What changed
FACT (cited rule): FDA published a final classification order (June 17, 2026) placing radiological machine-learning-based quantitative imaging software WITH a predetermined change control plan into class II with special controls. This creates a codified pathway where model updates ship under a pre-approved plan instead of a fresh 510(k).
Why now
FACT: the rule is weeks old, so the first cohort of vendors clearing under this device type is forming right now. INFERENCE: no incumbent packages PCCP per-update evidence as a lightweight per-device product for the long tail yet; the window is before eQMS incumbents bundle it.
Converging signals
One strong regulatory signal (the classification rule, FORCED BUYER) plus a pattern-transfer hypothesis (registry-gated recurring paperwork). No independent PAIN or HIRING/SPEND signals were provided, so the recurring-pain claim is currently a HYPOTHESIS, not evidence.
Customer pain
INFERENCE: each retrain/update under a PCCP triggers a recurring evidentiary duty — execute the pre-approved protocols and keep inspection-ready records. Deviation risks the device being deemed adulterated or forcing a new 510(k). Small vendors (likely <50 headcount) have ML engineers but no regulatory ops staff to produce this dossier every cycle. Unverified until discovery calls confirm it.
Who pays
The regulatory affairs lead or founder/CTO at a small AI radiology/quantitative-imaging software company that cleared (or is clearing) under this class II pathway. FDA's public AI-enabled device list enumerates the exact companies — a self-updating lead list.
Solved today
INFERENCE: general eQMS suites (Greenlight Guru, Ketryx, Matrix Requirements), regulatory consultants billed hourly, or ad-hoc spreadsheets/Confluence maintained by engineers. eQMS covers document control broadly but is not purpose-built around the PCCP update loop (retrain → protocol execution → evidence capture → passport for hospital procurement).
Why current solutions are bad
eQMS is priced and shaped for the whole quality system (annual contracts, heavy implementation), consultants are per-hour and don't scale with monthly retrains, and none produce a shareable 'update passport' hospital procurement teams can accept. HYPOTHESIS — must be falsified in discovery: incumbents may already cover this adequately.
Proposed product
Per-device subscription 'PCCP Evidence Locker': ingests each model update event (CI hook or manual upload), walks the vendor through their own pre-approved protocol checklist, captures validation-run outputs and performance-monitoring logs, versions rollback records, and renders an inspection-ready dossier plus a shareable compliance passport for hospital/GPO procurement.
MVP version
A structured dossier generator: template the special-controls evidence requirements from the codified rule, let a vendor register a device + its PCCP protocol, log one update cycle end-to-end, and export a polished audit PDF/portal. No deep ML-pipeline integration in v1 — manual upload of artifacts is acceptable.
30-day build
Validation only, before building: scrape FDA's AI-enabled device list and the new product-code listings; cross-reference LinkedIn headcounts to confirm the long tail; send 20+ outreach emails to regulatory/founding contacts; run ≥3 discovery calls probing who currently produces per-update PCCP evidence.
60-day build
If ≥3 calls confirm unstaffed recurring documentation work: build MVP with the codified special controls as the template spine; recruit 2-3 design partners at steep discount in exchange for protocol samples and inspection-readiness feedback.
90-day revenue plan
Convert design partners to paid at $300-600/device/month; begin outbound to every new clearance appearing on the FDA list (the list itself announces each new prospect). Realistic first paid contract in the 90-150 day range given regulated-buyer diligence.
Distribution path
Outbound to an enumerable, public, self-refreshing roster (FDA device listings / AI-enabled device list) — exactly the founder's proven ELDT motion. Secondary: RAPS/regulatory-affairs communities, content on 'what your PCCP obligates you to document,' and hospital procurement teams who can pull vendors toward the passport.
Pricing hypothesis
$300-600 per device per month, or per-update packs; anchor against a fractional regulatory consultant ($200+/hr) and eQMS seats ($15k+/yr). Per-device aligns with how the obligation accrues.
Technical difficulty
Moderate: forms/workflow engine, artifact storage, versioning, PDF/portal rendering, optional CI webhooks. Well within solo AI-assisted build capacity. The hard part is regulatory correctness of the templates, which likely requires a fractional RA consultant (founder has capital for this).
Legal / regulatory risk
Meaningful but manageable: the tool documents compliance, it is not itself a medical device. Must avoid implying FDA endorsement or guaranteeing inspection outcomes. Errors in templates could contribute to a customer's adulteration finding — needs disclaimers and a consultant-reviewed template spine.
Platform dependency
Low. Depends only on public FDA data and the customer's own pipeline. No app-store or platform approval anywhere.
Founder fit
Mixed. Matches the proven pattern partially: a regulation creates recurring compelled paperwork and a public registry enumerates the buyers (like ELDT). But unlike ELDT there is NO government portal to submit into — the deliverable is internal inspection-readiness, and the buyer is a regulated medtech company that must trust its evidence vendor. Founder has no medtech/RA track record, which raises the credibility bar in a way the ELDT play did not. The applicable lesson (gov-portal mandate = 8-9 fit, confidence 0.80) only half-applies; scored down accordingly.
Breakout potential
Good if validated: PCCPs are expanding across FDA's software-device universe beyond radiology, so the same locker generalizes to other SaMD categories; the passport could become the artifact hospital procurement standardizes on.
Final recommendation
VALIDATE BEFORE BUILDING — do not kill, do not build yet. The forced-buyer mandate is real and freshly codified, the lead list is public and enumerable, and the product shape fits a solo builder. But the two load-bearing claims (long tail of small vendors; per-update documentation is unstaffed pain not already absorbed by eQMS) are both unverified hypotheses. The stated testable prediction is cheap (~2 weeks, ~$0) and decisively resolves go/no-go.
Next action
Scrape FDA's AI-enabled device list + new product-code clearances, join against LinkedIn headcount, and send 20 discovery emails to regulatory contacts at sub-50-person imaging-AI vendors asking how they produce PCCP update evidence today; book ≥3 calls.