Home About Studio Products Agentic AI Publications YouTube Neuronia Founder Skills Certifications Contact Privacy Policy

Theme

Agentic Enterprise Command Center
Governed agentic workflows.

Every case below is a governed workflow: evidence retrieval, model reasoning, policy checks, human approval, cost control, and an audit trail. The browser only ever receives a safe projection. Nothing here is a black box.

This is a private evaluation entry point for selected workflows. Access is granted per company.

Demo cases run on free AI models and fall back to synthetic workflowsWorking trials run live models behind a company loginFail closed by default, never a silent downgrade

Workflows

Platform

Built on AWS

Live Amazon Bedrock behind a company login, with automatic failover to Anthropic Claude Direct.

live in gated trial flagship Company trialOpen demo

Thin-file newcomer lending

Mithras Bank

Lends to an expat or a foreign student with no local credit file, using alternative evidence, an affordability check, a fair-lending guardrail, an officer approval gate, and a step-up ladder. Two ways to run it: the trial workflow runs live models behind company trial access, and the demo workflow runs the same fourteen steps on a free or deterministic path, with the provider that answered labelled on every run.

Trial: Amazon Bedrock Haiku 4.5, live. Demo: free or deterministic, labelled per run.

Open workflow
live in gated trial Company trial

Payment reconciliation

Mithras Bank

Investigates a payment exception, weighs the evidence on the case, and recommends a governed resolution with the approvals it needs. The model recommends and governance decides.

Amazon Bedrock, with Anthropic Claude Direct as failover

Open workflow
Platform

Built on Azure

Retrieval-grounded policy advisory on Azure AI Foundry and Azure AI Search, when the Azure track is configured. A governed sample is returned and labelled when it is not.

live Company trial

HR policy assistant

StraitsWorks Group

A governed HR advisory. When the Azure track is configured, policy evidence is retrieved with Azure AI Search and reasoned over by Azure AI Foundry, both server side; if Foundry cannot answer, the same question goes to Anthropic Direct, and if neither can, a governed sample is returned and labelled as such. Every run names the provider that answered. The page shows the recommendation, the evidence used, what is missing, and the approvals required.

Azure AI Foundry when configured, Anthropic Direct as fallback; governed sample otherwise

Open workflow
Platform

Open demos

Where a case uses a model, the demo path makes real calls to free AI providers with automatic failover, and drops to a deterministic synthetic workflow if none answers. Every run names the provider that answered. No model cost to you.

live on synthetic data Open demo

Telco storefront

Singa Telco

Browse, cart, checkout, and a customer assistant that raises CRM tickets into the command center, where the payment and governance workflow picks them up.

Synthetic data

Watch the walkthrough
planned Open demo

Home-loan anomaly detection

Mithras Bank

Flags irregularities across a loan book, such as income inconsistency, valuation mismatch, or duplicates, with explainable reasons and a governed review queue for the severe ones.

Synthetic preview

Planned. No runnable workflow today.
planned Open demo

Insurance policy checks

FairEast Assurance

A governance-team view that checks policies against the controls they are meant to satisfy and surfaces the gaps that need a human sign-off.

Synthetic preview

Planned. No runnable workflow today.

How a governed workflow is put together

Three layers, deliberately separated so a second workflow reuses the engine rather than copying it.

Domain experience

The screens and the market pack for one industry. This is the layer that changes per use case.

Agentic workflow

The ordered steps, the checks between them, the attempt budget and the human gate. Declarative, so a new workflow is mostly configuration.

Enterprise AI engine

One governed path for every model call on the platform: provider selection, readiness, spend reservation, output validation against a bounded contract, and the audit trail.

  • The model recommends; rules and humans decide.
  • A failed attempt is recorded, not hidden.
  • Synthetic and live are labelled per run, never silently swapped.
  • Unlisted fields never leave the backend.
  • No provider key ever reaches the browser.

Controls on every trial

Time-boxed trialPer-trial call budgetHuman approval gateAudit trailProvider and path labelled per runDeterministic fallback

Thresholds, budgets and environment details are set per trial and are shared with the company that holds the trial, not published here.

How access works

Trial credentials are issued privately by ViRan Digital Labs, per company, for a fixed window. There is no public sign-up and no credential form on this page. Corrections, questions and access requests are handled by email, outside the product.

Request access

What this is, plainly

The companies shown (Singa Telco, Mithras Bank, FairEast Assurance, StraitsWorks Group) are fictional demo brands. No real customer data appears anywhere on this platform.

  • Synthetic and live are always labelled, per run.
  • Demo cases cost nothing. Working trials run under a daily spend cap.
  • Planned cases are listed as planned and have no runnable workflow today.

Governed Agentic AI and enterprise Cloud AI on Amazon Bedrock and Azure AI Foundry — built day to day with Claude, Microsoft Copilot Studio, Microsoft Foundry, Codex and NotebookLM.

Agentic AIAmazon BedrockAzure AI FoundryCloud AIAI GovernanceClaudeMicrosoft Copilot StudioCodexNotebookLM