
What you receive →
Evidence plan
Your study plan set against what cleared devices in your code actually reported — element by element, with the pages.
What you receive →
Validation gap review
Your reports read element by element against what cleared summaries in your code state — before the FDA reads them.
What you receive →
Submission drafting
The record half of a 510(k) or De Novo, assembled in days — every drafted sentence sourced and reviewed by a person.
What you receive →
Who it is for
The person with the deadline
The work is bought by whoever owns the date: a fundraise, a Pre-Submission meeting, a filing.
S1
Founders without a regulatory lead
Which code, which predicate, which pathway — before the next investor meeting.
S2
Regulatory and clinical leads
What the study must show, and the questions for the Pre-Submission meeting.
S3
Teams with a filing date
A second reader for the reports before the FDA becomes one.
S4
Small teams assembling a 510(k) or De Novo
The record half of the submission, assembled and sourced in days.
The method
The public record does the reading; a person does the judging
- 1
The record
Device Evidence has read the clearance summary of every AI-enabled device on the FDA's list for thirteen evidence elements, traced each predicate chain, and kept the page of every finding. It is public and free.
- 2
Your documents
Sent under a signed confidentiality agreement, read on the consultant's machine, never uploaded to a server, and deleted by a date you agree.
- 3
The deliverable
Every statement footnoted to an FDA page or to your document. Drafted text is reviewed sentence by sentence; anything without a source is listed, not included.
The consultant
Dr. Rafiq Muhammad, MBBS, MIHMEP, PhD
A medical doctor with a PhD in artificial intelligence in healthcare from Karolinska Institutet and a master's in healthcare management, economics and policy from SDA Bocconi. He has drafted regulatory submission documents for medical devices for the US FDA, and built Device Evidence and Precedium.
What we never say
No deliverable predicts the FDA's decision on a device — no clearance, grant or approval is promised, and no probability is given. The record describes past filings, not yours. The FDA decides; your regulatory signatory submits. That restraint is why the documents can be trusted.
Start with a free pathway brief
The first S1 brief for your device is free. A twenty-minute call is enough to start: what the device does, where you are, and the date you are working to.