Brane AIF User Handbook

This handbook explains Brane from your perspective: the workplace in your browser. It describes every feature you can see and use in the chat — and how you can tell, at any moment, where your request was processed. Installation, user management and system operations are your IT department's job and are covered in a separate administration handbook.

1. What is Brane?

Brane is your company's AI assistant. You use it like any familiar chat assistant: ask questions, draft texts and emails, have code explained, upload documents and ask questions about them, research and analyze. The difference from public AI services: Brane runs on a server inside your company — provided and managed by your IT department.

At its heart is the Guardian routing layer. It inspects every single message before it is answered and decides: will this request be answered by the local AI model directly on the box — or may it (if your IT has allowed it) go to a powerful cloud model? If Brane detects personal or sensitive data in the process, that data is — depending on your company policy — either kept out of the cloud entirely or anonymized first: names, email addresses, IBANs and other data are replaced with neutral placeholders before anything leaves the building — and put back into the answer afterwards. To you, the result looks like a normal chat; behind the scenes, the system keeps watch on every message.

For you this means: you can use Brane for real work — including customer data, contracts or internal documents — without having to second-guess what you are allowed to reveal in every message. And you don't have to trust the system blindly: below every answer, a small security badge shows where the message was processed and why (chapter 4). Transparency instead of a black box.

2. Your first login

2.1 Signing in

Your IT gives you the address of your Brane installation. On the login page you will find the „Anmelden“ (Sign in) card („Melden Sie sich mit Ihrem Unternehmenskonto an.“ — Sign in with your company account) with the E-Mail and Passwort (password) fields and the „Anmelden“ (Sign in) button.

Depending on your company's configuration, you may additionally (or instead) see a single-sign-on button: „Mit Microsoft anmelden“ (Sign in with Microsoft), „Mit Okta anmelden“ (Sign in with Okta) or „Mit SSO anmelden“ (Sign in with SSO). In that case you simply sign in with your usual company account — no separate Brane password is needed.

No account yet? The link „Noch keinen Zugang? Hier anfragen“ (No access yet? Request it here) takes you to the „Zugang anfragen“ (Request access) form (name optional, email, password with at least 8 characters). After submitting, the rule is: „Sie können sich anmelden, sobald Ihr Administrator den Zugang freigegeben hat.“ (You can sign in as soon as your administrator has approved your access.) Approval is handled by your IT.

For stubborn display problems, the „Cache & Cookies löschen“ (Clear cache & cookies) button at the bottom of the login page helps: it clears the app's local browser data and reloads the page.

2.2 The welcome tour

On your first visit from a device, Brane greets you with a short welcome dialog: „Hi, ich bin Brane.“ (Hi, I'm Brane.) Three cards sum up what to expect:

1. „Frag mich alles“ (Ask me anything) — write texts, draft emails, explain code, do research.
2. „Schick mir Dokumente“ (Send me documents) — upload PDFs, images, Word and Excel files and ask questions about them.
3. „Deine Daten bleiben bei euch“ (Your data stays with you) — Brane runs in-house; sensitive content is automatically anonymized before anything goes to the cloud.

„Loslegen“ (Get started) takes you into the chat. The dialog appears only once per browser/device (it reappears after clearing browser data or on a new device).

2.3 The interface at a glance

Area	What you find there
Sidebar, left	„New Chat“ (keyboard shortcut ⌘⇧U), „Search“ (opens the chat search, ⌘K), your projects („New project“) and your past conversations, grouped by time period
Chat area, center	In an empty chat: the greeting „Your secure AI assistant.“ / „How can I help?“, below it the tags „On-premise · Private · Sovereign“ and clickable suggestion categories
Input field, bottom	Text field „Type a message...“, next to it the round „+“ button with the tools (files, voice input, Tiefgang, search, image) and the send arrow
Header, top right	Light/dark toggle, „New Chat“, the Feedback button and your avatar with the profile menu: „Settings“, „Insights“ (your personal usage overview) and „About“

Depending on configuration, your IT may also display a custom welcome message on the start screen.

