Frequently Asked Questions

Your Office AI is an all-in-one, AI-native collaboration platform that combines team chat, video and audio calls, an intelligent AI assistant, knowledge, reusable AI agents, and visual workflow automation — all in a single app. The product name is "Your Office AI"; yoffice.ai is simply the website shorthand.

Unlike traditional productivity suites that bolt AI onto existing tools as an afterthought, Your Office AI is designed around an AI layer that understands your team's context: conversations, uploaded documents, connected integrations, and your running workflows. The result is an assistant that gives accurate, context-specific answers rather than generic ones.

It's available on the web and as native apps for iOS, Android, macOS, Windows, and Linux, and supports six AI providers — OpenAI, Anthropic, Google, Groq, and locally hosted Ollama and Ollama Cloud — for private or cost-controlled use.

Your Office AI runs on all major platforms:

• Web — No installation required. Open qa.qa.app.yoffice.ai in any modern browser (Chrome, Firefox, Safari, Edge).
• iOS — Native iPhone and iPad app.
• Android — Native Android app.
• macOS — Native desktop app for Apple Silicon and Intel.
• Windows — Native desktop app.
• Linux — Native desktop app.

All platforms share the same account and sync state in real time. You can start a conversation on your phone and continue it on your desktop without missing a message.

Yes. Your Office AI has native iOS and Android apps built from the same codebase as the desktop client, so the feature set is consistent. On mobile you can:

• Send and receive messages in channels and DMs
• Start and join audio and video calls
• Chat with the AI assistant and switch models
• Talk to the assistant by voice
• Browse and search Knowledge
• Receive push notifications for mentions, calls, and hails
• Run and approve workflows

The layout adapts to each screen — the home view is assistant-first on phones, while the canvas and workflow builder are richest on tablet and desktop.

Slack and Microsoft Teams are primarily messaging platforms that have added AI on top. Your Office AI is built around AI as a first-class part of every interaction, which leads to several meaningful differences:

• Context-aware AI: The assistant draws on your conversations, documents, connected integrations, and workflow history before responding — answers specific to your team, not generic internet knowledge.
• Model choice: You pick the model per conversation across six providers (OpenAI, Anthropic, Google, Groq, Ollama, Ollama Cloud). Slack and Teams lock you into a single vendor.
• Integrated video: Calls run on LiveKit with live captions and translation, screen sharing, and cross-device beaming — built in, not bolted on.
• Agents & workflows: A visual builder and reusable AI agents orchestrate multi-step work on a LangGraph engine, not just simple message forwarding.

Your Office AI needs an internet connection for most features — chat, video calls, and cloud AI models all require network access.

If you configure a local Ollama instance and select it as your model, AI inference stays on your own hardware rather than going to a third-party provider.

Yes. YOffice is built with a security-first architecture across every layer:

• Encryption in transit: All traffic between your client and YOffice servers travels over TLS (1.2+). Real-time video and audio are encrypted in transit via LiveKit's DTLS-SRTP (transport encryption to the media server).
• Data hosting: Organization data is hosted on EU infrastructure (Germany). Sensitive values such as API keys and access tokens are encrypted at the application layer with AES-256-GCM before being written to the database.
• Tenant isolation: Each organization runs in its own isolated tenant. Database rows are scoped by tenant ID, and every API call is validated against a short-lived, signed tenant JWT. One organization cannot access another's data.
• API key storage: Bring-your-own API keys (OpenAI, Anthropic, Google, Groq) are encrypted before being written to the database and are never returned in plain text after initial entry.
• GDPR & consent: We process data in accordance with GDPR, with consent versioning and an audit log. You can export or delete your data from Settings → Account → Privacy & data.

We conduct regular security audits and follow responsible disclosure practices. To report a vulnerability, email security@yoffice.ai.

Privacy is built into the product, not added on. From Settings → Account → Privacy & data you and your organization can control:

• Consents: Analytics, crashlytics, marketing, and AI-training consents all default to off; the terms, privacy policy, and data-processing consents are captured at sign-up, with versioning and an audit trail.
• EU data residency EU data residency by default.
• AI-training opt-out — your data isn't used to train models unless you opt in.
• Encryption posture Encryption posture surfaced in a dedicated card, plus links to the Privacy Policy, Terms, DPA, Cookies, and Sub-processors.

Yes. Your Office AI supports data-subject requests. From Settings → Account → Privacy & data you can Download my data (a DSAR export of your content).

To delete your account, choose Delete account in the same area. Deletion runs on a 30-day grace period and requires you to type a confirmation and re-enter your password, so an accidental request can be reversed.

