AI messaging is licensed with a usage allotment. AI Usage Management shows how much of that allotment has been consumed, so administrators can plan ahead rather than be surprised, and sends email alerts to chosen recipients when usage reaches a set percentage.

---

Accessing AI Usage Management

AI Usage Management is accessible to accounts with Admin permissions.

1. Navigate to AI Configuration > AI Usage Management
2. Current usage and the alert settings are shown in the modal

---

Viewing Usage

The modal shows the number of AI messages used in the current period against the account's message allotment.

---

Usage Alerts

1. Enable the email notification and set the usage percentage that triggers it — for example, 90
2. Enter the email addresses that should receive the alert
3. Save the changes

When usage reaches the set percentage, each recipient is emailed.

---

AI Governance and Data Privacy

Xima operates under a formal AI Tools Usage Policy designed to ensure that AI integration within CCaaS is responsible, ethical, and secure. This policy governs how AI processes customer data, how that data is isolated, and how AI behavior is constrained to authorized use cases.

Data Privacy and No-Training Guarantees

Xima's AI infrastructure is selected specifically to comply with data privacy regulations including GDPR and to ensure that customer data is never used to train external AI models.

The Control Layer — LiteLLM

LiteLLM serves as Xima's centralized AI gateway, sitting between CCaaS and the underlying AI models. It manages API keys centrally, enforces rate limits to prevent abuse, and ensures every AI request follows internal security protocols before reaching a model.

Audit and Transparency — Langfuse

Langfuse is used as Xima's AI observability and audit platform. Every AI interaction is traced, producing a complete record of the prompt, the model's reasoning path, and the final output, which supports ongoing accuracy monitoring and bias detection.

Operational Guardrails

Beyond infrastructure, AI behavior in CCaaS is constrained at the prompt level so AI agents remain focused tools rather than general-purpose assistants:

• Tool-bound logic — AI agents are limited to an authorized toolkit, such as knowledge base searches and skill transfers
• Scope enforcement — AI agents are instructed to decline tasks or questions outside their configured capabilities
• Hallucination mitigation — AI agents are prohibited from using knowledge outside the provided materials; responses must be grounded in the assigned knowledge base content

Compliance Summary