AI Usage in Eleveo Software

Overview

Eleveo integrates artificial intelligence (AI) capabilities into its software solutions to enhance Quality Management and analysis of voice-based interactions between agents and customers. These AI functionalities operate within a secure and private cloud infrastructure and are not customer-facing. No customer data is used for training the AI models. Customer data remains the sole property of the customer, and Eleveo does not claim any ownership, rights, or usage privileges over such data beyond what is necessary to provide the contracted services.

AI Capabilities

Eleveo AI offers several functions:

• Summaries – Generates a summary based on the transcription created by Eleveo Speech Recognition.

• AI-Rating – Assigns a score (1–5) with a brief explanation for agent-customer interaction quality.

• Highlights and Next Steps – Identifies key points and next steps from conversations.

• Flags – Indicates sentiment, churn risk, escalation, and repeat contact.

• Topics – Detects predefined topics defined by users.

• Automated Quality Management (AQM) – Enables 100% of conversations to be reviewed using Speech and Generative AI.

AI Model and Data Usage

Eleveo uses large language models (LLMs) such as Mistral 7B and Gemma 2-9B, both of which are state-of-the-art generative AI models based on transformer architecture. These models are pre-trained and do not receive any customer data for further training or tuning. All AI processing occurs within Eleveo’s private cloud infrastructure, ensuring that customer data is encrypted and never shared externally.

Mistral 7B is developed by Mistral AI, while Gemma 2-9B is developed by Google DeepMind. Both models are selected for their performance, reliability, and ability to integrate seamlessly into Eleveo’s Quality Management tools.

Bias Mitigation and Ethical Standards

Eleveo selects AI models from providers that demonstrate strong commitments to fairness and bias mitigation, such as Mistral AI and Google DeepMind. While these models are developed externally, Eleveo performs internal evaluations to ensure fairness, reliability, and suitability for its use cases. Eleveo complies with relevant ethical and regulatory standards, including the EU AI Act and GDPR, and implements safeguards such as human oversight and regular audits. Mistral AI and Google DeepMind, follow best practices to mitigate bias. These include using diverse datasets, applying fairness constraints, performing continuous audits and, in addition to that, Google DeepMind includes safety training with reinforcement learning (RLHF) and red teaming. Mistral AI and Google DeepMind comply with ethical guidelines and are committed to transparency, fairness, and accountability.

Explainability and Human Oversight

The system includes explainable AI mechanisms such as attention maps and rule-based logic to ensure decision transparency. Supervisors and managers review and validate AI outputs, and Eleveo provides training and support to ensure responsible AI use.

Security and Privacy

Customer data, including PII/PHI, remains encrypted and is never used to train AI models. Generative AI is used exclusively within the Eleveo software environment to support customer-facing use cases, such as call summarization and quality scoring, on a per-interaction basis. It is not used for Eleveo’s internal analytics, marketing, or training purposes.

Performance and Risk Management

AI capabilities are actively monitored. Human oversight, structured feedback mechanisms, and regular audits ensure model performance and mitigate risks related to decision automation. The tools enhance existing solutions and are designed for long-term integration rather than short-term obsolescence.

Compliance with the EU AI Act

Eleveo assesses its AI functionalities under the risk-based framework of the EU AI Act. The AI tools provided (such as summaries, sentiment analysis, and quality scoring) are classified as limited-risk systems under the EU AI Act. Where applicable, Eleveo ensures compliance with mandatory transparency obligations, including informing users when AI-generated content or scoring is used.

Eleveo does not deploy high-risk or prohibited AI systems. All AI functionalities are designed to support — not replace — human decision-making, and human oversight is maintained throughout.

For additional details regarding Eleveo’s AI usage, data protection practices, or regulatory compliance, customers may contact legal@eleveo.com.

This statement is for informational purposes only and does not constitute a contractual obligation or legal warranty. Please refer to Eleveo Terms of Service and data processing documentation for binding commitments.

Eleveo a.s.

AI Usage in Eleveo Software-v10.0.pdf