3. Chatting

3.1 Writing and sending messages

• Type into the „Type a message...“ field and send with Enter. Shift+Enter inserts a new line.
• The round arrow button („Send message“) also sends. While Brane is answering, it turns into a stop square („Stop generating“) — one click aborts the answer.
• A message may be up to 200,000 characters long; from roughly 90% a character counter appears above the input field. Signed-in users can send up to 1,000 messages per day, of which at most 500 to premium/pro models. In addition, if you send very many messages in very quick succession, Brane briefly asks you to wait a moment (overload protection).
• You can also send files without any text — e.g. a PDF with the implicit question „What's in here?“.

After sending, you can watch what happens: first, if applicable, „Uploading files...“, then „Analyzing your request...“ — this is the moment the Guardian layer makes its routing decision — then the answer starts to appear. New conversations automatically get their own address (/c/...) that you can bookmark. While an answer is streaming, Enter will not send a new message. And if you reload the page in the middle of a running answer: the part already generated is preserved.

3.2 Suggestions to get started

In an empty new chat, Brane shows eight suggestion categories: Code, Write, Privacy, Analysis, Security, Document, Image, Spreadsheet. Clicking a category fills a matching sentence starter into the input field and unfolds six example questions. Clicking an example question sends it immediately.

The Privacy category deliberately contains example requests with fictitious personal data (Max Mustermann, a sample IBAN) — try one of them and watch the security badge (chapter 4): it's a risk-free way to see anonymization in action. Your IT can replace the example questions with company-specific ones or add its own categories.

3.3 Working with answers

Below or on every answer you will find:

Action	Meaning
„Copy text“	Copy the answer to the clipboard
„Regenerate“	The last answer is replaced by a new attempt (not appended)
Thumbs up / down	Quick feedback on answer quality
Quote a passage	Select text in an answer to refer back to it
Security badge	Shows where the answer was processed — in detail in chapter 4

You can edit your own messages afterwards („Edit message“). Important: the edited message is re-sent, and the conversation from that point on is replaced — everything that came after it is discarded.

3.4 History: search, pin, rename

• ⌘K or „Search“ in the sidebar opens the chat search („Search chats...“). The search looks at conversation titles; hits are grouped into „Today“, „Last 7 days“, „Last 30 days“, „Older“ and „Pinned“. Hovering over an entry shows a preview of the conversation.
• Hovering over a chat reveals the actions „Pin chat“ / „Unpin chat“, „Edit“ (rename), „Export“ (see below) and „Delete chat“ (with an in-place confirmation).
• „Export“ downloads the whole conversation as a Markdown file — handy for archiving or sharing a result. If the chat contains detected personal data, the same safety prompt as for document downloads (chapter 5.4) appears first: the export contains the content in clear text and must be handled per your company policy. Every export is recorded in the audit log.
• If you switch chats while an answer is still being generated, Brane asks: „A response is currently being generated. If you switch now, the generation will be stopped. Do you still want to switch?“
• On a smartphone, the same features live in a slide-up history view.

3.5 Projects: chats with shared context

Projects bundle related chats and give them shared knowledge and shared rules.

• Create: Click „New project“ in the sidebar. In the „Create Project“ dialog („Give your project a name and optionally add instructions.“) you set the project name and, optionally, project instructions. Note in the dialog: „You can upload files directly on the project page after creation.“
• Instructions: On the project page under „Instructions“ — „These instructions tell the AI how to behave in this project. They apply to all chats.“ Ideal for tone, target audience or standing requirements.
• Knowledge: Under „Knowledge“ you upload files („Drop a file here or click to upload“ — „PDF, Word, Excel, PowerPoint, text files, images (max. 100 MB)“; up to 20 files per project). „Text is automatically extracted from uploaded files and used as knowledge in every chat.“ Each file shows its processing state: „Processing...“, „Processed“, „No text“ or „Error“. There is also the free-text field „Additional notes“ (up to 200,000 characters).
• Chatting in a project: The project page has its own input field; below it, the project's latest chats are listed („Recent chats“).
• File an existing chat into a project: You can move an existing conversation from your history onto a project in the sidebar by drag and drop — the chat is then assigned to that project and adopts its instructions and knowledge. The target project highlights as you drag.

