ONEiO and AI (ONEAi®)

How ONEiO Uses Artificial Intelligence Safely, Responsibly and Transparently

At ONEiO, our goal is to make integration work easier, smarter, and more efficient—without compromising the security, privacy, or reliability of your data. ONEAi brings selected AI-powered capabilities into the ONEiO Integration Automation Platform to help users work faster and resolve issues more effectively, while maintaining full human control and ensuring compliance with regulations such as the EU AI Act.

This page explains exactly what AI features we offer, how they work, what data they use, and how ONEiO protects your information.

Overview of ONEAi

ONEAi is an optional set of assistive, human-in-the-loop features embedded in the ONEiO SaaS platform. These capabilities are designed to enhance—not automate—your experience when building, maintaining, and troubleshooting integrations.

ONEAi includes:

Text Translation (AWS Translate)

Automatically translate text fragments inside mapping expressions.

Useful when working with multi-language integration content.

Integration Field Mapping Suggestions (Amazon Bedrock GenAI)

AI-assisted suggestions for mapping integration fields based on curated metadata.

Integration Failure Troubleshooting (Amazon Bedrock GenAI)

AI-assisted analysis of API error messages and configuration context, producing human-readable explanations and suggested next steps.

All AI features require explicit user action—NONE operate automatically.

What Data is Used for AI Features?

ONEiO follows strict data-minimisation principles. AI features only receive the minimal required information, and no integration message content is used for training or fine-tuning any model.

Text Translation

• Text provided by the user inside a mapping expression is sent to AWS Translate.
• No contextual integration message data is included.

Mapping Suggestions

• Uses a curated internal dataset of existing mapping metadata.
• Does not use customer integration message content.
• No customer-provided messages or payloads are ever sent.

Integration Failure Troubleshooting

Triggered manually by the user:

• The API error returned by the integrated system
• Relevant endpoint configuration metadata
  → are provided to the GenAI model to produce clear, actionable explanations.

Training and Data Use

• ONEiO does not train its own AI models.
• AWS Bedrock and AWS Translate do not use ONEiO customer inputs for model training.

• ONEiO has globally opted out of any customer-data-driven model improvement via AWS Organizations.

   

How ONEiO Protects Your Data

ONEiO applies the same strict security controls to AI-related data as to all customer data in the ONEiO SaaS.

Encryption

• All AI data exchanges use TLS 1.2+ encryption in transit.
• All data stored in ONEiO systems is encrypted at rest using AES-256.

No Model Retention

AWS Bedrock and AWS Translate do not store or reuse your data for training or service improvement.

Access Controls & Authentication

• All actions that trigger AI functions follow ONEiO’s strict access controls, MFA, and least-privilege principles.

Audit Logging

All AI-triggering events follow ONEiO’s comprehensive audit log policy (logs retained 5 years).

No Sharing With Third Parties

Customer data is not shared with external parties beyond AWS, as the AI processor, governed under strict contractual and technical controls.

ONEiO’s Compliance with the EU AI Act

ONEiO is classified as a Deployer under the EU AI Act.

All ONEAi features fall under Minimal or Limited Risk as defined by the Act.

ONEiO complies through:

• Human oversight: users must review and approve all AI suggestions.
• Transparency: clear explanations of what the AI does and the data it uses.
• Safety and monitoring: continuous model performance checks and risk assessments.
• Opt-in usage: AI is only invoked when a user chooses to use it.
• Data protection: strict isolation of customer data and zero training use.

No AI-based decisions are made that would have significant legal or personal impact.

Human-in-the-Loop and Explainability

ONEAi is explicitly built on human control.

• AI suggestions are never applied automatically.
• Users can review, edit, accept, or ignore suggestions.
• AI outputs (especially troubleshooting explanations) are written in clear language.

• ONEiO continuously evaluates output quality as part of our performance monitoring and ISMS risk management processes.

   

Performance Monitoring & Risk Management

ONEiO maintains a continuous evaluation process:

Performance Monitoring

• Samples of AI outputs are regularly reviewed by ONEiO engineers.
• Findings inform improvements in prompting and operational parameters.

Risk Management

• AI risks are handled within ONEiO’s ISO 27001:2022-certified ISMS.

• Risks are assessed and treated using the same methodology applied to all information assets.

   

Customer Choice: Opt-In and Opt-Out

You always control whether AI is used:

• Text translation: only used if configured in your mapping.
• Mapping suggestions: appear only when manually requested.
• Troubleshooting: only triggered by a user action.

You can use ONEiO fully without enabling any AI features.

Intellectual Property Protection

ONEiO protects customer intellectual property by ensuring:

• No customer content is used for the training or fine-tuning of any model.
• AWS processing is covered by strict contractual data protection terms.
• Prompts and results are processed securely and not stored by the AI model.

Summary

ONEAi enhances the ONEiO experience while preserving the principles our customers rely on:

Secure - Built on ONEiO’s strong encryption, logging, and access-control framework.

Compliant - Aligned with ISO 27001:2022, GDPR, and the EU AI Act.

Transparent - Clear explanations and human-in-the-loop oversight.

Optional - AI features activate only when explicitly chosen by the user.

Never trained on customer data - Your integration data is never used to train or improve models.

ONEAi is designed to support you—not replace you—by making integration work faster, clearer, and easier.