Each organization is an isolated tenant, so its data is scoped to that organization and never mixed with another's.

Your Office AI supports six providers at once, and you can switch models per conversation without leaving the chat. The model picker is data-driven, so administrators decide exactly which models appear. Current defaults include:

• Anthropic — e.g. claude-sonnet-4-5. Strong at nuanced reasoning and long documents.
• OpenAI — e.g. gpt-4o-mini. Fast and versatile.
• Google — e.g. gemini-2.0-flash. Multimodal with a large context window.
• Groq — Very low-latency inference.
• Ollama (local) — run models such as llama3.2 on your own hardware; Run models on your own hardware; inference stays on your machine.
• Ollama Cloud — hosted open models such as gpt-oss:120b. Hosted open models.

Administrators control which models are available and set per-org, per-provider spend limits. Models are routed per task type, and every numeric budget has a positive minimum — zero disables a provider rather than making it unlimited.

Yes. Administrators can supply organization API keys for OpenAI, Anthropic, Google, or Groq so usage is billed to your own provider account. Keys are validated and encrypted at rest, and are never returned in plain text after entry.

This is useful for organizations with pre-negotiated pricing, data-residency requirements, or existing provider credits. You can also keep inference entirely off third-party providers with Ollama (local) or Ollama Cloud.

Knowledge extracts text from a broad range of formats:

• Documents: PDF, Word (.docx, .doc, .odt), RTF, plain text, Markdown, email (.eml)
• Spreadsheets: Excel (.xlsx, .xls), OpenDocument (.ods), CSV, TSV — turned into an LLM-friendly structure and a queryable table
• Presentations: PowerPoint (.pptx)
• Code & data: JSON, YAML, XML, HTML, logs

After upload, the text is split into chunks and embedded into a pgvector store, so the assistant can retrieve and cite the most relevant passages — even across many documents. Note that images aren't text-extracted (no OCR). You can also link a website as an AI source, and folders hold up to 1,000 documents with optional watchers that auto-ingest new files.

Workflows automate multi-step tasks without writing code, using a visual node-canvas builder that connects a trigger to a series of nodes (around 28 node types span AI, control flow, actions, and a terminal step). Agents are persistent, reusable AI personas with their own tools, model, and knowledge. Both run on a Python LangGraph engine.

Common use cases include:

• AI digests: On a schedule, summarize recent activity from a connected tool and post the summary to a channel.
• Meeting prep: When a calendar event approaches, generate a briefing from related knowledge and send it out.
• Incident response: When a webhook fires, create a ticket, notify a channel, and kick off follow-up steps.
• Forms: A form-trigger workflow renders an in-app fillable form that starts the automation when submitted.
• Research agents: Run multi-step research that searches the web and your knowledge, then posts a cited report.

Triggers include manual, webhook, schedule (cron), integration events, and folder watchers. See the Workflows & Agents documentation for the full reference.

The integrations catalog includes 60+ tools across categories like communication and email, productivity, development, CRM, storage, analytics, marketing, support, finance, and social. Today, four connect with one-click OAuth: Gmail, Google Calendar, Google Drive, and Google Tasks — the rest of the catalog is marked "Coming soon" until their OAuth credentials are configured.

You can still reach any other tool today by defining your own custom integrations as HTTP actions with your own authentication (API key, bearer token, or basic auth). Administrators set a per-integration spend cap and a monthly call quota (a quota of zero disables the integration). See the Integrations documentation.

Open Settings → Organizations → Members and choose Invite. Enter one or more email addresses and assign a role — admin, editor, or member. Each person receives an email with a join link that lands on a page to accept and join your organization.

New members join with the member role by default; admins can change roles at any time and add people to specific workspaces. See Members & Roles for the full model.

Your Office AI supports sign-in with email and password, or continue with Google. There is no Apple or phone sign-in.

If you use email and password: If you use email and password: on the sign-in page, use the password-reset flow if you've forgotten it. After signing in, you pick an organization to enter — the app then requests a tenant token scoped to that organization.

If you continue with Google: If you continue with Google: your credentials are managed by Google, so there's no separate Your Office AI password to reset.

Yes. Each conversation can run in text, voice, or avatar mode. Voice is routed through the same chat engine as text, so spoken requests have full parity — the same tools, context, and memory. The voice pipeline uses a Deepgram or Google bridge, or a Gemini Live realtime bridge, and an animated avatar (powered by Simli) is available in avatar mode.

See Unified Voice + Chat for how the pipeline is chosen and configured.