Ugrás a tartalomhoz
← Back to Nortinia AI Asszisztens
AI
Nortinia AI Asszisztens · Enterprise AI assistant

Enterprise AI assistant with audit logs and an on-premise option

The enterprise variant of Nortinia AI Assistant adds SAML/SSO and SCIM provisioning into your existing identity provider, detailed audit logs of every question and answer, and fully local (on-prem) execution on a GPU box. Where data cannot leave the corporate network for even a millisecond, this is the version that delivers — while the user experience matches SaaS.

View details

What it does

SAML/SSO integration with Microsoft Entra, Okta, Google Workspace, ADFS
SCIM 2.0 user and group provisioning
Detailed audit log: who asked what, when, and what came back
Fully on-prem execution on a GPU box, no internet required
Choice of vector database (pgvector, Qdrant, Weaviate, Milvus)
Permission granularity down to module, field and record level

How people use it

Enterprise IT internal assistant

An enterprise IT team wants to answer colleagues fast from internal docs, ticket history and change logs. The Enterprise AI Assistant authenticates via SSO and only sees documents within that org unit — every answer comes back with source links, and IT leadership tracks usage through the audit log.

Financial sector

A bank-internal assistant works from regulatory reports, risk policies and internal directives. Data never leaves the network, on-prem execution gives full data sovereignty, and every question-answer pair is auditable for central-bank / regulatory inspection.

Public-sector customer

A public-sector customer wants to make legislation and internal directives searchable through an AI assistant for civil servants. The Enterprise AI Assistant runs in a Hungarian data centre, integrates via SAML with the national digital ID, and records every query in line with public-procurement audit requirements.

Multinational subsidiaries

A multinational parent wants one shared executive assistant used by 12 subsidiaries, with each user seeing only their own unit. SCIM automatically syncs users and groups from HR, and the audit log is segmented per subsidiary.

Frequently asked

What GPU hardware is needed for on-prem deployment?

It depends on model size. For Llama 3 70B or Mistral Medium-class open-source models, 2× NVIDIA H100 or 4× A100 (80GB) is recommended; for smaller 7B-13B models, a single RTX 6000 Ada is enough for a 10-20 person team. Nortinia provides a sizing study based on your target user count before the POC.

Which SAML / SSO providers are supported?

Microsoft Entra ID (former Azure AD), Okta, Google Workspace, ADFS, Keycloak and any standards-compliant SAML 2.0 + OIDC IdP. SCIM 2.0 provisioning is available with most of the above; manual group sync is an optional fallback.

How much data is needed for meaningful answers?

The RAG engine returns useful answers from as little as a few hundred pages and scales to millions of documents. ERP queries are always live lookups, so there is no lower data threshold there — existing data is immediately usable.

Can we run multiple environments (DEV, STG, PROD)?

Yes. The enterprise install supports three environments (DEV, STG, PROD) out of the box with separate databases, keys and audit logs, so testing and change management stay inside your standard enterprise process. Feature flags allow you to roll out experimental features between environments.

More Nortinia AI Asszisztens solutions

Interested? Let us chat for 15 minutes.

Send a quote request and hear back within 24 hours. No commitment, no pressure.

View details