Government-adjacent

Records that legally cannot leave the building.

Document intake and private knowledge search for ministries, regulators, SOEs and contractors whose files cannot be sent to a public model. Air-gapped or customer-controlled tenancy. Arabic and English. Open weights only.

The operational trigger

Tenders, correspondence, case files, inspection reports and citizen packs sit in silos. Staff search is slow. Outsourcing OCR or chat to a multi-tenant cloud is blocked by policy, classification, or both. The institution already operates hardware. It does not need another SaaS region. It needs inference that never egresses and an audit trail that stays on premises.

bizfylabs — use case

Air-gapped document intelligence architecture

Air-Gapped Architecture & Validation Flow

If the network cable is pulled, does the answer still exist, and can you show who asked?

The wedge

Document intake and private knowledge search on Foundation and DocxIntel: classify and map incoming packs; ask questions with citations back to source regions. AgentLab only when workflows and named approvers are defined, not on day one.

Enterprise boundaries / Guardrails

What we do not claim

Not a tender SKU for every jurisdiction. Not a replacement for the case-management system of record. Engagement starts with a named workflow and a security pack under NDA.

Why the cloud answer fails here

Classification and sovereignty arguments fail if keys or weights sit with a foreign-operated model. A residency checkbox on a public stack does not survive an air-gap test.

What deployment looks like

Air-gapped or locked-down customer tenancy. Offline update bundles. SBOM per release. Nothing calls home. The customer holds keys. A human remains in the loop for consequential actions.

Controls in this workflow

These controls run on Private AI Foundation for every document ask. AgentLab, if used later, shares the same path.

Packs stay inside the customer perimeter. Inference does not require a network path out.
You hold the encryption keys in every tier that touches your data.
Every document ask goes through the Model Gateway. There is no side door to a public model.
Policy and DLP evaluate the request before weights run.
RBAC names who may invoke which model class on which data class.
DocxIntel writes per-field confidence and citation back to the source region. Originals are preserved.
Consequential AgentLab actions wait for a named approver. Nothing silent-posts to your core systems.
Immutable audit is written on the same install: who, what, which model, which policy, outcome.
Apache 2.0 and MIT weights only. Licence text and an SBOM ship on disk per release.

The paid proof of value

Thirty to forty-five days. Fixed fee. In the customer environment. One document type. Agreed real volume. Success criteria and conversion price agreed in writing before starting. If it misses the threshold, the customer keeps the report and owes nothing further.

Success criteria may include offline operation as a hard gate. This is not a free trial and not a sandbox that pretends to be residency.

Key specs

Deployment: Air-gapped / On-prem
Egress: Zero external API calls
Audit: Immutable local log

Deploy in your perimeter

Talk to an engineer about air-gapped, on-prem or single-tenant deployment for this workflow.

Prove it with the cable pulled.

Book an offline demo of the full stack on one machine, disconnected from the network, on your own documents.

Book an offline demo

The full stack on one machine, disconnected from the network.

In your office, on your documents, with the cable pulled out. No competitor's sales engineer can do this.

Start a conversation