PharmaLectures.com

Courses › Pharma industry foundations › Computers in a GMP world: Annex 11, GAMP 5 and the EU AI Act, explained for pharma

Computers in a GMP world: Annex 11, GAMP 5 and the EU AI Act, explained for pharma

Your team explains every algorithm to an inspector, and your rollout dates hold.

GAMP 5 is the ISPE guide to Good Automated Manufacturing Practice, the most widely used method for validating computerised systems in pharma. Its second edition was published in 2022. It sorts software into categories by how much it is configured or custom-built, and scales testing to patient risk. GAMP 5 is industry guidance, not law: in the European Union the binding rule is Annex 11 of EudraLex Volume 4, and the EU AI Act, Regulation (EU) 2024/1689, now adds duties for some artificial intelligence systems. This course ties all three together.

  • 8 audio modulesthree lengths each
  • 1 h 20 min to 6 h 40 minbriefing to expanded
  • Workbook, transcript, certificateand a recorded quiz per module
  • IT, QA, validationwho it is written for

Why GAMP 5 training matters

An inspector asks how your automated batch review tool handles an edge case. Your validation engineer cannot tie the algorithm to EudraLex Volume 4 Annex 11, and the room goes quiet. The finding is major, the rollout stops, and the quality protocols get rewritten.

What changes after this course

Your engineers write validation plans that answer the inspector's questions before the audit starts. They map each algorithm to the European Union Artificial Intelligence Act and Good Automated Manufacturing Practice 5 (GAMP 5). You hold a certificate and a training record for every learner.

Watch the pitch

3:43 on how an audio course is bought, listened to and recorded. Full screen, one click.

Watch the pitch3:43, full screen, one click

GAMP 5 training: what this course gives your people

A new machine learning tool arrives and IT no longer argues about its category. The team places it in a GAMP 5 category, writes the risk assessment in one sitting, and links each choice to Annex 11. The paper trail exists before anyone asks for it.

Quality Assurance reads the validation package once and signs. The data governance section already shows how inputs, training data and outputs meet EudraLex Volume 4 Annex 11, so nothing comes back for rework. Worry about hidden bias in batch review drops, because the validation master plan states how stability is checked.

When the inspector asks about continuous compliance, the validation engineer opens the monitoring records and shows live performance against the limits your quality system set. The inspector sees a system in a validated state and moves on to the next topic.

Your Head of Digital Transformation stops losing arguments with regulatory affairs. Each new tool is checked against Article 6 and Annex III of the European Union Artificial Intelligence Act at the start, so the answer to whether you can deploy it arrives with its reasons attached.

Who needs GAMP 5 training

  • Information Technology Manager

    Your developers say quality rules slow them down. Now the team builds with Annex 11 validation and audit trail expectations in mind from the first sprint, and the rework stops.

  • Quality Assurance Specialist

    An inspector asks for the audit trail behind a machine learning decision. You hand over a record that already answers it in Annex 11 terms.

  • Validation Engineer

    You face a neural network and a blank validation plan. You place it in a GAMP 5 category, choose the testing, and write the plan to match.

  • Head of Digital Transformation

    The board wants faster automation, and regulatory affairs keeps braking. You bring a compliance route both sides accept, with the risk class of each tool already settled.

Roles: Quality assurance and Qualified Person, IT, data and validation

Businesses: every kind of pharma company

Countries: India, China, Türkiye, Korea, Japan, Taiwan, Egypt, Gulf and MENA, Brazil and Latin America, Pakistan, Bangladesh and Vietnam, United States, Europe

Selling into: Europe and the UK

Courses are in English. Other languages on request.

After this course your people can

  1. Classify a new software tool against the risk categories of the European Union Artificial Intelligence Act and record the reasoning.
  2. Write a validation master plan that covers machine learning models, from training data to release.
  3. Show an inspector the audit trail and data checks behind an automated decision, using Annex 11 wording.
  4. Place an adaptive algorithm in its GAMP 5 category and scale the validation effort to its risk.
  5. Document the training and testing of predictive maintenance software so Quality Assurance can approve it.
  6. Set up monitoring that tracks model performance over time and tells you when to retrain or withdraw it.

