Annex 11 with AI — every change triggers revalidation, and your quality team can no longer keep up.

The problem

The validation wheel never stops — and the rate of change only goes up.

Annex 11 was written when computerized systems were stable islands: one MES, one LIMS, one SCADA, annual patches. Today the reality is different: new integrations every quarter, quality dashboards that change, connectors with suppliers, cloud modules, AI in specific parts of the process. And every one of those changes, seen through Annex 11, raises the question “does this need revalidation?”. Here is how the quality team feels it:

1. Every minor change opens a file — and the file takes weeks, even if the change was a single parameter.
2. The IQ/OQ/PQs pile up — and the person qualified to sign them is the same person who has to write them.
3. Business pressure does not wait — so things end up approved with half-written dossiers, which multiplies the risk of a finding at inspection.

It is not that the quality team is slow. It is that the validation model was designed for a pace that no longer exists, and nobody yet had the tool to do it any other way. That is what has just changed.

How it fits the IRIS system

iLEAN does not replace QA — it takes the dumb work away so QA can sign the changes that matter with real judgment.

The Annex 11 promise is not to validate less; it is to validate well and leave an impeccable documentary trail. The problem is operational: nobody had agents capable of writing and maintaining the dossier while the system changes. That is what iLEAN does, without asking you to change your validated MES or your LIMS.

Tracer captures every change, every access, every release. The Writer agent drafts the GAMP 5 evidence pack. The quality team reads, adjusts and signs — on a dossier that is already built, not on a blank page.

The two iLEAN pieces applied to Annex 11 / continuous CSV:

• Tracer (audit trail + change capture) — reads the logs of your validated systems (MES, LIMS, SCADA, IoT, MOM), the configurations of every release and user access records, and normalizes them into an auditable format. Every change is tied to an author, a timestamp, a business reason and a potential impact — the four pieces of data Annex 11 asks for and that are almost never in the same place.
• Writer agent — takes everything Tracer captures and writes the evidence pack: change description, impact analysis, test plan, results, conclusion and recommendation to QA. Aligned with GAMP 5 category 4-5 and with ALCOA+. QA receives the dossier already structured: reviews it, adjusts whatever needs adjusting and signs. They do not start from zero.
• Connect (where applicable) — captures the changes that arrive by email from the software vendor, the service desk tickets and the integrator's WhatsApp messages, and puts them into the file at second zero. What used to be discussed by email stops getting lost along the way.

See the full IRIS architecture →

Before and after

CSV by heavy milestones vs. continuous CSV with iLEAN

Aspect	Classic CSV + Word + folders	With iLEAN Tracer + Writer
Minor change to a configuration	Opens a file, IQ/OQ/PQ, weeks	Evidence pack generated, QA reviews and signs
Audit trail	Scattered across logs, exports and spreadsheets	Normalized, tied to author, timestamp, reason, impact
GAMP 5 alignment	Rebuilt release by release	By design, built into every change
Legacy systems	Every touch triggers a full revalidation	iLEAN reads, it does not modify — legacy untouched
Inspection readiness	Weeks of scrambling the night before	A live dossier, exportable in minutes
Load on the QA team	The structural bottleneck	QA decides on a finished dossier — it does not write it

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 plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Pharma plant with a validated MES and LIMS, a legacy SCADA and between 20 and 100 computerized-system changes a year.
• Tracer + Writer pilot on the MES and a couple of satellite systems. First value expected within a few weeks: a validation dossier generated automatically for minor changes.
• Indicative payback between 4 and 9 months, depending on the fully loaded cost of the QA team and the volume of changes subject to revalidation.
• A ≥30% reduction in quality hours spent on validation documentation. The hard lever is freeing QA from the dumb work so they can sign off on the major changes with real judgment.

And the quality manager's reasonable doubt

“What if the AI hallucinates in the validation dossier?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data (reading a change log and packaging the evidence), the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN's three safety rings ensure the dossier goes to QA, and QA signs. The AI proposes, the person releases — exactly what Annex 11 asks for. See the architecture.

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