A Microsoft Copilot alternative you own

AI assistant

A private AI assistant for the documents that should not leave your network, on models you host, behind the Entra ID sign-on you already run.

Replace Microsoft 365 Copilot Book a 30-minute briefing

Replacing: Microsoft 365 Copilot, by Microsoft Corporation
With: Open WebUI, LibreChat, vLLM
Runs on: Your servers, or Pilae Cloud in Switzerland and the EU
Migration: Included in onboarding, at a fixed price

Why teams leave Microsoft 365 Copilot

The documents that matter most

Legal files, patient records, unpublished results. With the model on your own machines, a prompt and its attachments never leave your network.

Your choice of model

Open-weight models you pick and pin on your GPUs, a hosted API where you accept the trade, or both side by side. Switching is a configuration change, not a licence change.

Keep Microsoft 365, add a private option

Staff sign in with their Entra ID account. Copilot can stay for everyday work while the confidential material goes to the assistant you control.

Microsoft 365 Copilot vs Open WebUI operated by Pilae

What is being compared	Microsoft 365 Copilot	Pilae
Where prompts are processed	In the Microsoft 365 service boundary, within the EU Data Boundary for eligible EU customers.	On your premises or on dedicated machines in any of 12 Pilae Cloud regions, six of them in Switzerland and the EU. Air-gapped on request.
Where it works	Inside Word, Excel, PowerPoint, Outlook and Teams, grounded in your Microsoft 365 data through Microsoft Graph.	A chat interface in the browser and on mobile, with retrieval over the document sources you connect. Not built into Office apps.
Models	Models chosen, hosted and updated by Microsoft.	Open-weight models you pick and pin, plus any hosted API you choose to connect, each one a deliberate setting.
Use of your data for training	Prompts, responses and Microsoft Graph data are not used to train foundation models, per Microsoft.	Pilae does not train models on your data. Local models run on hardware you control, and chat history stays in your own database.
Sign-on and access	Entra ID, with Microsoft 365 permissions and sensitivity labels respected.	Entra ID or Keycloak over OIDC, with groups deciding who reaches which model and which document collection.
Operation	Fully run by Microsoft as part of Microsoft 365.	Run by the Pilae Agent and Pilae engineers: planned upgrades, daily backups restore-tested every month, 60-second probes.
Pricing	Per user, per month, as an add-on to eligible Microsoft 365 plans.	On request, sized to your hardware and your team. No per-prompt charge.

A private alternative to Microsoft 365 Copilot

Microsoft 365 Copilot is well integrated: it drafts in Word, summarises in Teams and answers from your mailbox, and it respects the permissions your tenant already has. For most everyday work that is a good fit. The harder question is the material some organisations may not send to any cloud service at all, whatever the contract says. Our jurisdiction page sets out the legal side fairly.

With Pilae, a private assistant runs where that material already lives: on your own servers, or on dedicated machines in any of 12 Pilae Cloud regions, six of them in Switzerland and the EU. Open WebUI gives staff a familiar chat interface, with LibreChat where several model providers need one front end. The model runs on hardware you control, served by vLLM. The full offer is on the private AI page, and the comparison with a hosted assistant is on our ChatGPT Enterprise alternative page.

Self-hosted AI next to Microsoft 365

You do not have to leave Microsoft to have a private option. Staff sign in with their Entra ID accounts, set up through our identity service, and groups decide who reaches which model and which documents. We size the hardware before installing anything. The Pilae Agent then plans every upgrade, waits for your approval and records each change. Monitoring runs every 60 seconds and alerts reach Pilae engineers around the clock.

Onboarding is fixed-price. Pricing is on request: tell us about your team and we come back with a sizing and a quote.

Moving from Microsoft 365 Copilot to Open WebUI

Decide what stays with Copilot
Which work is fine inside Microsoft 365 and which material should never leave your network. The line is drawn with you before anything is installed.
Size the hardware against the model
Which models your GPUs can serve, at what context length, for how many people at once. We say plainly where a local model is weaker than Copilot.
Put it behind Entra ID
The app is registered in your tenant, groups map to models and document collections, and nobody gets a new password.
Connect the confidential sources
The file shares and folders that hold the sensitive work are loaded into knowledge collections, each one open only to the groups that should see it. Answers cite the documents they come from.

The apps that replace Microsoft 365 Copilot

Open WebUI

A chat interface over models you host, so prompts and the documents people paste into them never leave your network.

Replaces ChatGPT Team, Microsoft Copilot

LibreChat

One chat interface over every model your organisation allows, local or hosted, with the keys, the history and the choice of provider held on your side.

Replaces ChatGPT Team, Poe

vLLM

An OpenAI-compatible inference server for open-weight models, run on GPUs in Switzerland, the EU or your own datacentre, so prompts and answers never reach a model vendor.

Replaces OpenAI API, Azure OpenAI

Questions

Does this replace Copilot inside Word and Outlook?

No, and we say so up front. Open WebUI and LibreChat are chat interfaces, not Office add-ins. What they give you is an assistant for the material you would not send to any cloud service, with the model on your own hardware.

Can we run both?

Yes, and for many organisations that is the right setup. Copilot stays for everyday work in Microsoft 365, and the private assistant takes the confidential documents. Both use the same Entra ID accounts.

Are open-weight models as good as Copilot?

For drafting, summarising and answering from your own documents, current open-weight models do the work most teams ask for. For the hardest reasoning tasks the frontier hosted models are still ahead. We size against your real use cases and tell you where the gap is before you commit.

Which of the two chat apps do we need?

Open WebUI is the interface most teams start with, with retrieval built in. LibreChat suits organisations that want one interface over several model providers with per-team limits. Behind either, vLLM serves the model to many people at once. Open WebUI ships under its own BSD-3-based licence with a branding condition above fifty users; LibreChat is MIT.

Do we need our own GPUs?

No. We run the models on dedicated GPU machines in the Pilae Cloud region you choose, if you do not want hardware on your premises. If you already have GPUs, we size against them first.