Responsible AI
Principles for applying AI and machine learning responsibly in real business systems.
Principles
Portalwiz approaches AI and machine learning as business systems—not magic. We aim to use models where they can improve decisions, experiences or operations while keeping appropriate human judgment, data governance and evaluation in the loop.
Authorised data use
Client or user data should only be used for agreed purposes. Training, fine-tuning, evaluation or prompt-processing choices should match the contract, data sensitivity, provider configuration and applicable law.
Evaluation before scale
ML and AI systems should be evaluated against relevant business and technical criteria before production use. Depending on the use case, this may include accuracy, robustness, bias/error analysis, latency, cost, drift, explainability and human review.
Human oversight
High-impact decisions should not be delegated to automation without an appropriate governance model. Portalwiz can design review steps, escalation paths, confidence thresholds and auditability into AI-enabled workflows.
Security and privacy
AI delivery should follow data minimisation, access control, environment separation and secure integration practices. Sensitive data should not be sent to third-party model providers without an authorised basis and appropriate safeguards.
Continuous monitoring
Models and AI workflows can change in performance as data, user behaviour or providers change. Production systems may therefore require monitoring, evaluation, versioning and controlled updates.