JNN Global Ltd (trading as “VERYX”, part of Groupe Nseya Digital) · Updated 12 June 2026
All AI activity flows through one gateway: model routing, approved prompt templates, pre-execution cost calculation (the ACU cost is always shown before you run an agent), usage logging, fallbacks, rate limits, safety filtering and output validation.
Anthropic Claude is the primary reasoning provider; OpenAI, Gemini and Vertex AI connect through the same abstraction when configured. Without live keys the platform runs a deterministic simulation so behaviour is always predictable and testable.
AI context (your organisation profile, recalled episodes, documents you submit) is scoped to your workspace and never crosses tenants. Your data is not used to train foundation models. AI memory can be reviewed by your admins and is deleted with your workspace.
Agents recommend freely but never execute high-impact actions without explicit approval: external communications, deletions, financial changes, permission changes, workflow submissions, external risk escalations, admin settings and paid premium tasks all wait for a human decision in the approval queue.
Every AI action is metered in ACUs and billed at the fixed multiplier of raw provider cost — uniform for every customer, with published volume bands. Budget caps and alerts keep spend under your control.