ARBI.
Architecture & roadmap

How ARBI works

An honest map of what exists today, what is illustrative, and what is planned. Transparency matters more to us than appearing finished.

Legend:Live nowalready built and usableIllustrativereal method, sample dataPlanneddesigned, not yet built
Current state

ARBI today

ARBI currently exists as an Experience Edition. This release is focused on getting the executive experience right — validating the decision workflows, the boardroom workspaces, the trust framework, scenario evaluation, and the overall information architecture — before any enterprise data is connected.

It demonstrates exactly how ARBI is expected to operate. To make that experience real and testable, the datasets, relationships, recommendations and scenarios are illustrative — built on a synthetic reference enterprise so leaders can evaluate the decision experience before full backend integration. What you judge today is the experience and the method, not the numbers.

Live nowExecutive experienceLive nowDecision workflowsLive nowBoardroom workspacesLive nowTrust frameworkLive nowScenario evaluationLive nowInformation architecture
Business view

From data to outcomes

In plain terms, ARBI moves along one path — and every step stays connected to the one before it.

Step 1
Business data
Step 2
Evidence
Step 3
Insights
Step 4
Decision options
Step 5
Business outcomes

Start with what's happening in the business, gather the evidence behind it, turn that into insight, weigh the decision options, and steer toward a better outcome — with the reasoning visible the whole way.

Trust view

Why you can trust what you see

Trust is built from four things. Here is the honest status of each.

ElementCurrent statusPlanned status
Source transparencySources and a source trail are shown for each topic — using illustrative data.Live lineage back to the originating enterprise systems.
Evidence transparency20+ benchmarked indicators per hero topic, each visible and traceable (illustrative values).The same indicators, measured from real data.
Confidence scoringA real, deterministic reliability score (evidence quality, source coverage, framework support, freshness) — on illustrative inputs.Identical method, computed on live evidence.
ExplainabilityLive today: the full chain and key assumptions are always visible.Extended with AI-assisted plain-language explanation.
Technical architecture

The seven layers

Each layer rests on the one below it; data flows upward from the sources to the executive experience. Today, only the top layer is live.

L7Executive Experience
Live now
Purpose

Boardroom-ready decision workspaces — the part leaders actually use.

Technologies

Decision workspaces, lenses, trust panels

Planned evolution

Connected to live enterprise data

L6AI Assistance
Planned
Purpose

Help leaders explore, interpret and explain — never to decide for them.

Technologies

LangGraph · RAG · LLM services

Planned evolution

AI-assisted navigation and explanation, human-in-control

L5Decision Intelligence
Illustrative
Purpose

Turn evidence into recommendations, scenarios and reliability scores.

Technologies

Python · business rules · simulation models

Planned evolution

Operational, data-driven decision services

L4Relationship & Context
Planned
Purpose

Connect workforce, capability, operational and business concepts.

Technologies

Neo4j · knowledge-graph concepts

Planned evolution

Contextual relationship intelligence

L3Transformation
Planned
Purpose

Shape raw data into clean, business-ready datasets.

Technologies

dbt

Planned evolution

Automated transformation + semantic business models

L2Data Platform
Planned
Purpose

Centralized, analytical foundation that scales.

Technologies

BigQuery

Planned evolution

Enterprise-scale analytical data platform

L1Data Sources
Planned
Purpose

The enterprise systems where workforce and business data lives.

Technologies

HRIS · ATS · Learning · Workforce Planning · ERP · Operations · Labor-market

Planned evolution

Connected enterprise datasets

Why BigQuery & dbt (Layers 2–3)

BigQuery gives an enterprise-scale, analytical foundation that handles large workforce and operational datasets without heavy infrastructure. dbt turns that raw data into clean, well-defined, business-ready models — so every number has a consistent, documented definition feeding the layers above.

Why a relationship layer (Layer 4)

Workforce risk is rarely one number — a skills gap connects to attrition, to capacity, to a business outcome. A relationship layer (Neo4j) captures those connections so ARBI can reason about context, not just isolated metrics.

On AI (Layer 6)

ARBI is not intended to replace decision-makers. When AI is added, it will support exploration, interpretation, explanation and scenario understanding — helping leaders grasp the evidence faster. Human leaders remain responsible for the decisions. The reasoning will stay visible and inspectable, never a black box.

