MADAI // ENTERPRISE INTELLIGENCE SYSTEM
One governed AI system. Six entry points into business transformation.
Select a point in the map. Each one opens a concrete system layer — what it solves, how it is governed and where to continue for the full detail. No invented KPIs and no dependency on a single model.
● SELECT A POINT / ARROW KEYS MOVE / ESC CLOSES DETAIL
SYSTEM MAP / V40 INTERACTIVE
EVIDENCE FIRST GOVERNED AI DETAIL →
01 / FOUNDATION KNOWLEDGE FACTORY data → governed knowledge
02 / CONTROL PRIVATE AI on-prem / private cloud
03 / WORKFLOW AI AGENTS draft → approve → act
04 / PROOF EVALUATION Gold set / regression
05 / OVERSIGHT GOVERNANCE ACL / audit / ownership
06 / ENTRY COMPASS & PILOT the smallest proof step
01 ON-PREM / PRIVATE CLOUD / HYBRID02 ACL + CITATIONS + PROVENANCE03 GOLD EVAL SET + REGRESSION04 HUMAN-IN-THE-LOOP
SELECTED SYSTEM LAYER ESC
01 / KNOWLEDGE FACTORY A model is not knowledge. Fix the input first. We turn company sources into a versioned, citable knowledge layer with access control while preserving a trace to the original.
hash, provenance and source ownership classification, metadata, entities and relationships hybrid retrieval, reranking and citation verification Open Knowledge Factory detail →
02 / PRIVATE AI Data should determine architecture. Not model marketing. The same knowledge and evaluation layer can use local and selected cloud models. Placement follows sensitivity, latency, performance and economics.
on-premise for maximum control and offline work isolated private cloud environment hybrid policy routing by task and data class Open Private AI detail →
03 / AGENTIC WORKFLOW The agent prepares. An accountable person decides. Agentic workflows connect knowledge, tools and business rules. Sensitive actions remain approval-gated with rationale and an audit trail.
role- and context-aware tool permissions a proposed next step with supporting evidence approval before transactions or data changes Open AI agents detail →
04 / EVALUATION Quality should be measured, not claimed. For the pilot we create a Gold evaluation set and measure groundedness, citation correctness, completeness and abstention.
before / after measurement on a real workflow regression after model, prompt or retrieval changes explicit GO / NO-GO criteria for production Open evaluation detail →
05 / GOVERNANCE Governance is part of architecture, not paperwork at the end. Identity, permissions, monitoring, human oversight and incident procedures are designed with the workflow. Every sensitive action has an owner.
purpose, owner and allowed context ACL before retrieval and tool use audit trail, monitoring and incident response Open governance detail →
06 / COMPASS & PILOT Start with the smallest step that can produce evidence. We first test the process, data, risks and economics. A larger implementation follows only after technical and commercial proof.
AI Compass from CZK 49,000 excl. VAT AI Pilot from CZK 229,000 excl. VAT Private AI from CZK 590,000 excl. VAT Open full pricing →