Omni Assistant
Omni is Omnisnia’s built-in AI assistant. It answers questions grounded in your organization’s own data, and it can carry out work on your behalf — but only ever after you confirm. Omni proposes; you confirm; it runs as you.
What Omni is
Omni is a chat assistant that understands your CRM and collaboration data. You can ask it questions in plain language and it answers using your organization’s records and documents, not generic knowledge. Because everything Omni sees is scoped to your current organization , it can only ever read and act on your own tenant’s data.
You can also work with Omni inside a channel : prompt it there and it answers in the channel, grounded in that channel’s conversation and the documents in its scope.
Configuration required
Omni depends on an inference endpoint that your administrator configures for the instance. This is the AI model Omni talks to.
- When an inference endpoint is configured, Omni is available and reports its status as ready.
- When it is not, Omni reports that it is not configured on this instance, and cannot answer questions or propose actions.
For richer, meaning-based search over documents, an administrator can also configure embeddings. Without embeddings, Omni still works and document search falls back to keyword matching. If Omni is unavailable, ask your administrator whether the inference endpoint is set up.
Asking questions grounded in your data
Ask Omni a question about your organization — your pipeline, a client, what’s closing soon, what a document says — and it answers using your live data, citing what it drew on. Because the answer is grounded in your records, it reflects the current state of your organization rather than a generic response.
The confirm-before-execute action model
Omni can do more than answer: it can perform actions. But it never acts on its own. The model is always:
- Propose. You ask Omni to do something. Omni turns your request into a specific, structured action and proposes it, showing you exactly what it will do. Nothing has happened yet.
- Confirm. You review the proposal and confirm it (or cancel it). Only the person who asked can confirm or cancel their own proposed action.
- Execute. On confirmation, Omni runs the action as you — under your own permissions and in your current organization. The same rules that apply when you do something by hand apply here, including the administrator check for administrator-only actions.
This means Omni can never do something you could not do yourself, and never does anything without your explicit go-ahead. A proposed action that you do not confirm simply never runs.
What Omni can do
When you ask, Omni can propose any of these actions:
| Action | What it does |
|---|---|
| Create a client | Add a new client (account) |
| Create a lead | Capture a new lead |
| Convert a lead | Turn a lead into a client, optionally opening a deal |
| Create a task | Open a task, optionally linked to a record |
| Create a product | Add a product to the catalog |
| Create a case | Open a support case |
| Create a channel | Start a new channel |
| Post a message | Post a message to a channel |
| Create a deal | Open a deal |
| Send an email | Send an email to a contact |
| Generate a quote | Produce a quote for a deal |
Administrator-only actions
Two actions can only be proposed and run by an administrator :
- Create a reasoning rule
- Create an automation
If you are not an administrator, Omni will not carry these out for you, just as you could not perform them by hand.
Related pages
- Collaboration — prompt Omni inside a channel
- Workflow Automation — automated rules use the same action catalog
- Getting Started — organizations and tenant isolation
- Account & Organizations — roles and administrator actions