3.6 Which model answers? — Brane decides

You do not pick a model in Brane — there is deliberately no model switcher in the chat. The Guardian layer decides per message whether the local model on the box or a cloud model approved by your IT answers; which models are used is configured centrally by your IT. Which model actually produced your specific answer is always visible in the detail view of the security badge (chapter 4.2).

3.7 Tiefgang: making it think harder

Behind the „+“ button you'll find the „Tiefgang“ toggle (brain icon, glows amber when active): „Force deep reasoning for this message“. With it, the model takes noticeably more time to think the next message through — great for tricky analyses, complex texts or hard decisions. If web search is also active, this becomes a more thorough deep search.

Tiefgang applies to that one message only and switches itself off automatically after sending. When the tool area is collapsed, a small amber brain badge shows that Tiefgang is active.

3.8 Voice input: dictate instead of typing

Also behind the „+“ button: the microphone („Start voice input“). One click starts recording (red button with a timer), another click („Stop recording“) ends it. Then „Transcribing...“ appears, and the recognized text is inserted into the input field at the cursor position — it is not sent automatically. So you can dictate, tidy up and then send deliberately.

The first time, the browser asks for microphone permission; if it is denied, the button shows „Microphone blocked“. Possible error messages: „Microphone access denied“, „No microphone found“, „Speech service unavailable“, „Transcription failed“, „No speech detected“. Recordings are capped at 25 MB; if the browser has no microphone support, the button is not shown at all.

4. Reading the security badge

Below every answer sits a small, clickable badge — the security badge. At a glance, it answers the question: where was this message processed? So you can always see whether your request left the box — and if so, in what form.

4.1 The badge levels

Badge	Meaning
Local (shield)	Processed on the box. No sensitive data detected.
Local (PII) (lock)	Personal data detected — therefore deliberately processed locally.
Local (Thinking)	Processed locally, with Tiefgang.
Cloud (cloud)	Answered by an approved cloud model; no sensitive data involved.
Cloud (Thinking)	Cloud, with Tiefgang.
Cloud + PII protection (shield with check)	Answered in the cloud — but the detected personal data was anonymized first. The cloud only ever saw placeholders.
Cloud + PII protection (Thinking)	As above, with Tiefgang.
Cloud (direct)	Routed straight to the cloud due to a special configuration by your IT.
Local (No Key) (amber)	Cloud would be allowed in principle, but no cloud access is configured — Brane answered locally.
Local (Uncertain) (amber)	The analysis of the message was inconclusive — when in doubt, Brane always stays local.
Tiefgang unavailable (amber)	You requested Tiefgang, but it could not be delivered for this message.
Quality reduced (second badge)	High system load — a smaller model is answering temporarily; answer quality may be reduced.

Gray badges are normal states; amber badges are notices, not errors — they show that the system erred on the side of caution. Ground rule: whenever a decision is unclear or a piece of information is unavailable, the message is shown and processed as Local — the safe path is always the default.

4.2 The detail view: why was it decided this way?

Clicking the badge opens the detail view (on a desktop directly under the answer, on a smartphone as a slide-up panel; tinted green for local, blue for cloud processing). From top to bottom:

1. A plain-language reason — for example: „No sensitive data detected.“ · „Personal data detected – processed locally.“ · „Personal data was anonymized before cloud processing.“ · „Personal data blocked for cloud – processed locally.“ · „This conversation stays local.“ · „Mode is set to local only.“ · „Analysis uncertain – processed locally for safety.“ · „Company documents stay on this server.“

2. The active policy — a line like „Cloud preferred · Anonymize“. The first part is your installation's routing mode („Local only“, „Local preferred“, „Cloud preferred“, „Cloud only“), the second is the PII rule („Block“ or „Anonymize“).

