Frequently asked questions
What data do AI agents have access to?
Agents can access the data within your Userorbit workspace — feedback submissions, help center articles, changelog entries, announcements, feature requests, and customer information stored in your account. They do not access data from other workspaces, third-party services, or anything outside your Userorbit environment unless you have explicitly configured an integration that shares data.
Can agents publish content without my approval?
Only if you explicitly allow it. In assistant mode, agents never take action without your direct instruction. In supervised mode, every proposal requires human approval. In autonomous mode, agents can publish automatically but only when their confidence score exceeds the threshold you set. You have full control over which mode each agent operates in and what the threshold is.
How do I reset an agent's context?
If an agent seems to be referencing outdated information or producing unexpected results, you can reset its context. In the AI assistant, start a New conversation to clear the chat context. For agent guidance and configuration, go to Settings > Agents, select the agent, and review or update its guidance instructions. Changes to guidance take effect immediately for all subsequent actions.
Are there rate limits on agent usage?
Yes. Agent usage is tied to your plan's AI credit allocation. Each message, proposal, and automated action consumes credits. The amount varies by model — advanced models use more credits per action than fast models. You can monitor your credit usage in Settings > Billing. If you approach your limit, you will receive a notification. Agents pause when credits are exhausted and resume when more are available.
Can I use agents with multiple team members?
Yes. All team members in your workspace can interact with the AI assistant and access the agent inbox. Conversations are private to the team member who created them, but inbox proposals are visible to everyone with the appropriate permissions. Agent configuration (modes, thresholds, guidance) is managed by workspace administrators.
What happens if an agent makes a mistake?
For content that was published by an agent, you can unpublish or edit it from the relevant module (help center, changelog, or announcements) just as you would with any manually created content. For categorization or tagging actions, you can recategorize or re-tag the item. We recommend starting with supervised mode or a high confidence threshold so you can catch issues before they reach customers.
Can I use agents for customer-facing live chat?
Userorbit agents are designed for internal workflows — generating content, categorizing feedback, and automating tasks within your workspace. They do not directly interact with customers in real time. However, they can draft responses that you then send through your existing customer communication channels.
How do agents handle multiple languages?
Agents can generate content in multiple languages if prompted to do so. Include language preferences in your agent guidance — for example, "Always respond in French when the customer's feedback is in French." The quality of non-English output depends on the AI model selected.
Can I undo an agent's action?
Most agent actions can be reversed. Published content can be unpublished or edited. Categories and tags can be changed. Feedback responses that were sent cannot be unsent, which is why we recommend routing customer-facing actions through the inbox for review before sending.
How do I improve agent quality over time?
Three approaches work well together:
- Refine guidance — Update your agent guidance with more specific instructions based on patterns you notice in proposals.
- Review rejections — Look at what you reject and why. This reveals gaps in the agent's understanding that guidance can address.
- Adjust thresholds — As quality improves, gradually lower confidence thresholds to give agents more autonomy on tasks they handle well.
Where can I get help?
If you have questions not covered here, reach out to our support team through the help widget in your workspace or email us directly. We are happy to help you get the most out of AI agents.