Validation of AI Systems
- GxP and IT Expertise from a Single Source
- 30+ Years of Experience in Regulated Industries
- SAP Silver Partner
AI Validation According to EU AI Act & ISPE GAMP® AI Guide
Artificial intelligence is increasingly being used in regulated environments – from drug discovery through quality control to production optimization. With the EU AI Act and the new ISPE GAMP® AI Guide, clear regulatory expectations for the use of AI in GxP-relevant systems are emerging for the first time.
For management, QA and IT, this means that AI applications must be risk-classified, data quality and model transparency must be regulatory verifiable, and validation does not end with go-live. At the same time, existing governance structures must be adapted, as AI systems are not classical deterministic IT systems and therefore cannot be managed with traditional validation approaches alone.
DHC supports you in systematically and practically reconciling AI innovation and regulatory security.
Challenges in the Validation of AI Systems
Data Integrity & Bias
Training data, model assumptions and data preparation directly influence product quality and patient safety.
Continuous Control
AI systems deliver probabilistic results and change their performance – without monitoring, regulatory risk arises.
EU AI Act & GxP Conformity
New regulations require traceable models, risk classification and transparent decision logic.
Risk-based AI Validation According to GAMP & EU AI Act
CAQ functionalities directly in SAP S/4HANA
CAQ with SAP maps all central quality processes directly in the SAP system – without additional specialized CAQ software and without redundant data management. Based on QM with SAP, inspection processes are seamlessly linked with purchasing, production, logistics and maintenance.
Typical CAQ processes in SAP:
- Inspection planning & inspection lots
- Incoming goods, in-process & final inspections
- Measurement data acquisition & Statistical Process Control (SPC)
- Test equipment management
- Deviation & complaint management
- Quality KPIs & dashboards
CAQ with SAP is thus the operational foundation of an integrated eQMS with SAP – specifically for manufacturing companies.
Result for management:
- Real-time transparency on quality KPIs
- Reduced quality costs
- Higher process & audit security (e.g. ISO, IATF)
- Future-proof architecture based on SAP S/4HANA
DHC ePaper: ISPE GAMP® AI Guide
DHC Consulting Services
- Initial Maturity Check:
Ranking of maturity level with our proprietary AI Scorecard - Strategic Alignment:
Clarification of goals and application areas for AI in the corporate strategy. - Data Strategy:
Analysis of workflows, data flows, data sources and data management.
- Governance and Data Sovereignty:
Clarification of authorization concepts and copyrights. - Infrastructure:
Consulting for digital infrastructure requirements. - Corporate Culture:
Assessment of corporate culture regarding openness to innovation and competence in data capture and qualification.
- Create clarity:
Which AI systems, use cases and business processes should be validated? - Understand regulatory requirements:
Which guidelines (GxP, EU AI Act, FDA, ISO standards etc.) are relevant? - Conduct risk assessment:
Which critical areas need special attention?
- Define roles and responsibilities:
Who is involved in the validation process, who is responsible for what? - Foundation for a roadmap:
Define action steps, resource requirements and timeline.
- Risk-based Validation:
We analyze and assess the risks of your AI system and develop validation strategies based on them. - Data Validation:
We verify the quality, completeness, fairness and trustworthiness of your data.
- Documentation and Traceability:
From the definition of user requirements to the creation of the validation report, we ensure all regulatory requirements are met. - Continuous Monitoring & Framework:
We help you implement processes and SOPs for the entire lifecycle of your AI systems.
Why DHC – Your Specialist for AI Validation
DHC combines expertise and deep GxP know-how with a deep understanding of regulated industries. For more than 30 years we have been supporting companies in the validation of computerized systems.
Our approach unites compliance and innovation: We help you use AI technologies safely and in a regulatory-compliant manner, without limiting the opportunities of new technologies and innovation potential.
Validation of AI Systems
Trust through Experience
We are the specialist for all QM, QA and QC solutions based on SAP ERP.
“DHC developed a tailored validation concept together with us, which was then successfully implemented and audited. The collaboration so far has been outstanding, and we would gladly rely on the expertise of DHC’s experts for future projects.”
“The collaboration with DHC was always goal-oriented and very pleasant. Thanks to the high expertise in GMP and CSV as well as the development of a suitable risk management and framework, Hamilton Bonaduz AG was able to become the first Swiss life science company to achieve successful validation completion for Microsoft Dynamics D365FO in Switzerland.”
Frequently Asked Questions about AI Validation (FAQs)
When does AI require validation in the life sciences environment?
Whenever AI systems influence product quality, patient safety, data integrity or regulatory decisions.
How does AI validation differ from classic CSV?
AI systems deliver probabilistic results and can change through model drift. Therefore, training data, bias risks and monitoring mechanisms must also be assessed.
What role does the EU AI Act play?
The EU AI Act defines risk classes for AI systems and requires additional governance, documentation and monitoring requirements for high-risk AI.
How is model drift managed from a regulatory perspective?
Through structured monitoring concepts, defined performance limits and clearly regulated re-validation processes.
Who bears the regulatory responsibility for AI systems?
Responsibility remains with the regulated company – regardless of whether AI is developed internally or used as a cloud service.
Is retrospective validation of existing AI systems possible?
Yes. Existing systems can be assessed, documented and transferred into a structured governance and monitoring model.
Validation of AI Systems