3. PII chips — if personal data was detected, a colored chip shows each category. Detected categories include:

   Group	Categories
   Identity & contact	Person, Email, Phone, Address, Date of birth, Organization
   Finance & IDs	IBAN, Credit Card, Tax no., VAT ID, Trade register, Insurance no., SSN, Passport, Reference no.
   Technical & secrets	IP Address, Password, Secret/API Key
   Specially protected (Art. 9 GDPR)	Health Data, Political opinion, Religion, Ethnic origin — deliberately colored in a serious red family
   Catch-all	Sensitive Data

4. „Show what the cloud sees“ — the most important trust feature: if anonymization was applied, this button unfolds the exact text that left the box — with placeholders instead of your data (hide again via „Hide anonymized view“).

5. Warnings — e.g. „High load → temporarily smaller model (response quality may be reduced)“ or, in red, „Security fallback → local“ if the anonymization check did not fully pass and the message was therefore processed locally.

6. Footer — the model that actually answered (with a provider icon). Via the flag icon you can report a routing decision you consider wrong directly to your IT. The report dialog explains the decision again step by step — including your installation's short rule (e.g. „Short rule: Local; cloud only when needed (personal data is anonymized).“) and, if anonymization was applied, the anonymized view.

4.3 Anonymized passages in detail

In the „Show what the cloud sees“ view, every replaced passage is highlighted as a colored chip with a lock icon and a category label (e.g. Person, Email, IBAN; the replacement label reads „Protected“). Hover over a chip to see the technical placeholder (e.g. <EMAIL_1>).

Important to understand: your own message in the chat history always shows the original text. The highlighted view is not your chat — it is what left the building. The answer you read has already been „rehydrated“: the placeholders were replaced with the real data, without the cloud ever having seen it.

4a. Insights: your personal overview

Alongside the badge under each individual answer, there is a second transparency view for the big picture: Insights. You reach it via your avatar → „Insights“. The page („Brane Insights“ — „Your private overview of how Brane keeps you sovereign and productive.“) shows only your own usage — no one else sees your Insights, and you don't see theirs. At the top you switch the time range (e.g. last week, last month).

The overview is organized into zones:

• Safety — „Your data stayed sovereign“: the split of your requests across local, anonymized and cloud, plus a breakdown of which categories of protected data were detected. This is where you see in black and white how much of your work never left the box.
• Models & routing: which models answered your requests. Individual rows expand to show timestamp, model and reason — the same transparency as the badge, but across many messages.
• Productivity — „What you made“: your generated outputs (documents, spreadsheets, images, code, diagrams) and highlights such as your busiest day.
• Levels — „Your Brane journey“: a playful overview of your path with Brane.

Insights is view-only — you configure nothing there. It is where the many small badge decisions come together into one understandable picture.

5. Documents & files

5.1 Getting files into the chat

Three ways lead there:

1. the paperclip button („Add files“) behind the „+“ button,
2. drag & drop straight onto the input field („Drop files here to attach“),
3. pasting images from the clipboard (e.g. screenshots). Pasted text, by contrast, always stays plain text in the input field and never becomes an attachment.

Attached files appear as removable chips in the input field and are uploaded when you send.

Supported formats:

Category	Formats
Images	JPEG, PNG, GIF, WebP, SVG, BMP, ICO
Documents	PDF, Word, Excel, PowerPoint, ODT/ODS/ODP, RTF, EPUB
Code	around 45 common source-code formats (e.g. .py, .js, .ts, .java, .sql)
Data	CSV, TSV, JSON, XML, YAML, TOML, log files
Text & web	TXT, Markdown, HTML, CSS, LaTeX and more
Configuration	common configuration files

Not supported — the app explains why whenever it rejects a file: video files („Video files cannot be processed by AI models“), audio files („Use the voice input button for audio“), executables („Executable files are blocked for security“) and archives („Extract archive contents and upload individual files“).

