EU CPNP cosmetic notification: the dossier is no longer weeks of copying data between spreadsheets.

The problem

Five sources that don't talk to each other, and a submission that allows no errors.

Loading a cosmetic into the CPNP means stitching together five realities, each living on its own island:

1. The qualitative-quantitative formula — in the R&D software, or in the lab spreadsheet only its keeper understands.
2. The composite INCI — assembled piece by piece from each supplier's technical data sheets, every one in its own format.
3. The Annex III allergens — with their thresholds (0.001% in rinse-off, 0.01% in leave-on for declarable fragrances), which have to be validated against the actual composition.
4. The safety tests — challenge test, stability, packaging compatibility, microbiology — in PDFs from the external lab.
5. The CPSR — the assessor's report, which has to cite all of the above coherently and in the format of Annex I of the Regulation.

The regulatory team knows it, but they cannot be in five places at once. And at launch time — campaign ready, retailer waiting — the bottleneck is the dossier. What gets loaded into the CPNP is the official truth of the product. A coherence error between INCI, formula and CPSR forces you to notify all over again, with everything that drags behind it.

How it fits the IRIS system

iLEAN does not add a sixth system — it seals the cracks between the five you already have.

The CPNP dossier does not fail for lack of information: it fails because the information lives on islands and gets stitched together by hand every time. iLEAN acts as the putty that fills the gaps between R&D, regulatory, supplier and external lab — without asking you to change your formulation software or your Annex I templates.

Connect captures every source wherever it lives. iLEAN Writer drafts the dossier with a traceable citation to the original document. The assessor reads, corrects and signs. The CPNP submission always carries a human signature.

The specific pieces:

• iLEAN Writer — an agent specialized in compiling regulatory dossiers. It cross-references the formula, supplier data sheets, lab tests and the Annex I template, and delivers a navigable draft with every data point linked to its original source. AI reliability on anchored tasks (recontextualizing a data point from one system to another) drops below a 1.5% error rate in the best models^[1].
• Connect — captures the five sources wherever they are: R&D software, lab spreadsheet, supplier PDFs, the external lab's email, the assessor's templates. It does not force you to change anything — it adapts.
• Three safety rings — Writer drafts in the outer ring at full power. The assessor signs in their own environment. What gets loaded into the CPNP is what the person approves, not what the AI infers.

See the full IRIS architecture →

Before and after

Manual dossier vs. dossier compiled with iLEAN Writer

Aspect	Manual compilation	With iLEAN Writer + Connect
Formula source	R&D software or a spreadsheet — copied by hand	Direct capture, every data point cited to source
Composite INCI	Assembled piece by piece from supplier sheets	Generated and validated against Annex III at second zero
Allergen coherence	Manual cross-check in the final review	Automatic verification against Regulation thresholds
CPSR draft	From a blank template, days per SKU	Navigable draft ready for the assessor
Errors between versions	Frequent in parallel launches	Batch-to-document traceability; changes flagged
Regulatory time per SKU	Weeks of copy-paste + review	Hours of review over a structured draft

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your portfolio. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Brand with a catalog of several dozen active SKUs and recurring EU launches (skincare, color, hair care).
• Writer pilot on the family with the most launches per year (the low-hanging fruit). First navigable, source-traceable draft within a few weeks.
• Indicative payback between 4 and 9 months, depending on annual notification volume, the regulatory team's hourly cost and the % of rework caused by INCI/CPSR coherence errors.
• Expected reduction in regulatory time per SKU ≥ 30% — because the draft is no longer written from scratch. A defensible floor; we refine the ceiling once we have your SKU Pareto.

And the CAIO / regulatory lead's reasonable doubt

“What if the AI makes up a data point and we end up with a wrong INCI in the CPNP?” — hallucination is a problem of free generation, not of tasks anchored to a source. iLEAN Writer does not generate formulas: it reformats data that already exists in your systems, with the citation to the source document visible. The best models brought the error in that class of tasks below 1.5% ^[1]. And the CPNP submission does not go out without the responsible person's signature — the three rings are there precisely for this.

^[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.