What the assistant can do
Office files, data, charts, and approved requests — enabled per workspace.
Workspace admins turn capabilities on under workspace collaboration → assistant capabilities. The platform must also enable them. People get outcomes, not a list of switch names.
Common capabilities
| After it is on | What people can do |
|---|---|
| Basics / office files | Read and write text and Markdown; read and create Word, Excel, PowerPoint |
| Data connection | Query a business database with predefined queries; writes can require approval |
| Charts | Create bar, line, and pie charts in the conversation, and place them in office files |
| Approved requests | Call authorized business APIs from a skill |
| Scripts | Process office files under a skill, with per-member authorization |
Admins can also add and manage data connections in conversation with the help assistant or a work agent.
Is asking data safe?
A data connection must be enabled, and people must be authorized. Day-to-day work uses predefined queries. Secrets stay on the server and are not echoed in chat.
Improve by talking
Repeated work can become a skill. Skills and app interfaces support: draft a revision → compare → confirm, so accidental changes are fewer.
Tables, documents, and images can sit on the right of the conversation while you keep changing the request on the left. A work-agent conversation also keeps a topic that states what you are trying to finish.
Related: Messages and agents · Embed in existing systems