Keep Jarvis available when one AI service fails
Use AI models to create a small, approved fallback order and test it before the team depends on it.
Outcome
Chat can use an approved second connection when the preferred AI service is unavailable.
Main feature: AI models
The situation
The finance team uses Jarvis throughout close week. Its preferred AI service had a short outage last month, which left the team unable to finish time-sensitive analysis.
Before you start
An administrator must approve the provider, data terms, cost, model, and credentials for every connection. A fallback is not permission to connect an unreviewed personal account. If this is a new workspace, finish Set up a workspace before building the fallback order.
Create the order
- Open Settings > AI models. Use the Settings guide to confirm which sections are personal and which require an administrator.
- Check the current connection and its health.
- Add the approved second connection using the method offered for that provider, such as a supported subscription or API key.
- Put the preferred model first and the approved fallback second.
- Save and wait for connection work to finish.
- Open a new Chat with Auto selected.
- Run an ordinary, low-risk question and check the reply.
- Review the status again under Settings > General.
Test without creating risk
Use a read-only question for the first test. Do not test continuity by asking for a payment, deletion, or large batch change. If the provider order is changed during a live conversation, begin a fresh test so the result is easy to explain.
Done when
Both approved connections show a usable status, their order is clear, Chat works on Auto, and the team knows who owns provider problems and cost review.
The same question can receive a different answer from a fallback model. Availability does not remove the need to check important figures and drafts.
Use it again
Review the fallback during provider, plan, or policy changes. Remove a connection as soon as the company no longer approves it.