Skip to main content
Blueprint

The KUOS platform, by layer

A working view of how outcomes, specialised agents, models, runtime and data fit within deployment and governance boundaries chosen for the organisation.

Top — Outcomes & productsMiddle — Orchestration + agentsBottom — Foundation
01

Outcomes & products

What you ship to your customer

Design

Owned by

PRISM · PIXEL · ECHO

Plan

Owned by

ATLAS · Taylor · MAESTRO

Run

Owned by

FORGE · BRIDGE · ORION

Govern

Owned by

SENTINEL · VAULT

Care

Owned by

AVA · SPARK

ATLAS · Each outcome is owned by a small agent team. The team that runs Design is different from the team that runs Govern — accountability is split by capability, not by org chart.
02

Orchestration + specialised digital capabilities

How AI supports the work

Conductor
MAESTRO
MAESTRO plans + dispatches + synthesises. Every other agent reports up here.
ADAExecutive Secretary & Dispatch Agent
CUSTOSData Protection Officer (DPO)
ATLASStrategy Agent
NOVAIntelligence & Industry Research Agent
SAGEKnowledge Architect
AVASolution Architect
FORGEEngineering Agent
ORIONInfrastructure Agent
SENTINELGovernance Agent
PIXELExperience Agent
ECHOMedia & Publishing Officer
SPARKInnovation Agent
AURACommercial & Proposals Agent
PRISMReadiness & Assessment Agent
BRIDGEChange Management & Adoption Agent
VAULTQA & Validation Agent
TAYLOR · MAESTRO reads a request → picks a crew → runs the steps → pauses at the configured human approval gate. Each major step is recorded for review.
03

The foundation

Where the work happens

03.1

Model & Deployment Advisory

FORGEBRIDGE
LLMProvider-neutral model selection
ORCloud, EU-hosted & local deployment paths
MCPMCP & Tools · Native Protocol Servers
CFCloudflared Secure Tunneling
03.2

Running / Routing

ORION
HWExisting AI PC / NPU / GPU assessment
LANOn-premise, sovereign EU or cloud deployment
SLASLOs · on-call handoff · runbooks
03.3

Chat Matrix

SENTINELCUSTOS
MTTenant and role isolation boundaries
DAControlled data-access boundaries
RVConfigurable review and approval gates
03.4

Data

SAGENOVA
documentsdatabasespoliciesknowledgeapproved APIsresearchcustomer dataoperational datalogscontentworkflowscataloguesfinancemarketregulatory
Structured + unstructured. SAGE turns it into knowledge; NOVA turns it into research.
HUMAN CONTROL · The foundation remains under customer control. Deployment and data boundaries are selected for the organisation's requirements; agents operate within those boundaries.

End to end

request → answer
  1. 1. Customer
    Submits a question at /[locale]/ai-workflow
  2. 2. Router (ADA)
    Classifies intent + urgency
  3. 3. Orchestrator (MAESTRO)
    Picks the crew using policy weights
  4. 4. Workers
    ATLAS / SENTINEL / CUSTOS / VAULT draft, review, score
  5. 5. Memory gate
    Long-term vs skip vs drop, based on utility + surprise
  6. 6. Human approval
    Reviews, approves, returns or rejects
  7. 7. Mailbox
    Email the customer, log to erp_activities
  8. 8. Memory
    Feed the next run

Every step writes to agent_messages and erp_activities. The Learning dashboard (RPE + policy store + memory gate) reads from the same stream. The audit log is not a separate product — it is the workforce.

A working reference implementation of the KUOS Platform

This site demonstrates the platform through a multilingual public experience, CMS, learning environment and governed agent workflows. It shows how knowledge, agents, applications and human approvals can operate on one reusable foundation.

Start with a conversationBuild capability togetherKeep people in commandTest before productionRun it in operationsFit real workplace work

Start with a conversation

01 / 06

We value your privacy

We use cookies and anonymous analytics to understand how visitors use KUOS and improve the experience. Only non-essential cookies are used for usage analytics. You can change or withdraw consent at any time via Cookie settings. Cookie policy · Privacy policy

KUOS Platform — Enterprise AI Enablement