On-site GMP document review for EU medicines manufacturers
2 days on site · 4 hours of training · a report with a source for every finding
Find outdated references and conflicting limits, roles or deadlines across your GMP documents. We review an agreed set of documents at your site with fixed checks and local AI. Your QA team verifies each possible finding and decides.
What a finding looks like
Assay acceptance limits differ between the testing procedure and the specification.
An assay result of 92 % would be inside the limit in the procedure and outside the specification. Found by local AI comparing the content of the two documents. Your QA decides: confirmed, rejected or sent to change control, with the reason recorded.
See the full sample report → Scope, 16 possible findings, suggested actions and space for the QA decision.
From our published test 2 on a synthetic quality system (“Demo Pharma”, 25 documents). Not a client document. The quotes are translated from the Estonian test set.
How we check
Fixed rules check document codes, versions, dates and signatures. Local AI compares the content of documents with each other and with EU GMP text. The work is done at your site, and your documents are not sent to any cloud service. Working copies are deleted before we leave, and by default we take with us only the signed handover record and a work log that contains no text from your documents. Read how we check →
The review stands on its own. If it proves useful, we help you turn it into a second brain for your quality system: a knowledge base on a machine at your own site that local AI can read. See the service →

“In every plant I built, the hardest part of the quality system was not writing it but keeping it consistent.” Veiko Saluste, Qualified Person 1999–2026, reviews your documents himself. Read why he does this →
Ten questions from GMP inspections
Recurring themes from our own inspection experienceAre your records complete, and is every decision written down?
EU GMP 4.8Empty fields, a report without a pass or fail decision, an entry not made at the time of the activity.
Does your APS reflect the interventions and worst cases of real production?
EU GMP Annex 1 §9.34, §9.44 · aseptic manufacturingInterventions that vary from run to run, worst-case conditions left out, incubation conditions not justified.
Does each SOP take effect only after approval, and are the people who use it trained on it?
EU GMP 4.3, 2.11An effective date before the approval date, a new version in force with no training record.
Does the qualified state of your equipment match what is in use today?
EU GMP Annex 15Calibration certificates missing, pressure differentials far from the qualification values, retired equipment still named in SOPs.
Is monitoring data recorded so that trends can be evaluated, and are critical alarms active?
EU GMP 6.9; Annex 1 §4.16 for critical pressure differentialsWarning for a critical pressure differential not active, ventilation data not stored, manual readings never reviewed for trends.
Do the GMP activities that need a written procedure have one?
EU GMP 4.29Typical gaps: IT access rights, entry requirements for clean areas, complaints.
Does your training plan match each person's actual tasks?
EU GMP 2.11Training done but not linked to the job description, effectiveness never assessed.
Does your self-inspection plan cover every GMP area, and is it followed?
EU GMP Chapter 9Areas missing from the plan, audits postponed without a recorded reason.
Does every change reach all the documents it affects?
EU GMP Chapter 1, ICH Q10The SOP was updated; the form, the specification and the training record were not.
Are CAPAs completed on time and checked for effectiveness?
ICH Q10 §3.2.2Overdue actions, effectiveness checks pointed to a process that does not perform them.
Take the ten-question self-check → Five minutes, nothing is stored.