AI Executive Office — CXO Assistant Platform (On-Premises)  ·  View 02 of 30  ·  1 · Context and scope

High-Level Architecture

The shape of the platform in one picture: six stages from an executive's question to a governed action.

Editable source SVG draw.io All views
Executive experience Mattermost app brief · approvals Executive web React on Kubernetes Mobile MDM managed Edge and identity HAProxy + WAF TLS · rate limit Kong Gateway tenant quota Keycloak token exchange · MFA Orchestration and decision Executive Orchestrator LangGraph runtime Specialist agents 10 capability packs Decision engine situation to option set Investigation runs Temporal workflows Grounding and action Tool plane MCP · caller token OpenSearch hybrid · ACL trimmed Semantic model the only KPI source KServe endpoints forecast · anomaly Execution plane Camunda + Camel Data and knowledge MinIO + Iceberg bronze · silver · gold Trino warehouse canonical model Entity graph Neo4j Kafka CDC · signals Decision store PostgreSQL · append-only Enterprise systems ERP and finance Procurement Projects Document estate approved write CDC outcome check AI Executive Office — High-Level Architecture Application we own Interface / broker Security / platform Data store Queue / topic External / third party synchronous event / async batch The model never reaches a system of record. Every arrow into the right-hand column is a typed tool or an approved action. v 1.0 · owner Data & AI Global Practice · date 2026-09

The load-bearing decision

  • The language model sits in the orchestration stage and nowhere else. It never appears in the data stage, and it holds no credential for any store — everything it learns arrives through the tool plane, carrying the caller's identity
  • Reads fan out widely; writes converge on a single execution plane. That asymmetry is the whole governance model, and it is why the right-hand column has one solid arrow into it and many dotted ones out
  • The tool plane is a set of first-party Model Context Protocol servers. MCP is the wire contract between the model and every read; it is not a new capability, it is the standardisation of the one that was always there

Deliberate omissions

  • Guardrails, evaluation and audit are on every path but drawn only once, in view 29. Repeating them here would triple the arrows and say nothing new
  • The detection pipeline that produces the morning brief runs beside this flow rather than inside it — view 18

Numbers

  • Simple KPI question: p95 under 5 seconds, two model calls
  • Cross-system investigation: asynchronous, 30–180 seconds, progressively disclosed rather than made to wait
  • Morning brief: generated by 06:30 local, delivered 07:10