# Your own AI chat, inside your company network. Connected to your systems, with the model you choose. > Your own AI chat: in your cloud, behind your VPN, connected to your systems through MCP, with the model you choose. In the browser and on the phone. Source: https://schuetze-web.com/integrations/private-ai-chat One chat for your whole team: it runs in your cloud, is reachable only through your VPN and works with your knowledge, data and systems. Which AI answers in the background you decide, and you can change it at any time. ## What changes ### One app for everything Knowledge, data, website and your systems in one chat. Staff don’t switch between tools, and they don’t copy company data into public chat apps. ### Under your control The chat runs in your cloud, behind your VPN, with your company logins. Every request is logged, and you decide how long it is kept. ### Free choice of model Which AI answers you decide, per team or per task. If a better or cheaper model comes along, you switch, and nothing changes for your staff. ## In the browser, on the computer, on the phone One web app that runs in the browser and installs like an app on computers and phones. Same login and same history everywhere, reachable only through your VPN. ### In the intranet Open it in any browser on the company network, without installing anything. ### On the computer Installed as an app on Windows and macOS, with its own window and icon in the dock or taskbar. ### On the phone Installed from the browser on iPhone and Android, with your VPN switched on. No app store needed. ## How it connects ### Your team Browser, computer and phone Through your VPN and company login ### Private AI chat Runs in your cloud. Applies the rights of each person, masks personal data where you want it and logs every request. ### Your systems Knowledge, data, website, email, CRM, ERP Connected through MCP, each with the least rights it needs ### The model Cloud models in your region, a European provider or an open model on your own server See the options below ## Which AI answers The chat works with the models your company prefers, and can use several side by side. ### Directly from the providers Claude from Anthropic, the ChatGPT models from OpenAI. ### Through the cloud you already use Microsoft Azure, Amazon Bedrock or Google Vertex AI, each in a region you choose, for example in your own country. ### A European provider Mistral, from France. ### On your own server Open models such as Llama, Qwen, DeepSeek or Mistral. Nothing leaves your network. - HR questions · Open model · Own server - Contract review · Strong cloud model · Local region - Translations · Smaller cloud model · Local region Rules like these are set once by you; staff simply ask. ## One question, two ways Support wants a reply to an angry customer email. ### Today - Someone copies the customer email into a public chat app - Name, address and order number leave the company with it - The answer knows nothing about your guidelines or the order - It is pasted back by hand and adjusted - Nobody knows which data went where Fast, but blind to your systems, and the data protection officer finds out last. ### With your own AI chat The same speed, with your order data and guidelines, personal data masked and every step logged. ## Secure by default ### Behind your VPN Reachable only from your network or through a zero-trust access. Not on the open internet. ### Company logins Sign-in with your existing accounts and two-factor authentication. Rights come from the groups in your directory. ### Personal data masked Names, addresses, email and account numbers can be masked before a request goes to a cloud model, and put back in the answer. ### Not used for training Through business contracts with the providers your data is not used to train their models. ### Logged and deletable Every request is logged. You set retention periods, and conversations can be deleted. ## Set up in three steps ### Choose cloud and models We set the chat up in your cloud, for example Azure, AWS or Google Cloud, and connect the models you choose. ### Connect login, VPN and systems Company login and VPN in front, your systems behind it through MCP, with the rights from your directory. ### Roll out to the team Staff open the chat in the browser or install it on their computer and phone. ## You stay in control An admin view shows which models each team may use, what they cost and what happened. Budgets per team keep costs predictable, and you pay the models by use instead of a licence per person. - Sales · Cloud models, local region · €400 / month - HR · Own server only · No limit - Management · All approved models · €800 / month ## Questions ### Why not simply the business version of ChatGPT or Claude? Those are good products. Your own chat makes sense when you want it in your own cloud and network, several models side by side, your own systems connected through MCP, local models for sensitive data, or personal data masked before it leaves. ### Where is the data processed? The chat runs in your cloud. Requests go to the models you approve, in the regions you choose, or stay on your own server with a local model. ### Which models can we use? Claude, the ChatGPT models, models through Microsoft Azure, Amazon Bedrock or Google Vertex AI, Mistral, and open models such as Llama, Qwen or DeepSeek on your own server. Several can run side by side. ### Do we need a VPN? The chat should not be reachable from the open internet. Most companies use their existing VPN; a zero-trust access works as well. ### How does the app get onto phones? It is a web app: staff open it in the browser and add it to the home screen. No app store, and it only works with the VPN on. ### How does masking personal data work? Before a request goes to a cloud model, names, addresses, email and account numbers are replaced with placeholders. The answer comes back with the details filled in again, only inside your network. ### What does it cost to run? The hosting in your cloud and the use of the models, paid by use rather than per person. The admin view shows the costs per team, and budgets set an upper limit. --- Contact: book an intro call at https://schuetze-web.com/book or write to simon@schuetze-web.com.