The AI assistant that stays in your office.
Sovereign, GDPR-compliant and fit for professional secrecy. Berta runs on its own hardware on your premises — not in someone else's cloud. Your client data does not leave the building.
The ground under cloud AI has started to shift
2026 has shown how little you can plan around externally hosted AI. If you work under professional confidentiality, you can no longer rely on it.
Availability is not guaranteed
Even leading US providers had to suspend access to their most capable models practically overnight — worldwide, for every customer. Building your firm's operations on a service like that means building on someone else's ground.
Zero retention can no longer be taken for granted
For the newest generation of models, providers now require that inputs and outputs may be stored for a retention period — otherwise access stays closed. For data under professional secrecy, that is simply not acceptable.
The real costs are still coming
Access to the strongest models sits in the hands of a few providers. Their enormous investments will be recouped through prices over the coming years — borne by customers. None of that is predictable.
For confidentiality-bound professionals, the conclusion is clear: non-local AI can no longer be used responsibly.
Berta closes that gap — predictable costs, 100 % local, no residual risk.
Why most AI tools are a non-starter for your firm
Professional-secrecy risk
ChatGPT, Copilot & co. now offer EU data residency — but only via a data-processing agreement with the provider, and a residual risk remains: processing at a third party, its sub-processors, its access. For client and patient data under special professional secrecy (German §43a BRAO, §203 StGB, medical confidentiality), that residual risk is precisely the sore point.
A limit or a compromise
There are flat rates for businesses — but with a usage cap. The GDPR-compliant tiers are more expensive and capped; the cheap standard plans aren't legally compliant. On top of that, providers keep changing models, prices and terms. That has little to do with planning for the years ahead.
Vendor lock-in for life
Your memory, your workflows, your custom GPTs — it all sits with the provider. Switch, and you lose everything. Change the terms tomorrow, swap out the model, get sanctioned — and you lose everything then, too.
Berta is different.
The same work you'd expect from good AI — but under your control, in your building, at a fixed price.
| Cloud AI (ChatGPT, Copilot …) | Berta |
|---|---|
| EU residency only via a processing agreement with the provider — residual risk remains | Data stays in the building — no provider, no residual risk |
| Tiers capped or not GDPR-compliant; terms keep changing | Fixed monthly price, no usage limit |
| Negotiate a processing agreement with the provider | GDPR-compliant, you are the controller |
| Model updates without warning — access revocable at any time | You control when (and whether) updates happen — and access stays |
| Office integration only with extra licences | Microsoft 365 + Google Workspace built in |
| Generic “AI for everyone” | Tailored to professional workflows |
What Berta does for your firm
Client correspondence starts as a draft
Berta reads incoming messages and proposes replies — in your tone, with reference to the file. You confirm or edit, then it goes out.
File research in seconds
Hand over a PDF, Word or Excel file, ask “what does it say about X?” — and get an answer with a page reference. Not invented, but cited.
Meeting prep, automated
Berta pulls the upcoming client meeting from Outlook or Calendar, fetches the latest files for it and suggests talking points.
Multimodal processing
PDF, scan, photo, Word, Excel — Berta understands it all. Including the handwritten note in the margin.
Template drafting
Mandatory details, standard letters, processing-agreement templates — Berta fills them with client data, in your house style.
Your firm's knowledge base
What Berta learns once — your tone, your templates, your standard answers — it keeps. But only for your firm. Not for the provider.
Berta runs on your hardware. At your place.
During setup we deliver your Berta hardware to your premises, connect it to your network, configure the link to your Microsoft 365 or Google Workspace and train your team. After that, Berta is ready to go. The device sits with you — you have physical control.
- Plug and play. We install, you use. No major IT project.
- Updates come from us. We test, we deliver, you decide when they're installed.
- Hardware is part of the solution. You don't buy a software subscription plus extra hardware — it's all in the package.
No arrow pointing outward — except to services your firm already uses.
As local as the potatoes from the farm down the road.
What grows here stays here. No long journey, no middleman, no surprise when you unpack it.
Two sizes — matched to your firm
Berta Box
For 1–3 power users who work with Berta every day.
- Compact hardware with generous unified memory
- Compact, local language model
- Multimodal: text, image, PDF
Berta Studio
For up to 5 power users, high file volumes, complex research.
- Powerful hardware with maximum unified memory
- Large, local language model, open weights
- Long contexts — around 400 pages at once
Berta is not a self-service product.
We come to you, take a look at your firm, work out whether Berta is a fit — and then deliver the right tier. No subscribe button, no sales funnel. A conversation.
Or call directly: Fabian Kübler, +49 [phone to follow]