AI Usage Disclosure

EMEGA may use configured AI services to detect language and intent, retrieve approved Knowledge Base content, draft or translate replies, summarize conversations, identify missing information, suggest routing, classify records, extract contact details, and propose CRM updates.

Customer-facing answers should be based on approved organization knowledge, structured Branch Knowledge Profiles, conversation context, and deterministic policy checks. The system should ask a clarification question or hand off when the required fact is missing, ambiguous, sensitive, or insufficiently supported.

Internal instructions, retrieved excerpts, confidence calculations, and system prompts must not be exposed as customer answers.

Organizations can require human approval by risk, channel, workflow, intent, confidence, or action. Authorized staff remain responsible for admissions, employment, discipline, fees, discounts, complaints, safeguarding, financial matters, official records, and other significant decisions.

An AI suggestion does not itself create a promise, admission, contract, official record, or staff decision.

AI can misunderstand context, dialect, names, branches, grades, or provider events and may produce incomplete or incorrect text. Users must review high-impact output and report errors. EMEGA uses controls to reduce unsupported answers but cannot guarantee that every output is correct.

Prompts and outputs may contain conversation and CRM context needed for the configured task. Only the minimum necessary context should be sent to an approved AI provider. Provider, region, retention, and training settings must be documented by the operator and customer agreement.

Users must not place unnecessary sensitive data, passwords, payment details, confidential examinations, or unrelated student records in prompts.

AI may propose a name, phone, branch, grade, intent, assignment, task, or note. Changes must pass field validation, identity matching, permissions, provenance logging, and configured approval. Destructive actions and protected-record changes must not be performed solely from a free-form model response.

AI modes can be limited or disabled by the organization. A customer may ask to speak with a person, and the conversation should be routed according to the configured handoff path. Questions about an AI-assisted decision may be submitted through the Contact and Support page.