AI literacy

Overview

Auroranexis mixes three classes of assistance: generative AI (large language models that draft text), deterministic logic (rules, SQL aggregates, fixed thresholds), and predictive scoring (models or heuristics that rank risk, health, or retention signals). Only generative surfaces require EU AI Act Article 50-style transparency disclosures in the product UI.

Optional AI features send prompts and trusted operational context to configured providers (for example OpenAI) when enabled. Deterministic dashboards and many executive intelligence findings do not call generative models even if they appear under an “intelligence” heading.

Core Concepts

• Generative AI — drafts or transforms text (Copilot, Report Assistant, executive summary drafts, risk/incident AI analysis, knowledge generation, automation NL translation).
• Deterministic — fixed rules and calculations (many metrics, filters, workflow engines without LLM calls).
• Predictive — scores or rankings that estimate likelihoods; may be heuristic or model-based but are not free-form text generation.
• Hallucination risk — generative models can invent facts not present in verified context; always check citations and source records.
• Human review — required before publishing customer reports, enabling automations, or acting on risk/incident recommendations.
• Personal data — do not paste unnecessary personal data into prompts; treat AI logs as potentially sensitive.
• Unsafe reliance — do not use AI outputs as sole basis for legal, medical, employment, or high-stakes contractual decisions.
• Security — treat model providers as sub-processors when enabled; escalate suspected data leakage to security/legal contacts.