The OfficeDocs AI workspace connects document work with configured model services for chat, writing and knowledge retrieval. Choose the services your team needs and review their output in context.
Last updated
Shared workspace 4 people editing
PDA
Q3 PRODUCT LAUNCH
Make the next release easy to understand.
One shared brief keeps the story, owners and decisions in the same place.
Bring context into every handoff.
✦ Piggy·AI
Live edit · Changes saved just now
The handoff, made visible
One task. Several minds.
An illustrative collaboration scenario: select a contributor below to explore the handoff. The animation is a preview, not a live AI session; available AI features depend on your configuration.
Q3 launch briefShared with 4 collaborators
Q3 launch brief
Piggy & Danco · A shared plan for the next release
The story Piggy·AI
Teams need one clear place to turn ideas into decisions. Bring docs, decisions and deadlines into one shared workspace.
Key message Danco
Lead with the customer’s problem. Keep the promise clear, specific and easy to believe.
Next steps Danco·AI
Finalize the launch story Piggy
Prepare the customer email Danco
Share the review with the team Piggy
Connect the AI capabilities you need
AI configuration separates four services: a base model for chat and writing, an image model for image generation or editing, online search for external references, and embeddings for knowledge retrieval. You do not need to configure every service.
Base models must support the OpenAI Responses protocol and streaming. Knowledge retrieval requires an embedding service with dimensions that match its output. Online search is configured separately from the base model; turning search off does not prevent requests to your configured model or embedding endpoints.
Configure and verify your workspace
In the MDP Operations Platform, select OfficeDocs Suite, open Tenant Management, then choose AI Configuration.
Enter your base model provider, endpoint and API key. Add an available model and set it as the default.
Configure image generation, online search or embeddings only for the features you plan to use. Confirm each service endpoint meets your network and data-handling requirements.
Save, wait 1–2 minutes and refresh the page. Test a conversation and each optional capability you enabled; an enabled switch alone does not confirm a working connection.
Agents work inside the document people are already editing rather than in a separate chat window. Each action appears with the agent identity, a live cursor and an entry in the edit history, so a reviewer can see exactly what changed and take over at any point.
Can we use our own AI model?
Yes. The AI configuration layer points at a model endpoint you choose. Point it at a model running inside your own infrastructure and prompts and document context stay there; point it at a provider you have approved and that data flow becomes a decision you made and can document.
Is document content sent to an AI provider?
Only to the endpoints you configure. The AI capabilities — a base model, an image model, embeddings and an optional online search service — are connected in the operations platform. Online search is a separate, optional service: leave it unconfigured and no outbound retrieval happens at all.
Can the agents use our own knowledge base?
Yes. Knowledge retrieval is one of the AI capabilities connected in the operations platform, alongside the base model, the image model and the optional online search service. It runs through an embeddings service you provide, so the index the agents retrieve from can live inside your own environment rather than being built by a vendor.
Does the free plan include AI?
The free plan for teams of up to five people includes an AI workspace preview. AI assistance across the workspace is part of the paid Team plan, alongside advanced permissions, single sign-on and audit logs.
