Sovereign AI infrastructure

We provide the foundation. You keep control.

Many AI projects fail before they even begin, because servers, interfaces and security are missing. We provide the entire AI environment and operate it, in your tenant or in our infrastructure in the EU. You can steer it in the chat, with our DevOps agent BRIDGE OPS.

Book a call See the tiers

Every capability that uses AI needs a secure foundation. Companies without a large IT department should be able to use AI too, without tying themselves to a single provider. That’s why you get everything from us: the environment, models, integration and operations.

Name: Decided on 28 Aug: externally, the DevOps agent is called “BRIDGE OPS” and the infrastructure is only described as “Sovereign AI infrastructure”. “Rebecc.AI” and “RebeccAI” are not used externally (a person’s name, risk of confusion; README item 17 done). Neither name is in the TechCo IP catalog. Clarify the hosting model (our own EU infrastructure or operation by the provider), an open question from Lucas’ check-out on 31 Jul.

Three tiers

As sovereign
as you need it.

You decide where your models run and where your data is stored. If your needs change, you can switch tiers.

Managed Cloud

Start quickly

You use major cloud language models in an EU environment that we operate in line with ISO 27001. Monitoring and cost control are part of it. This tier suits smaller teams. Live

Private AI

In-house models

Local language models run in your environment, you get your own knowledge base, and you can add cloud models if you wish. This tier suits larger user groups. Pilot

Edge and Air-Gap

No internet at all

Large language models run on your own hardware, fully offline. This tier is meant for areas where no data may leave your premises. In development

ISO 27001 certified and GDPR compliant. You can choose between several cloud providers and local models and are not locked into any one provider. Name providers? (Azure AI Foundry, AWS Bedrock Frankfurt, Mistral, local models; platform note, June) · Confirm C5 status

Tiers from the “Capability Development” offer (Phase 6): Starter up to 20 users (€590 per month), Professional up to 100 users (€1,490 per month), Enterprise unlimited (on request), each plus cloud costs. Confirm the status of each tier. Showing prices publicly: decision pending.

What’s included

We provide everything,
and one person runs it.

BRIDGE ONE, BRAIN and every agent you use run on this infrastructure, and with them every capability on the capability landscape.

In your tenant

BRIDGE runs in your cloud environment or in ours. Your data stays within the boundaries you set.

Ready in days

We set up a prepared AI environment, largely automatically. So your first use case doesn’t need a major IT project. Check the claims “start in one day” and “around 95% infrastructure as code” (Huddle 5 Jun)

No lock-in

You can choose from several language models, including locally run ones. If you switch models, your knowledge stays where it is.

Traceable

Every answer shows its source, and every action is logged. Approvals and human oversight are built in.

Connects what you have

BRIDGE connects to SharePoint, Teams, Jira, Confluence and your line-of-business systems. Check the list of available connectors

One contact person

One person knows your environment and takes care of operations, security and further development. BRIDGE OPS takes on the routine work.

Operations by agent · BRIDGE OPS

You steer your environment in the chat.

You write in Teams or Slack what you need, just as you would write to a colleague. BRIDGE OPS, our DevOps agent, gets it done in your Microsoft environment, with exactly the permissions you give it. Before it changes anything, you see what would happen.

In use at BRIDGE itself

Teams · BRIDGE OPS

Anna Berger joins the sales team on Monday. Please set up her access the same way as for the others on the team.

Preview, nothing has been changed yet:

Create an account in Entra ID
Add her to the Sales and CRM groups and the Sales channel in Teams
Assign a Microsoft 365 license

Permission change: needs your approvalCan be rolled back

ApproveDecline

Illustration, no customer data.

What it handles
Joiners and leavers, access and permissions, setting up and shutting down systems, certificates, incidents and costs per company. On top of that, you get a report every month.
What it may not do
It never acts with more permissions than you give it. An approval catalog defines what runs without asking, what needs approval and what never happens automatically.
Preview and rollback
Before every change, you see what would happen, and every change can be rolled back. Every step is logged, including declined attempts.
Read first, then act
On the first day it can only read, and that is enforced technically. From the second week, it takes on two or three routines, and from the second month, it also responds to alerts from monitoring and the ticket system.

BRIDGE OPS works in Microsoft Azure and during business hours. Production, machines and on-site hardware are not part of it, and it replaces neither ERP nor the help desk.

Based on the BRIDGE OPS customer info and pitch (30 Aug) and the landing page concept (28 Aug). Name decided on 28 Aug. According to Heiko on 21 Oct, Teams and Slack are in use and a customer is asking for it; the August documents still call this a target. Before publishing, document: a positive list per area (mainly Azure has been tested in production), log export and second approval for deletions and permission changes (so far contract items, not shown as working features). Only give prices (setup €5,900, €1,450 per environment and month) on request.

Where should your models run?

Together with your IT team, we work out which tier fits and set up the environment.

Book a call Capability Landscape