Current vs future

What's real today, what's coming

CapabilityCurrent Experience EditionFuture production platform
Data sourcesIllustrative reference dataset (a synthetic enterprise)Connected live enterprise systems
RecommendationsCurated and rule-based, designed to validate the experienceGenerated from live data and analytics models
ScenariosDeterministic, pre-modelled trade-offsData-driven scenarios on real baselines
BenchmarksIllustrative PY / target / industry / best-in-classSourced from real internal history and market data
Confidence (reliability) scoresReal deterministic method, illustrative inputsSame method applied to live evidence
AI assistanceNot yet — interactions are structured, not generativeAI-assisted exploration and explanation
Decision workspacesImplemented and usable todayUnchanged experience, powered by live data
Trust frameworkLive: source trail, evidence, assumptions, reliabilityExtended with real system lineage
Implementation roadmap

From experience to enterprise

Six phases. We are at Phase 1.

Phase 1 · Experience Edition· Decision workspaces
Live now
Objective

Validate the executive experience, decision flow and trust model.

Deliverables

Boardroom workspaces, lenses, scenarios, trust framework.

Expected benefit

Proves the experience and aligns stakeholders before any data cost.

Phase 2 · Data Foundation· BigQuery · dbt
Planned
Objective

Stand up the analytical platform and business-ready datasets.

Deliverables

Central data platform; transformed, modelled datasets.

Expected benefit

A scalable, trusted single source for everything above it.

Phase 3 · Relationship & Context· Neo4j
Planned
Objective

Connect workforce, capability, operational and business concepts.

Deliverables

Relationship layer linking the objects ARBI reasons over.

Expected benefit

Context-aware insight instead of isolated metrics.

Phase 4 · Decision Intelligence· Python · analytics · simulation
Planned
Objective

Make recommendations, scenarios and reliability data-driven.

Deliverables

Operational decision services replacing curated content.

Expected benefit

Recommendations grounded in the enterprise's own data.

Phase 5 · AI Assistance· LangGraph · RAG · LLM
Planned
Objective

Assist exploration, interpretation and explanation.

Deliverables

AI-assisted navigation and plain-language explanation.

Expected benefit

Faster understanding — with humans still deciding.

Phase 6 · Enterprise Deployment· Live data · workflows
Planned
Objective

Run on live enterprise data inside real decision workflows.

Deliverables

Connected, governed, production deployment.

Expected benefit

ARBI becomes a durable layer of how the enterprise decides.

Technical FAQ

Straight answers

Is ARBI already connected to enterprise data?

No. The current Experience Edition runs on an illustrative reference dataset — a synthetic enterprise designed to validate the decision experience. Connecting live data is Phase 6.

Is BigQuery already operational?

No. BigQuery is the planned data platform (Phase 2). It is not yet part of the running product.

Is dbt already operational?

No. dbt is the planned transformation layer (Phase 2). It is not yet running.

Is Neo4j already operational?

No. The relationship layer is a conceptual design today (Phase 3).

Is AI already operational?

No. Today's interactions are structured, not generative. AI assistance (Phase 5) is planned, and even then it will assist exploration and explanation — it will not make decisions.

Are recommendations generated from real company data?

No. Recommendations are currently curated and rule-based to validate the experience. They become data-driven in Phase 4.

What is simulated today?

The datasets, evidence values, benchmarks, recommendations and scenarios are illustrative. The reliability scoring method, the decision workflow, the lenses and the trust framework are real.

What is planned for future phases?

A live data platform (BigQuery/dbt), a relationship layer (Neo4j), data-driven decision intelligence (Python/analytics/simulation), AI assistance (LangGraph/RAG/LLM), and enterprise deployment.

How trustworthy is the current Experience Edition?

It is trustworthy as a faithful demonstration of how ARBI is designed to operate — and it is honest about its data. Every figure is labelled illustrative; nothing here should be read as a measurement of a real company yet.

Our principle: transparency over hype. If something here isn't built yet, we say so. What you can rely on today is the experience, the decision method, and the trust framework — and a clear, honest plan for everything else.