What the company gets

  • Your plants deploy automated tools with a validation file that already answers the first questions an inspector asks.
  • Quality approves new software rollouts from complete packages instead of sending them back for another round of revisions.
  • Inspectors find a clear line from each algorithm to the requirements of EudraLex Volume 4 Annex 11.
  • Rollout dates hold, because incomplete validation plans stop being the reason a launch slips.

The evidence: questionnaire, training record, certificate

EU GMP Chapter 2, section 2.11, and 21 CFR 211.25 both ask for training that is documented and shown to work. This is what each learner leaves behind.

A recorded questionnaire per module

Every one of the 8 modules ends with a multiple-choice questionnaire. The result is recorded per person with a timestamp, whichever of the three lengths the person listened to.

A per-person training record

The export lists each learner, each module, the result and the date, which is the training record EU GMP Chapter 2 and 21 CFR 211.25 ask you to keep and the first thing an inspector asks for.

A certificate per learner

Name, company, course, hours, the guideline version taught and the date. It documents in-company training; it is not a professional qualification or an accreditation.

Your pass mark, your rule

There is no fixed pass mark. Your quality or human-resources manager decides what counts as passed and when a module is retaken, exactly as your training procedure says.

Delivered within 72 hours of payment, with the workbook, the full transcript and a source for every fact in the audio.

How long it takes, by who is listening

Nobody spends a day in a classroom. Every module exists in three lengths; a person takes the length their role needs, and every length ends with the same recorded questionnaire, so the training record is complete either way.

Briefing1 h 20 min10 min a module, for executives, sales and anyone who must know the rules, not run them
Standard3 h 20 min25 min a module, for regulatory, quality and business-development staff
Expanded6 h 40 min50 min a module, for technical staff who run the procedure day to day
Who takes this courseLengthTotal time
IT, data and validationBriefing1 h 20 min
Quality assurance and Qualified PersonStandard3 h 20 min

GAMP 5 course outline: 8 modules

8 modules. Under each title are the three things the listener can do afterwards; the audio itself runs 10 to 50 minutes depending on the length chosen.

  1. 1Map the European Union Artificial Intelligence Act to your operations

    • Why the same algorithm can fall into a different risk tier depending on the decision it supports.
    • How to build a one-page system inventory that shows which tools the Act touches and which it leaves alone.
    • What to do when IT, QA and the vendor each read the Act's requirements differently.
  2. 2Apply EudraLex Volume 4 Annex 11 to your artificial intelligence systems

    • How to show an inspector the audit trail behind a model's output, in Annex 11 terms.
    • Why Annex 11 expects built-in checks on data exchanged electronically, not only on data people type.
    • How to apply Annex 11 change management so a model update gets the right level of re-validation.
  3. 3Structure your validation with Good Automated Manufacturing Practice 5

    • How to place a machine learning component in a GAMP 5 category without a long debate.
    • Why risk-based scaling lets you skip scripts that add pages but no assurance.
    • What to do when the supplier trained the model and will not show you the training data.
  4. 4Pinpoint the high-risk artificial intelligence systems in drug production

    • Why a visual inspection tool on the packaging line and one in quality control may land in different places.
    • How to test a system against Article 6 and Annex III of the Act and write down why it passed or failed.
    • What to do when a predictive maintenance tool starts influencing a step that touches product quality.
  5. 5Create data governance that meets Annex 11 and the Artificial Intelligence Act

    • How to trace a prediction back to its source records so the inspector stops asking for more.
    • Why historical batch data carries bias, and the check that exposes it before the model learns it.
    • The ownership rule that tells you who approves a change to the training data, QA or IT.
  6. 6Write a validation master plan for artificial intelligence systems

    • The sections of a validation master plan that an adaptive model needs and a fixed program does not.
    • How to set acceptance criteria for a model that sets production parameters, before anyone trains it.
    • What to put in the release record so the inspector's first question has a page number.
  7. 7Set up continuous monitoring that keeps software compliant after go-live

    • How to tell a drift alert from a change request, so monitoring does not flood your change control.
    • The handful of performance measures that show the model is still in its validated state.
    • When to retrain, when to roll back, and when to take the model off the line.
  8. 8Answer inspector questions during an artificial intelligence system audit

    • The three documents to have on the table when the inspector hears the words automated decision.
    • How to explain model weighting to an inspector who learned on rule-based systems, in two minutes.
    • What to say when the question is how you stop data drift from changing product quality.
The founders

One of us was inspected. The other one pays for the training.

