artiplane
Artiplane · enterprise AI platform

An enterprise AI layer IT can govern.

Five microservice layers, deployment on Azure or customer infrastructure, no copies of source data and no writes without authorisation.

Enter the IT demoFive-layer architecture
Governance & Risk: monitored subjects, policy blocks, guardrails, events and EU AI Act conformity.
Agents & Workflows: enterprise AI registry with models, providers, environments, usage and cost.
  • 1.323monitored subjects
  • 230registered AI elements
  • 5layers, one model

For IT and Chief AI Officers

Eight answers before the security review.

Architecture

Five microservice layers with explicit contracts: Knowledge, Semantic, Organization, Orchestration, Experience. Versioned APIs and events between layers; only the first touches systems.

Deployment

Managed Azure in an EU region, customer cloud (Azure, AWS, Google Cloud) or on-premise Kubernetes. Network, keys and identity remain under customer control.

Integration

Connectors via API, OData, CDC, files and events; first access read-only with least-privilege credentials. Versioned mapping to the canonical model.

Data

No parallel data lake: systems remain the source of truth. Every value keeps source lineage, freshness and permissions; retention is configurable.

Identity and access

Enterprise SSO, roles, object and action permissions, market and brand scopes. The server revalidates every sensitive request.

LLM gateway

Azure OpenAI, OpenAI, Anthropic Claude, AWS Bedrock and open-weight models chosen per task. Customer LLM contracts integrated, with budgets and costs by area.

Governed writes

Agents propose, they do not execute: every write is a registered action with contract, approval threshold, idempotency and execution record.

Lifecycle and audit

Draft → Test → Production with Change Requests, versions and rollback. SHA-256 hashed logs, configuration audit and EU AI Act controls.

Control Plane

Architecture, connectors and governance.

Admin overview: applications, AI, users, processes, Orchestrators and Super Apps with open alerts.
Control Plane · architecture mapOpen this screen
Governance & Risk: monitored subjects, policy blocks, guardrails, events and EU AI Act conformity.
Control Plane · governance and EU AI ActOpen this screen

Deployment

Runs inside the customer perimeter

Azure reference, deployable on AWS, Google Cloud, private cloud or on-premise. Data stays where it is; models are chosen by contract and policy.

Deployment targets
  • Microsoft Azure
  • AWS
  • Google Cloud
  • Private cloud
  • On-premise

Customer perimeter · tenant, VNet/VPC, identity provider

Experience
  • Super Apps
  • Admin console
  • Chat
  • API
Artiplane runtime · containerised microservices
  1. 1Knowledge
  2. 2Semantics
  3. 3Organisation
  4. 4Orchestration
  5. 5Experience
Customer systems
  • SAP S/4HANA
  • Salesforce
  • Microsoft 365
  • PLM
  • WMS
  • POS
  • Data lake
LLM gateway
  • Anthropic Claude
  • OpenAI
  • Azure OpenAI
  • AWS Bedrock
  • Google Gemini
  • Private models
  • n8n

Product reference architecture. Topology, sizing and controls are confirmed with the customer’s IT during technical review.

Operational control

What IT sees every day

Console, AI registry, governance, audit and identity.

Admin overview: applications, AI, users, processes, Orchestrators and Super Apps with open alerts.
  • Alerts per system, user and agent
  • Every number opens its detail
  • Same model for IT and business
6.060
processes
304
users
43
Orchestrators

Real product screens · Atelier Nova Group reference tenant

Controls

Security control matrix

Product specification: each control is validated on the customer tenant.

ControlLayerHow it applies
Single sign-on and MFAExperienceSAML/OIDC to the customer identity provider; MFA by role.
Role and attribute accessOrganisationRoles, areas, countries and restrictions from the org registry.
Least-privilege accountsKnowledgeBridge blocks campaigns with excessive permissions.
Secrets and encryptionKnowledgeCredentials encrypted in the customer vault; TLS in transit.
LLM policies and guardrailsOrchestrationAllowed model, data and tools per task; blocks recorded.
Human approvalExperienceSensitive actions confirmed or four-eyes before any write.
Signed auditAllEvent, subject, policy and outcome exportable for review.
Data residencySemanticsNo bulk copy: the semantic model points to source systems.

Assess Artiplane on your landscape.

Systems, identity, AI models and deployment constraints: we define scope and reference architecture together.

Enter the IT demoTechnical resources