Limits: at most 100 MB per file (message: „File too large — maximum …“) and 500 uploads per day („Daily upload limit reached“). Oversized or unsuitable files are rejected immediately at selection time, not only when you send.

5.2 What happens to your files

Text is automatically extracted from office documents and PDFs and handed to the model — even scanned PDFs without a text layer are read via optical character recognition (OCR), locally on the box.

5.3 Having documents and spreadsheets created

If the feature is active, Brane creates real files on explicit request: „Turn this into a CSV“, „Summarize this as a Word document“, „Export this as a PDF“. Without a concrete file request, the answer stays normal chat text — tables and code then simply appear in the answer.

• Formats: PDF, Word (.docx), Excel (.xlsx), CSV, Markdown, JSON, TXT, YAML, XML, SVG and common source-code formats. If you don't name a format, you get a PDF; pure tabular data becomes .xlsx.
• During creation you see „Creating document...“ or „Creating spreadsheet...“.
• Multi-step requests work in one go too, e.g. „Research X and create an overview as a PDF“.
• If your IT has stored a corporate design, Word and PDF documents are automatically generated in the company layout (logo, colors, fonts, footer).
• Every file creation and every download is logged in an audit-proof way.

5.4 Artifacts: preview and download

Created files appear as an artifact chip under the answer — with the real file name, format and size (e.g. „PDF · 12.3 KB“). Longer code blocks (from 15 lines) and SVG graphics from normal answers also become artifacts automatically.

• Clicking the chip opens the preview in a side panel: documents with a switchable „Text“/„PDF“ view (the finished PDF is rendered right inside the app), spreadsheets as an Excel-like grid with column headers and row numbers. Very large tables are shown truncated: „Preview truncated to 1,000 rows — download for the full data.“ If no preview is possible: „No preview available — please download.“
• „Download“: for documents you choose the format — Word (.docx), PDF or Markdown (.md); spreadsheets, code and SVG download the file or source directly.
• „Open in new tab“ shows the artifact in its own sandboxed view.

All generated files are delivered over an authenticated connection only — only you can download your files.

6. Using knowledge

6.1 Answers with source citations

With the knowledge base active, Brane can draw on the company documents your IT has stored whenever a question fits. Answers from this knowledge come with evidence:

• Under the answer appears a „Quellen:“ (Sources:) bar with numbered chips [1] [2] …. The colored edge signals reliability: green = high, amber = medium match, dashed = speculative. Hovering shows the title, page („S. n“ — p. n) and match percentage.
• Clicking a chip opens the source view with document title, file name, reliability and a text preview of the cited passage.
• Answers that used web sources (chapter 7.2) additionally show an expandable „Sources“ row with the pages visited.

6.2 The wiki: maintaining knowledge yourself

Part of the knowledge base is the wiki („Knowledge Base“), reachable at the /wiki address of your Brane installation. Here you maintain pages yourself in three areas: „Company“, „Department“ and „Personal“ — „New page“ creates new ones, and tags help with organizing.

The trick: „Wiki pages are automatically included as AI context — the AI then knows your processes, structures and information.“ Whatever you document there, Brane knows in every one of your chats — without you having to explain it each time.

If the knowledge base is not activated, the wiki page shows the notice „Wiki not available“ — „The wiki is part of the knowledge base, which is not enabled for this installation. Contact your administrator if you need this feature activated.“

7. Generating images, web search, more tools

7.1 Generating images

• Behind the „+“ button, activate the „Generate image“ toggle (active: „Image“). The input field switches to „Generate an image of...“ — simply describe what you want to see and send. There is no separate dialog.
• During generation you see „Generating image...“ with an animated canvas that morphs into the finished image.
• Image mode stays active in the chat until an image has been delivered, then switches itself off; switching chats resets it.
• The finished image appears right in the answer; hovering reveals download and a full-screen view. If the image service rephrased your description, the revised version is shown alongside. Generated images, too, are retrievable only by you.
• By the way, Brane also understands casual image requests in normal messages („Show me what X looks like“) — the toggle is just the direct route.