Every claim below is sourced on the about page: the medicines register, the Lithuanian Pharmaceutical Union board list, the university directory.

Justina Krukonytė-Racz

Justina Krukonytė-Racz

Co-founder. Licensed pharmacist.

GDPResponsible PersonWarehouse & logistics

Licensed pharmacist. Responsible Person at Tamro (PHOENIX group), where she led the GDP set-up of the distribution centre.

The Responsible Person is the named person the regulator holds accountable for every pallet that leaves the building.

She writes the modules around the procedures a GDP inspection checks. Board member of Lithuanian Pharmaceutical Union.

Listed in the Lithuanian medicines register as Pharmacist.

Prof. Ben Racz

Prof. Ben Racz

Co-founder. Builds the platform and is responsible for delivery, records and pricing.

Platform & deliveryQuantitative researchCarnegie Mellon

He was Professor of the Practice at the College of Engineering at Carnegie Mellon University, where he taught Advanced Quantitative Finance and Algorithmic Trading to MSc students.

He is the founder and Managing Director of Algorithmic Sciences, a scientific research and consulting firm. Earlier he co-founded a data-driven growth consultancy, a data-science based wholesale trading company and an information security firm.

More about Ben at benracz.com

Questions about GAMP 5 training

Is GAMP 5 a legal requirement?

No. GAMP 5 is guidance published by ISPE, an industry association. What inspectors enforce in the European Union is Annex 11 of the EU GMP guide, and in the United States 21 CFR Part 11 and the drug GMP rules. GAMP 5 is widely used because it gives a practical, risk-based way to show that a computerised system meets those rules, so inspectors expect to recognise its logic.

What are the GAMP 5 software categories?

GAMP 5 sorts software into category 1 for infrastructure software, category 3 for standard products used as supplied, category 4 for configured products, and category 5 for custom software. Category 2 was dropped in GAMP 5. The higher the category, the more specification and testing are normally needed, always scaled to the risk the system poses to patients, product quality and data integrity.

Is there a certificate?

Yes. Every learner who finishes the course receives a certificate in their own name, and every module ends with a recorded multiple-choice questionnaire, so the certificate sits on top of a per-person training record. The certificate documents in-company training; it is not a professional qualification or an accreditation, and your quality or human-resources manager sets the pass mark and decides when a module is retaken.

How long does the course take?

8 modules, each in three lengths: a 10-minute briefing, a 25-minute standard run and a 50-minute expanded walk-through. The whole course takes 1 h 20 min in the briefing length and 6 h 40 min in the expanded length. People take the length their role needs, and every length ends with the same recorded questionnaire, so the training record is complete either way.

Who should take this course?

It is written for IT, QA, validation. Owners, managers and sales staff hear the briefing length, regulatory, quality and business-development staff take the standard run, and the people who run the procedure every day get the expanded version. One licence covers one named person on this course for a year, so the same title can be rolled out across a whole team at once.

Is this training accepted as a GMP training record by an auditor?

EU GMP Chapter 2 and 21 CFR 211.25 require documented, effective training, and that is what the course produces: a recorded multiple-choice result per module and per person, a certificate per learner and a transcript of what was taught. No provider can promise acceptance by a specific inspector; the record is built to answer what inspectors ask for, and your quality manager decides the pass mark.

How is it delivered and how fast?

The course arrives as audio files with a companion workbook, a full transcript and the questionnaire per module, within 72 hours of your payment arriving. You host the files on your own system or share them as you like; there is no platform to log into and no learner management system to set up. The free sample module arrives on request, before any purchase.

What does it cost per learner?

The list price is 9 euros per module, per learner, per year, so this course is 72 euros for one person for twelve months. The price per learner falls as more people take the same course: 10 learners pay 10 percent less and the discount keeps growing to 50 percent at 1000 learners. Type a number on the price panel and the yearly total appears.

Get this course for your team

One licence is one person on this course for a year. All three lengths, the workbook, the certificate and the training record are included in every purchase. Pick how many people, add it to the cart, and you have an invoice in your inbox.

Free sample module

Hear one full module before you spend a cent. Send your company and the role you want trained, and a complete sample module comes back with its workbook and transcript.

Connect on WhatsApp

or email [email protected], or ask for a written quote for your team