What changed
FACT (per convergence description): two June 2026 FDA classification rules created new class II special-controls categories for AI radiology/monitoring software, and PCCP guidance is finalized — converting each post-clearance model update into a recurring, auditable documentation duty (verification data, performance thresholds, labeling updates, traceable records) plus ongoing real-world performance monitoring.
Why now
The duty just became concrete and recurring rather than hypothetical: every model release by a PCCP-cleared device maker now requires a dossier executed exactly per the pre-approved protocol, and the obligated class grows with each new special-controls classification. HYPOTHESIS: the window is before eQMS incumbents ship a pipeline-integrated PCCP module.
Converging signals
(1) New special-controls classifications for AI imaging/monitoring devices; (2) finalized PCCP guidance making per-update evidence mandatory; (3) INFERENCE: a long tail of seed/A-stage AI-device startups with no regulatory-affairs department now carries this duty. NOTE: the signals array and demand_evidence array in this input are EMPTY — no primary source text was provided, so the classification-rule facts are taken on the convergence description's authority and everything downstream is hypothesis.
Customer pain
HYPOTHESIS (explicitly untested — zero demand_evidence supplied): founders manage PCCP/update evidence in ad-hoc docs and spreadsheets, burning 10+ engineering hours per model release and risking clearance-invalidating drift outside the approved change envelope. The convergence's own testable prediction (4 of 10 interviewed startups confirm) has NOT been run.
Who pays
Seed/Series-A AI radiology and patient-monitoring software companies with a cleared or in-process 510(k) citing a PCCP — the founder, VP Engineering, or fractional regulatory consultant. Plausibly dozens-to-hundreds of firms; the FDA 510(k) database can enumerate them by name, which is a genuinely reachable, listable buyer set.
Solved today
HYPOTHESIS: general eQMS platforms (Greenlight Guru, Matrix Requirements, Ketryx, Enzyme) plus fractional regulatory consultants plus hand-built spreadsheets and Confluence pages; Ketryx in particular already markets AI/ML-lifecycle compliance automation for medical software.
Why current solutions are bad
eQMS tools are document-centric, not ML-pipeline-native; they don't watch the model CI system, don't auto-assemble per-release verification evidence against the approved change envelope, and are priced/structured for firms with QA staff. Consultants are per-hour and don't scale with release cadence. UNPROVEN: that small firms feel this acutely enough to buy a third tool.
Proposed product
A 'PCCP evidence vault': CI/CD integration (GitHub Actions, MLflow, W&B) that on each candidate model release runs the pre-approved verification protocol, checks performance thresholds, assembles the FDA-structured update dossier, maintains the cumulative audit file, and hard-flags any change that exits the approved change envelope before it ships.
MVP version
A pipeline-agnostic CLI/GitHub Action + hosted vault that ingests eval metrics and artifacts, renders one FDA-expected dossier template per release for ONE device category (AI radiology triage), with an envelope-drift checklist. No eQMS integration, no real-time monitoring in v1.
30-day build
Run the falsification tests FIRST: query the FDA 510(k) database for clearances citing PCCPs and build the named prospect list; audit Greenlight Guru/Ketryx/Matrix for a shipped pipeline-integrated PCCP module; interview 10 startups from the list against the 4-of-10 prediction. Kill or proceed on that evidence.
60-day build
If validated: build MVP with 2-3 design partners from the interview set at a founding-customer rate; get a fractional regulatory consultant (paid, founder has budget) to bless the dossier template against the final PCCP guidance.
90-day revenue plan
Convert design partners to $400-800/mo subscriptions; sell via the FDA-database-derived named list and the fractional-RA-consultant channel (consultants recommend tooling to multiple clients). Realistic first revenue day 90-150 — inside the founder's 180-day window only if validation starts immediately.
Distribution path
Named-account outreach from the public 510(k)/PCCP list (small, enumerable universe — a strength); RAPS forums, MedTech LinkedIn, and fractional regulatory consultants as a reseller/referral channel. No ad spend needed, but this is credibility-gated outreach, not self-serve.
Pricing hypothesis
$400-800/mo per device program, or per-release fee (~$500/dossier) mirroring the founder's proven per-filing model. HYPOTHESIS — no willingness-to-pay evidence supplied.
Technical difficulty
Moderate: dossier templating, CI hooks, and threshold checks are solo-buildable in weeks. The hard part is regulatory correctness of the template — buyable via consultant review, which the founder can now fund.
Legal / regulatory risk
Material but manageable: the tool must not misstate what FDA expects (customers' clearances are at stake). Product is documentation tooling, not a medical device — no clearance needed itself, but errors create liability exposure and reputational kill risk in a small community.
Platform dependency
Low. FDA rules are public; CI integrations are commodity. Main dependency is the stability of PCCP guidance, which cuts favorably (more rules = more obligated buyers).
Founder fit
MIXED, and weaker than it first looks. Matches the proven pattern only partially: a mandate compels a class of actors — but the duty is maintaining an internal audit dossier, NOT submitting through a government portal per-transaction, so the ELDT wedge (portal automation, per-upload fee) doesn't transfer directly. It also collides with two stated avoidances: heavily regulated medical products (adjacent, not identical — this is tooling for regulated firms) and multi-year trust-building (FDA-adjacent tooling sold to firms whose clearance is on the line is trust-heavy). Strengths that DO apply: reading a federal mandate, systems/automation thinking, demonstrated-value selling to an enumerable buyer list.
Breakout potential
Real if the pattern generalizes: each new special-controls classification adds an obligated cohort, and the vault could extend to EU AI Act technical documentation. But incumbents (especially Ketryx) are natively positioned to absorb this feature.
Final recommendation
DO NOT BUILD YET — VALIDATE. This is a well-formed hypothesis with an enumerable buyer list and a real structural mandate, but it currently has zero attached demand evidence, a live falsification threat from Ketryx-class incumbents, and only partial founder fit (mandate-shaped, but not portal-shaped, and trust-heavy). Spend 2-3 weeks and modest budget running the stated evidence plan before writing any product code.
Next action
Query the FDA 510(k) database for clearances referencing PCCPs to size and name the obligated class; simultaneously audit Ketryx, Greenlight Guru, and Matrix Requirements for a shipped PCCP update module; then interview 10 startups from the list against the 4-of-10 pain prediction.