7.2 Web search

The globe toggle „Search“ behind the „+“ button is active by default in every session (blue chip). It allows Brane to pull in current information from the internet when needed; pages used appear under the answer as „Sources“. Combined with Tiefgang, this becomes a thorough deep search.

Important to understand: the toggle is a request, not a guarantee. Whether a search actually happens is decided by your installation's security policy, per message. No search takes place if

• your IT has disabled external search,
• your installation runs in pure local mode (a sealed box — nothing leaves the building),
• external search has been switched off for your department, or
• the current message contains sensitive or personal data — in that case there is categorically no external search, no exceptions.

In these cases Brane simply answers without web results — no error message appears.

7.3 Short video clips

If enabled, Brane generates short video clips of 4, 6 or 8 seconds on explicit request (formats 16:9, 9:16 or 1:1; 720p or 1080p). Generation takes roughly 30 to 180 seconds; you receive the result as a download link in the answer — there is no playback preview in the chat.

8. Your settings

Click your avatar at the top, then „Settings“. The dialog has three sections: „General“, „Connections“ and „Appearance“.

8.1 General

• Profile — your display name and email address (view only; changes are made by your IT).
• „About me“ — your personal standing context: „The AI considers this information in every conversation.“ Example from the input field: „e.g. I am a software developer and prefer short, technical answers.“ After „Save“: „Saved. Will be used in new chats.“ Worth doing — set your role, department and preferred answer style once, and never explain them again.
• „Display language“ — English, Deutsch or Français. „Applies to menus and settings. Chat replies follow the conversation language.“ The choice is stored on your account and follows you to every device.
• „Export my data“ — your GDPR self-service export, in detail in chapter 9.1.
• „Account“ / „Sign out“ — „Log out on this device“. Signing out also deletes all app data cached locally in this device's browser — good to know on shared computers.

8.2 Connections

Here you link your personal work accounts to Brane so the assistant can access, say, your mail or your calendar. Available — if set up by your IT — are Microsoft 365 (Outlook mail, calendar, SharePoint, Teams), Google Workspace (Gmail, Google Calendar, Google Drive), Slack, GitHub, Notion and Jira/Confluence.

• „Connect“ opens the sign-in window of the respective service — you sign in there with your own work account. Afterwards the card shows „Connected“ together with the account address.
• „Disconnect“ removes the link at any time.
• Only services your IT has set up will appear. Otherwise: „No connectors enabled by the administrator yet.“ Further possible services are listed under „More connectors (admin must enable):“.

8.3 Appearance

Under „Theme“ you choose between „System“, „Light“ and „Dark“ — the choice takes effect immediately and applies per browser/device.

9. Your data, your rights

9.1 GDPR self-service export: your data at the press of a button

Under Settings → General → „Export my data“ you download all your data without going through IT: „GDPR Art. 15 — download all chats, messages, feedback and profile data linked to your account as a JSON file.“ One click on „Export my data“ produces the file brane-data-export-<date>.json containing:

Area	Content
Profile	email, creation date, department
Chats	titles, timestamps, flags (e.g. „contains sensitive data“)
Messages	your complete chat history incl. the model used
Feedback	your thumbs ratings and comments
Projects	names, descriptions, timestamps
API keys	provider and timestamp only — the keys themselves are expressly not included

9.2 What is stored — and for how long

Brane stores your conversations so you can find them again later — on the server inside your company, not with a cloud provider. For how long is up to your IT. A notice at the start of every conversation tells you:

• „Chats are automatically deleted after X days.“ — your installation uses an automatic retention period.
• „This chat is not saved. Messages are deleted when you close the page.“ — your installation runs in ephemeral mode.
• No notice — chats are kept until you delete them yourself.

If you are assigned to a department, the same area shows its name — departments can have their own, possibly stricter rules.

