Microsoft Foundry

AI in your own Microsoft environment

Most organisations already work with Microsoft 365. That is why we build in your own Azure environment, with Microsoft Foundry. Where data must stay in-house, AI runs locally, on your own hardware.

Why Microsoft

Your data stays yours

When we build in your own Microsoft environment, these five things are a given.

  1. 01

    You already work with it

    Microsoft 365, Teams and Outlook are already there, with your own accounts and permissions in Entra ID. The AI is added to the same environment.

  2. 02

    The model comes to you

    Not your data to the model, but the model to your data. It runs in your own Azure environment, behind your own sign-in and your own network.

  3. 03

    Within the EU

    You choose a European data zone and processing stays there. For each use case we record which region, which model and which storage apply.

  4. 04

    The AI provider sees nothing

    What you enter does not go to OpenAI, Meta or the other AI providers in Foundry. They do not see it and do not train on it. This is set out in the contract. Claude is the exception; see below.

  5. 05

    All the models you know

    More than ten thousand models in one catalogue. If another model becomes better or cheaper, you switch without rebuilding your use case.

In your own Microsoft environment

You stay in control

The model comes to your data, not the other way round. For each use case we record which model, region, permissions and logging apply. Everything stays within the boundary of your own Microsoft environment.

In your own Microsoft environment Inside your own Microsoft environment there are four layers. On top is what we build: AI skills, AI agents, AI automations and the AI knowledge base. Below that Microsoft Foundry, Microsoft Azure and Microsoft 365 with Entra ID. Guardrails and governance run alongside every layer: logging and monitoring, prior agreement on what AI may and may not do, an employee who decides when in doubt, and access only for those with permission. Everything we build is yours, defined as Infrastructure as Code. In your own Microsoft environment within the EU What we build AI skills Reusable skills for recurring work AI agents Agents that carry out tasks AI automations Connections between your systems AI knowledge base Business knowledge safely available to AI Microsoft Foundry Enterprise quality, without needless complexity. Cost and flexibility Microsoft Azure Your data stays in your own Azure environment, within the EU. Data and privacy Microsoft 365 and Entra ID AI only sees what the user is allowed to see Users and permissions Guardrails and governance Logging and monitoring Every action is logged We agree in advance what AI may and may not do If unsure, the agent stops and an employee decides Only people with access can see it Everything we build is yours, defined as Infrastructure as Code. In your own Microsoft environment In your own Microsoft environment What we build AI skills Reusable skills for recurring work AI agents Agents that carry out tasks AI automations Connections between your systems AI knowledge base Business knowledge safely available to AI Cost and flexibility Microsoft Foundry Enterprise quality, without needless complexity. Data and privacy Microsoft Azure Your data stays in your own Azure environment Users and permissions Microsoft 365 and Entra ID AI only sees what the user is allowed to see Guardrails and governance Logging and monitoring Every action is logged We agree in advance what AI may and may not do If unsure, the agent stops and an employee decides Only people with access can see it Everything we build is yours, defined as Infrastructure as Code.
  • Data and privacy

    Your data stays in your own Azure environment, within the EU.

  • Users and permissions

    AI only sees what the user is allowed to see, with your own sign-in through Entra ID.

  • Logging and monitoring

    Every action is logged, so you can always trace what happened.

  • Guardrails and governance

    We agree in advance what AI may and may not do, and where a person decides.

  • Infrastructure and source code

    Everything we build is yours, defined as Infrastructure as Code.

  • Cost and flexibility

    You see what each model costs and decide for yourself when to expand or switch.

No black box and no needless vendor lock‑in.

Not all AI is the same

From the cloud to fully private

Everyone knows ChatGPT. For businesses there is a safer option: the same top models in your own Microsoft environment, where the AI provider cannot see your data. Whatever must never leave the building runs on your own hardware, even offline.

Consumer AI

ChatGPT, Gemini, Claude.ai

Everyone knows it and you can start straight away. What you type goes to the provider of the app, under the terms of the subscription.

Where
With the provider of the app, often outside Europe
Access
The provider, under its own terms
Cost
Fixed amount per user per month
  • ChatGPT
  • Gemini
  • Claude.ai
  • Le Chat

Enterprise cloud

Microsoft Foundry

The same top models, in your own Azure environment. Microsoft runs them in its own data centres. OpenAI, Meta, DeepSeek and xAI do not see your data and do not train on it.

Where
Microsoft data centres, in the EU if you choose
Access
Your own company accounts; the AI provider sees nothing
Cost
Pay per use: you pay for what you ask
  • OpenAI GPT
  • Meta Llama
  • DeepSeek
  • xAI Grok
  • Mistral

Claude is also available through Foundry, but there Anthropic processes the requests itself and there is no European data zone yet.

Private cloud

Microsoft Foundry, dedicated capacity

An open model on computing capacity that runs only for your organisation, reached over a private connection instead of the open internet.

Where
At Microsoft, on capacity nobody else uses
Access
Only your own organisation
Cost
Fixed price per hour
  • Llama
  • DeepSeek
  • Mistral
  • gpt-oss from OpenAI

On-premises

LM Studio, Foundry Local

The model runs on a workstation or server on the company network. Data does not leave the building. We set it up and keep it up to date.

Where
On your own hardware, on the company network
Access
Only your own organisation
Cost
One-off hardware cost, no price per request
  • LM Studio
  • Foundry Local
  • gpt-oss
  • Llama
  • Mistral
  • Qwen

Foundry Local works per device, for one user. For a team we run LM Studio as a server on the company network.

Air-gapped

No internet connection

The same set-up, with no connection to the outside world. Nobody outside the organisation can reach it, foreign governments included.

Where
On your own hardware, without internet
Access
Only those with physical access
Cost
One-off hardware cost, fixed cost for management
  • LM Studio
  • Foundry Local
  • open models

We bring in new models and updates under controlled conditions.

It is usually a mix: the cloud for most of the work, local for the most sensitive data. We advise on which AI suits which data.

The catalogue includes models from

  • OpenAI
  • Anthropic
  • Meta
  • Mistral AI
  • DeepSeek
  • xAI
  • NVIDIA
  • Hugging Face
  • Microsoft

Our sandbox

See it running for yourself

One question goes to five models from five providers at once. Each answer comes with proof that it ran in our own environment.

Our sandbox, the Proeftuin: one prompt goes to five models from five providers at once, showing for each model whether the data stays within the EU
foundry.outdeliver.ai, open by invitation only

Take a look

Half an hour behind the scenes

Our sandbox, the Proeftuin, is where we run everything we build. Tell us who you are and we will invite you for a half-hour look behind the scenes. No sales pitch, unless you ask for one.

The button opens an email with your details filled in. If that does not work, email hallo@outdeliver.ai.

Three ways to start

What does Microsoft mean for you?

Start small, with one process that comes up every week. Choose how: from half an hour behind the scenes to a session at your offices.

01

Look behind the scenes

Free

Half an hour, by invitation

See everything we build running in our sandbox, the Proeftuin. No sales pitch, unless you ask for one.

Request an invitation

02Recommended

Introductory call

Free

Half an hour on Teams

You tell us what you already do with AI. We tell you what fits, or whether it is still too early.

Call with Marcel Logtenberg

Book a call

03

Get started

From €495

At your offices, excluding VAT

An AI quick scan, an AI assessment or an AI assessment plus inspiration session, at a fixed price.

Get started