← All S2 solutions

PRIVATE AI / PRIVATE INFRASTRUCTURE

S2 Private AI

Your AI in your business. Ready for the first question, open to everything that follows.

Indicative target from €20,000

Hardware target excludes VAT and requires configuration, availability and scope confirmation. AI Care is proposed from €300/month; warranty and SLA are scoped separately.

PRIVATE AI / LOCAL
Infrastructure under your control

WHAT YOU GET

Business AI on infrastructure you control.

A private AI environment built around your needs, with client-owned hardware and an open path to integrations, retrieval and workflow automation. S2 delivers assessment, installation, integration, support and engineering.

01

You own the infrastructure

Purchase the device outright. S2 earns through delivery, integration and support.

02

A useful first deployment

Start with named users, an agreed communication channel and a practical use case.

03

Room to grow

Add approved data sources and workflows as your requirements mature.

PROPOSED ARCHITECTURE

Open components. Controlled integration.

The S2 AI Stack is a technical starting point. Final choices depend on requirements, licensing and implementation review.

RECOMMENDED

NetBird

Private network access; configure and review access boundaries.

RECOMMENDED

vLLM

Local model serving for a validated hardware and model combination.

RECOMMENDED

Open WebUI

Private user interface; initial scope targets up to 20 named users.

RECOMMENDED

Keycloak

Identity and access management for the deployment.

RECOMMENDED

PostgreSQL + pgvector

Structured data and approved retrieval context.

PILOT

n8n

Workflow orchestration; review license terms for the intended use.

PILOT

MCP gateway

Controlled tool access with explicitly scoped permissions.

PILOT

Hermes Agent

Optional pilot only after license, runtime and security validation.

Element/Matrix is the recommended private business channel. WhatsApp and Signal may use external infrastructure, so their data flows and provider terms must be reviewed separately.

FAQ

Frequently asked questions

Who owns the hardware?
The client purchases and permanently owns the device.
Does every conversation stay on-premises?
Local model processing can keep prompts in the deployment, but external channels or integrations may process data elsewhere. We map these flows before deployment.
Can an agent update an ERP?
Only through an explicitly scoped integration. Read access does not imply permission to alter records or initiate transactions.

THE NEXT STEP

Let’s talk about your AI infrastructure.

We will review your goals and systems, then propose a realistic scope.

Talk with the S2 team
Chat

AI Assistant