Two more points worth knowing:

• Chat titles are cleaned automatically: if the automatically generated title of a conversation contains personal data, the server replaces it with placeholders — so that, for example, the history list gives away no real names.
• AI labelling: Your IT can activate a permanent notice strip above the chat clarifying that you are communicating with an AI assistant and that answers may contain errors (labelling obligation under the EU AI Act).

9.3 What you control yourself

• Delete individual chats — any time via the history (chapter 3.4).
• Export individual chats as Markdown — via the chat menu in the history (chapter 3.4).
• Review your usage — the Insights page (chapter 4a) shows you at any time how much was processed locally, anonymized or in the cloud.
• Export all your data — chapter 9.1.
• Disconnect connections — chapter 8.2.
• For everything else (correcting profile data, password resets, account deletion), your administrator is the right person to ask.

10. When something doesn't work

10.1 Common situations

You see …	What's behind it	What you can do
Badge „Local (Uncertain)“	The content analysis was inconclusive — Brane answered locally to be safe.	Nothing — the system is working as designed.
Badge „Local (No Key)“ persistently	Cloud would be allowed, but no cloud access is configured.	Answers keep coming locally; tell IT if needed.
Badge „Quality reduced“	High load — a smaller model is answering temporarily.	Try again later; repeat important tasks if necessary.
„Tiefgang unavailable“	Deep reasoning could not be delivered for this message.	Send again, or work without Tiefgang.
Your request is refused or answered noticeably cautiously	Brane also screens inputs for hidden manipulation attempts (e.g. instructions „smuggled“ into copied text, documents or web content). Depending on your IT's settings, such requests can be blocked.	Rephrase the request in your own words; drop the suspicious source. If it's a false alarm: tell IT.
No image button in the tool area	Your installation cannot generate images (local-only operation or no cloud access).	That's configuration, not a bug — raise it with IT if needed.
Web search active, but no web results	The policy did not allow the search for this message (e.g. sensitive content or local mode).	Intended behavior — see chapter 7.2.
„Wiki not available“	The knowledge base is not activated on your installation.	„Contact your administrator if you need this feature activated.“
„Unfortunately, I can't create a Word document“ or similar	Document creation is not enabled.	Use the content as chat text; raise it with IT if needed.
Upload rejected	Unsupported format, file over 100 MB („File too large — maximum …“) or „Daily upload limit reached“ (500 files/day).	Check format/size; unpack archives and upload files individually; continue the next day if needed.
„Microphone blocked“ / voice error messages	Browser permission missing or the speech service is unreachable.	Grant microphone permission in the browser settings; otherwise tell IT.
At login: „Ihr Zugang wurde noch nicht freigeschaltet. Bitte wenden Sie sich an Ihren Administrator.“ (Your access has not been approved yet. Please contact your administrator.)	Your access request has not been approved yet.	Ask your administrator to approve you.
No security badge under answers	Your IT has deliberately disabled the display.	Not an error; your IT can answer questions about it.
Persistent display problems	A stale local cache.	Use „Cache & Cookies löschen“ (Clear cache & cookies) on the login page.

Ground rule for anything unexpected: when in doubt, Brane decides „local“. An amber badge or a cloud feature that stayed silent never means data left the building — it means the opposite.

10.2 A feature from this handbook is missing on your installation?

Then it is not (yet) activated there — the knowledge base, document creation, image generation, video clips, connections and SSO are deliberate choices made by your IT. Ask them; this handbook marks the affected sections with „Optional — must be enabled by your IT“.

10.3 Who helps?

1. Your IT / your administrator — first stop for access, feature activation and configuration.
2. The Feedback button at the top right of the app — for requests, praise and bug reports straight from your work context; for individual answers use the thumbs rating, for routing decisions the flag icon in the badge detail view.
3. AI-Z support — if your IT brings in the vendor: support@ai-z-group.com (general: info@ai-z-group.com; security reports: security@ai-z-group.com).

AI-Z GmbH · Königstraße 26, 70173 Stuttgart · info@ai-z-group.com · ai-z-group.com