The engines under the hood.
An extended intelligence layer powers every ALTeX Hub product. Each accelerator does one job well, and every product can use it.
01The accelerators
Anode
Anomaly detection on any dataset. Upload a file and Anode works out what the data is, recommends the fields to test and runs an ensemble of unsupervised models (Isolation Forest, One-Class SVM, z-score and IQR) to surface the lines that don’t fit. No set-up, no context needed.
Chat with Data
Plain-English questions on live data, with automatic charts. No SQL.
Correlator
Maps where control failures cluster across business units and periods, revealing patterns no single test would show.
ML Engine
Scores how likely each exception is to be real, forecasts trends and retrains continuously on your decisions.
ALTeX SDK
Reusable integration components and governed agent development, so new insights reach production faster.
Six layers of smart technology, one governed stack
Each layer catches what the one below it misses.
- 06AI agents
Orchestrate the stack: gather, test and draft, supervised.
ExampleA whistleblower tip triaged in 30 minutes
- 05LLMs
Read and write language: contracts, claims, emails.
ExampleBilling above contract clause 7.2
- 04Machine learning
Find what is unusual in many ways at once.
ExampleA vendor ranked 1 of 2,140 by risk
- 03Process mining
Show the real process flow from system logs.
ExampleInvoice before order; approvals bypassed
- 02Analytics
Statistics against a baseline.
ExampleBenford’s law; a scrap rate twice that of peers
- 01Rule-based insights
Known schemes written as tests.
ExampleThe same invoice paid twice; split orders
- 00Data foundation
ERP plus dark data: logs, documents, plant and dealer systems, through read-only connectors.
Agents gather, test and draft. People decide.
Supervised autonomy: agents never decide, contact anyone or sign off. Every step is logged and replayable, and data stays read-only in your environment.
What the explainer shows.
- Every ALTeX Hub product runs on one governed stack. Each layer catches what the one below it misses.
- A tip is logged, the data pulled read-only and the case scored. Then it waits for a person: the auditor approves.
- Agents never decide, contact anyone or sign off. Every step is logged and replayable, and the data stays read-only.
Six delivery accelerators, ready to order
Control intelligence library
750+ configurable rule-based and analytic insights, each mapped to a scheme, a control and a data source.
Get a quote →Read-only connectors
SAP ECC, S/4HANA and other sources into one common data model.
Get a quote →Explainable ML scoring
Vendors, employees and plants scored daily, each with plain-language reason codes.
Get a quote →LLM document reader
Contracts, invoices and claims read in many languages, with confidence scores.
Get a quote →Supervised agents
Agents gather, test and draft; every step is logged; people approve every decision.
Get a quote →Issue-to-closure and governance
Owners, deadlines and closure evidence, with privacy and EU AI Act documentation ready.
Get a quote →Five shifts that will define the next audit cycle
Continuous assurance
Testing every transaction becomes the default; annual sampling becomes the exception.
Agents under supervision
Agents gather, test and draft; auditors review, judge and decide.
Smart technology against smart fraud
Document forensics and deepfake checks built into payment controls.
Audit of algorithms
Models used in finance and audit need assurance themselves.
Dark data in scope
Logs, documents and plant data join the ERP as standard audit evidence.
Agents in anti-fraud work: 8% today, 31% expected by 2028 (ACFE and SAS, 2026).
Where is your money hiding?
Three minutes to an estimate. Thirty minutes to a plan. Six weeks to valued findings on your own data.