Every leading AI model in one workspace. Governed from the first prompt.

AIMan is an AI workspace for the whole company. Everyone signs in once and chats with the leading hosted models — Claude, GPT, Gemini, Cohere — and with open models that run on dedicated GPU servers in Germany, in the same interface. Answers can come from your own documents, web pages and wikis, or from current web results, with the sources cited. IT decides which models are on offer, who may use what, and holds the vendor keys centrally, so nobody needs a private account to work with AI.

Two kinds, in one picker. Open models run on our own GPU servers: Mistral NeMo (the default — fast and good in German), Qwen 3.5 and Gemma 4 with image input, gpt-oss, Phi-4 and Phi-4-reasoning, DeepSeek-R1, Llama 3.1, two code models and a vision model. Vendor models run at the vendor under your company's key: Anthropic Claude, OpenAI GPT, Azure OpenAI, Google Gemini, Cohere and Ollama Cloud. Admins decide which families and versions people see; new models are added as they appear.

It depends on the model you choose, and AIMan makes that choice visible: local models and vendor models are separate tabs in the picker. With an open model, the prompt, the answer and any document passages stay on brandcompete's dedicated GPU servers in Germany, or on your own servers if you run AIMan on-premise. All document processing — reading, indexing, searching and reranking — runs on those servers too, whichever model answers. With a vendor model, the prompt and the passages it needs are sent to that vendor under your company's contract and key. Translation goes to DeepL, and web search sends a search query to a search service. Each customer has a separate installation with its own database.

You create a data source — uploaded files, web pages, or a connected Confluence, Jira or Bitbucket — and attach it to a chat or a project. AIMan splits the documents into passages, indexes them for both meaning and keywords, and for every question picks the most relevant passages, reranks them and hands them to the model with the instruction to cite them. The answer carries numbered references to document and page. If nothing relevant is found, AIMan says so instead of letting the model guess. You can also restrict a chat to the documents only. Supported formats are PDF, Word, PowerPoint, Excel, text, Markdown, HTML and CSV; scanned PDFs without a text layer are not read yet.

Yes, with web search switched on for the chat. Claude, Gemini and GPT then use their vendor's own search; open models search through AIMan, which reads the best pages and ranks them before the model answers. Either way the answer lists its sources. Every model is also told today's date and what it cannot know, so without web search it says when a question is beyond its training rather than guessing. Admins can switch web search off for an installation.

A project is a shared chat: colleagues, teams or a whole organisational unit work in it together, each tab with its own model and data source. Prompts that work go into the prompt repository, with placeholders for the parts that change, and can be shared the same way, as can data sources and custom instructions. A chat tab can also be exported as a chatbot for an intranet page or an application.

People sign in with Microsoft Entra ID or Google single sign-on, restricted to the email domains and directories you name, or with email and password. Every account has a role, and roles carry the rights to create chats, data sources, prompts or chatbots, manage models, users or keys. What people create is private until they share it with named colleagues, teams, organisational units or the company, for reading, writing or administration. Admins can also allow an account the user interface, the API, or both.

Yes. AIMan has a REST API with streaming answers, and any deployed model can be called by name, so an application can switch from one model to another without code changes. Access to the API is a setting per account. A chat tab can also be embedded on a website as a chatbot.

Three ways. As a managed service, brandcompete runs your installation on its dedicated GPU servers in German data centres. On-premise and managed, brandcompete installs and operates AIMan in your own infrastructure. On-premise and self-operated, your IT runs it under licence, with releases and support from brandcompete. In every case each customer has its own installation, with its own database, search index and network.

That depends on how you run it and on how many people use it. Write to sales@brandcompete.com and we will put a figure against your case. Vendor models are used under your company's own keys, so their usage is billed by the vendor to you, without a mark-up from us.

With a pilot. About one week to set up your installation with your single sign-on, your branding, the models you want and a first knowledge base; four to six weeks in which one department works with it on real tasks; then about a week for a usage report and a roll-out plan. You decide on experience rather than on a demo.